Upload folder using huggingface_hub

images/0909.3691v1/gt.json

@@ -0,0 +1,30 @@
+[
+  {
+    "id": 0,
+    "page": 3,
+    "bounding_box": [
+      51.2728328704834,
+      82.6710205078125,
+      294.9151560465495,
+      151.70098876953125
+    ],
+    "latex_content": "\\begin{table}[!t]\n\\caption{Genotype-Phenotype Connections in Genetic Disorders}\n\\centering\n\\begin{tabular}{ | c || p{2.2cm} || p{1.4cm} || p{2.2cm} | }%\n\\hline\n\\bf{Alleles} & \\bf{Genotype} & \\bf{Dominant Disorder}  & \\bf{Recessive Disorder} \\\\[1ex]\n\\hline\n&&&\\\\\n\\bf{AA} & Homozygous WT & Normal    & Normal   \\\\[1ex]\n\\bf{Aa} & Heterozygous        & Affected  & Carrier   \\\\[1ex]\n\\bf{aa} & Homozygous mut. & Affected  & Affected  \\\\[1ex]\n\\hline\n\\multicolumn{3}{l} {A  - normal allele,  a  - mutant allele}\n\\end{tabular}\n\\label{table_genotype_phentype}\n\\end{table}",
+    "extracted_content": [
+      [
+        "Alleles",
+        "Genotype",
+        "Dominant\nDisorder",
+        "Recessive\nDisorder"
+      ],
+      [
+        "AA\nAa\naa",
+        "Homozygous WT\nHeterozygous\nHomozygous mut.",
+        "Normal\nAffected\nAffected",
+        "Normal\nCarrier\nAffected"
+      ]
+    ],
+    "similarity_score": 0.6548223350253807,
+    "table_image": "images/0909.3691v1/table_0.png",
+    "page_image": "pages/0909.3691v1/page_3.png"
+  }
+]

images/1005.2197v1/gt.json

@@ -0,0 +1,42 @@
+[
+  {
+    "id": 0,
+    "page": 21,
+    "bounding_box": [
+      230.6191234588623,
+      276.37799072265625,
+      380.6290054321289,
+      364.84600830078125
+    ],
+    "latex_content": "\\begin{table}[htpb]\n  \\caption[EEG Results]{\\emph{EEG Analysis with Incomplete Data:} The\n    similarity between the columns of $\\M{A}, \\M{B}, \\M{C}$ and the\n    columns of factor matrices extracted by CP-WOPT. The similarity is\n    measured in terms of FMS defined in \\Eqn{FMS}.}\n  \\label{tab:EEG}\n  \\begin{center}\n  \\begin{tabular}{|c | c |  }\n    \\hline\n     Number of  &  CP-WOPT \\\\\n     Missing Channels &  \\\\ \\hline \\hline\n    $1$  &  $0.9959$ \\\\ \\hline\n    $10$ &  $0.9780$ \\\\ \\hline\n    $20$ &  $0.9478$ \\\\ \\hline\n    $30$ &  $0.8949$ \\\\ \\hline\n    $40$ &  $0.6459$ \\\\ \\hline\n  \\end{tabular}\n  \\end{center}\n\\end{table}",
+    "extracted_content": [
+      [
+        "Number of\nMissing Channels",
+        "CP-WOPT"
+      ],
+      [
+        "1",
+        "0.9959"
+      ],
+      [
+        "10",
+        "0.9780"
+      ],
+      [
+        "20",
+        "0.9478"
+      ],
+      [
+        "30",
+        "0.8949"
+      ],
+      [
+        "40",
+        "0.6459"
+      ]
+    ],
+    "similarity_score": 0.9036144578313253,
+    "table_image": "images/1005.2197v1/table_0.png",
+    "page_image": "pages/1005.2197v1/page_21.png"
+  }
+]

images/1012.0269v1/gt.json

@@ -0,0 +1,50 @@
+[
+  {
+    "id": 0,
+    "page": 9,
+    "bounding_box": [
+      82.46077897813585,
+      109.06103515625,
+      516.9070095486111,
+      410.3310241699219
+    ],
+    "latex_content": "\\begin{table}\n\\centering\n\\begin{tabular}{| l |p{8cm}|}\n\\hline\n\\multicolumn{1}{|c|}{\\R function} & \\multicolumn{1}{c|}{Description}\\\\\n\\hline\n\\code{f.analyzeFMRI.gui()} & Starts an \\R/TclTk based GUI to explore, using the \\pkg{AnalyzeFMRI} package functions, an fMRI data set stored in ANALYZE format. \\tabularnewline\n\\hline\n\\code{f.icast.fmri.gui()} & The GUI provides a quick and easy to use interface for applying spatial or temporal ICA to fMRI data sets in NIFTI  format.\\tabularnewline\n\\hline\n\\code{f.plot.volume.gui()} & TclTk GUI to display functional or structural MR images.  This GUI is useful for instance to display the results performed with \\code{f.icast.fmri.gui()}.\\tabularnewline   \n\\hline\n\\code{f.read.header(file)} & Reads ANALYZE or NIFTI (\\texttt{.hdr} or \\texttt{.nii}) header file. The format type is automatically detected by first reading the  magic field. \\tabularnewline\n\\hline\n\\code{f.read.volume(file)} & Reads ANALYZE or NIFTI image file and puts it into an array. Automatic detection of the format type. \\tabularnewline\n\\hline\n \\code{f.write.analyze(mat,file,...,)} & Stores the data in ANALYZE format: creation of the corresponding  \\texttt{.img/.hdr} pair of files.\n \\tabularnewline   \n\\hline\n \\code{f.write.nifti(mat,file,size,...)}   & Stores the data in NIFTI format: creation of the corresponding  \\texttt{.img/.hdr} pair of files or single \\texttt{.nii} file.\n  \\tabularnewline   \n  \\hline \n\\end{tabular}\n\\caption{Seven main functions of our package with their description.\\label{table1}}\n\\end{table}",
+    "extracted_content": [
+      [
+        "R function",
+        "Description"
+      ],
+      [
+        "f.analyzeFMRI.gui()",
+        "Starts an R/TclTk based GUI to explore, using\nthe AnalyzeFMRI package functions, an fMRI\ndata set stored in ANALYZE format."
+      ],
+      [
+        "f.icast.fmri.gui()",
+        "The GUI provides a quick and easy to use in-\nterface for applying spatial or temporal ICA to\nfMRI data sets in NIFTI format."
+      ],
+      [
+        "f.plot.volume.gui()",
+        "TclTk GUI to display functional or struc-\ntural MR images. This GUI is useful for in-\nstance to display the results performed with\nf.icast.fmri.gui()."
+      ],
+      [
+        "f.read.header(file)",
+        "Reads ANALYZE or NIFTI (.hdr or .nii)\nheader file. The format type is automatically\ndetected by first reading the magic field."
+      ],
+      [
+        "f.read.volume(file)",
+        "Reads ANALYZE or NIFTI image file and puts\nit into an array. Automatic detection of the for-\nmat type."
+      ],
+      [
+        "f.write.analyze(mat,file,...,)",
+        "Stores the data in ANALYZE format: creation\nof the corresponding .img/.hdr pair of files."
+      ],
+      [
+        "f.write.nifti(mat,file,size,...)",
+        "Stores the data in NIFTI format: creation of the\ncorresponding .img/.hdr pair of files or single\n.nii file."
+      ]
+    ],
+    "similarity_score": 0.9058663028649386,
+    "table_image": "images/1012.0269v1/table_0.png",
+    "page_image": "pages/1012.0269v1/page_9.png"
+  }
+]

images/1107.1445v1/gt.json

@@ -0,0 +1,98 @@
+[
+  {
+    "id": 0,
+    "page": 35,
+    "bounding_box": [
+      189.86535862513952,
+      317.08099365234375,
+      419.6570129394531,
+      482.8590087890625
+    ],
+    "latex_content": "\\begin{table}[htbp]\n  \\centering\n  \\begin{tabular}{|c|c|}\n    \\hline\n    {\\bf Strategy} & {\\bf Sequence} \\\\\n    \\hline\n    \\multicolumn{2}{|c|}{Repeated measurements not allowed} \\\\\n    \\hline\n    ME Sampling & $10~~1~~9~~2~~3~~8~~5~~7~~4~~6$ \\\\\n    \\hline\n    IM Sampling & $~1~10~~9~~2~~3~~8~~7~~4~~6~~5$ \\\\\n    \\hline\n    RND Sampling & $~6~10~~3~~1~~5~~9~~8~~2~~4~~7$ \\\\\n    \\hline\n    \\multicolumn{2}{|c|}{Repeated measurements allowed} \\\\\n    \\hline\n    ME Sampling & $10~~1~10~~1~~1~10~10~~1~~1~~1$ \\\\\n    \\hline\n    IM Sampling & $~1~10~~1~10~10~~1~10~10~~1~~1$ \\\\\n    \\hline\n  \\end{tabular}\n  \\caption{Results for example $1$, the additive noise case. Design sequences yielded by the three strategies for one particular Monte Carlo run}\n  \\label{tab:ex1_add_seq}\n\\end{table}",
+    "extracted_content": [
+      [
+        "Strategy",
+        "Sequence"
+      ],
+      [
+        "Repeated measurements not allowed",
+        null
+      ],
+      [
+        "ME Sampling",
+        "10 1 9 2 3 8 5 7 4 6"
+      ],
+      [
+        "IM Sampling",
+        "1 10 9 2 3 8 7 4 6 5"
+      ],
+      [
+        "RND Sampling",
+        "6 10 3 1 5 9 8 2 4 7"
+      ],
+      [
+        "Repeated measurements allowed",
+        null
+      ],
+      [
+        "ME Sampling",
+        "10 1 10 1 1 10 10 1 1 1"
+      ],
+      [
+        "IM Sampling",
+        "1 10 1 10 10 1 10 10 1 1"
+      ]
+    ],
+    "similarity_score": 0.49162011173184356,
+    "table_image": "images/1107.1445v1/table_0.png",
+    "page_image": "pages/1107.1445v1/page_35.png"
+  },
+  {
+    "id": 1,
+    "page": 36,
+    "bounding_box": [
+      191.50399780273438,
+      149.093994140625,
+      418.74798583984375,
+      314.87298583984375
+    ],
+    "latex_content": "\\begin{table}[htbp]\n  \\centering\n  \\begin{tabular}{|c|c|}\n    \\hline\n    {\\bf Strategy} & {\\bf Sequence} \\\\\n    \\hline\n    \\multicolumn{2}{|c|}{Repeated measurements not allowed} \\\\\n    \\hline\n    ME Sampling & $~5~~6~~7~~4~~1~~3~~8~~9~~2~10$ \\\\\n    \\hline\n    IM Sampling & $~1~10~~9~~7~~2~~8~~4~~3~~5~~6$ \\\\\n    \\hline\n    RND Sampling & $~6~~1~~8~~5~10~~9~~4~~2~~3~~7$ \\\\\n    \\hline\n    \\multicolumn{2}{|c|}{Repeated measurements allowed} \\\\\n    \\hline\n    ME Sampling & $~5~~6~~6~~5~~5~~5~~6~~5~~6~~5$ \\\\\n    \\hline\n    IM Sampling & $~1~10~~1~10~10~~1~~1~~1~~1~10$ \\\\\n    \\hline\n  \\end{tabular}\n  \\caption{Results for example $1$, the multiplicative noise case. Design sequences yielded by the three strategies for one particular Monte Carlo run}\n  \\label{tab:ex1_mult_seq}\n\\end{table}",
+    "extracted_content": [
+      [
+        "Strategy",
+        "Sequence"
+      ],
+      [
+        "Repeated measurements not allowed",
+        null
+      ],
+      [
+        "ME Sampling",
+        "5 6 7 4 1 3 8 9 2 10"
+      ],
+      [
+        "IM Sampling",
+        "1 10 9 7 2 8 4 3 5 6"
+      ],
+      [
+        "RND Sampling",
+        "6 1 8 5 10 9 4 2 3 7"
+      ],
+      [
+        "Repeated measurements allowed",
+        null
+      ],
+      [
+        "ME Sampling",
+        "5 6 6 5 5 5 6 5 6 5"
+      ],
+      [
+        "IM Sampling",
+        "1 10 1 10 10 1 1 1 1 10"
+      ]
+    ],
+    "similarity_score": 0.47075208913649025,
+    "table_image": "images/1107.1445v1/table_1.png",
+    "page_image": "pages/1107.1445v1/page_36.png"
+  }
+]

images/1109.6846v2/gt.json

@@ -0,0 +1,48 @@
+[
+  {
+    "id": 0,
+    "page": 12,
+    "bounding_box": [
+      151.30999755859375,
+      85.44000244140625,
+      460.69014630998885,
+      171.51702880859375
+    ],
+    "latex_content": "\\begin{table}\n\\label{table:sham}\n\\begin{center}\n\\begin{tabular}{|c||c|c|}  \\hline\nobserved degree &  \\# predicted ($E[N_{\\delta,\\rho^*}]$) & \\# actual ($N_{\\delta,\\rho^*}$)  \\\\ \\hline\n$d_i\\geq \\delta =1$ & 8531  & 8492 \\\\ \\hline\n$d_i\\geq \\delta =2$ & 1697 & 1635 \\\\ \\hline\n$d_i\\geq \\delta =3$ & 234  & 229 \\\\ \\hline\n$d_i\\geq \\delta =4$ & 24  & 28 \\\\ \\hline\n$d_i\\geq \\delta =5$ & 2 & 4 \\\\ \\hline\n\\end{tabular}\n\\caption{Fidelity of the predicted (mean) number of false positives and\n  the observed number of false positives in the realization of the sham\nNKI dataset experiment shown in Fig. \\ref{fig:pv1}}.\n\\end{center}\n\\end{table}",
+    "extracted_content": [
+      [
+        "observed degree",
+        "# predicted (E[N ])\n\u03b4,\u03c1\u2217",
+        "# actual (N )\n\u03b4,\u03c1\u2217"
+      ],
+      [
+        "d \u03b4 = 1\ni \u2265",
+        "8531",
+        "8492"
+      ],
+      [
+        "d \u03b4 = 2\ni \u2265",
+        "1697",
+        "1635"
+      ],
+      [
+        "d \u03b4 = 3\ni \u2265",
+        "234",
+        "229"
+      ],
+      [
+        "d \u03b4 = 4\ni \u2265",
+        "24",
+        "28"
+      ],
+      [
+        "d \u03b4 = 5\ni \u2265",
+        "2",
+        "4"
+      ]
+    ],
+    "similarity_score": 0.44315992292870904,
+    "table_image": "images/1109.6846v2/table_0.png",
+    "page_image": "pages/1109.6846v2/page_12.png"
+  }
+]

images/1201.1277v1/gt.json

@@ -0,0 +1,58 @@
+[
+  {
+    "id": 0,
+    "page": 17,
+    "bounding_box": [
+      173.84100341796875,
+      233.2869873046875,
+      439.0249938964844,
+      342.8760070800781
+    ],
+    "latex_content": "\\begin{table}\n\\centering\n\\begin{tabular}{| l || c | c | c || c | c |} \\hline\n\\multicolumn{1}{| c || }{} & \\multicolumn{3}{c || }{Static Match Rate} & \\multicolumn{2}{c | }{Runtime Precise Rate}\\\\\n\\hline\nBenchmark           & \\hspace{1mm} Region \\hspace{1mm} & \\hspace{1mm} Shape \\hspace{1mm} & Injectivity & \\hspace{1mm} Shape \\hspace{1mm} & Injectivity\\\\\n\\hline\n\\bench{power}        & 100\\%  & 100\\%  & 100\\%     & 100\\% & 100\\%  \\\\\n\\bench{bh}           & 100\\%  & 90\\%   & 87\\%      & 100\\% & 100\\%  \\\\\n\\hline\n\\bench{db}           & 100\\%  & 100\\%  & 81\\%      & 100\\% & 100\\%  \\\\\n\\bench{raytracer}    & 80\\%   & 85\\%   & 83\\%      & 89\\%  & 98\\%   \\\\\n\\hline\n\\bench{luindex}      & 95\\%   & 95\\%   & 82\\%      & 100\\% & 91\\%   \\\\\n\\bench{lusearch}     & 93\\%   & 90\\%   & 84\\%      & 96\\%  & 89\\%   \\\\\n\\bench{runabs}       & 97\\%   & 98\\%   & 87\\%      & 94\\%  & 90\\%   \\\\\n\\hline\n\\end{tabular}\n\\vspace{3mm}\n\\caption{Static Match is percentage of each property correctly predicted by the static analysis \nwhen compared to \\emph{perfect analysis}. Runtime Precise is the percentage of properties that \nthe \\emph{perfect analysis} captures precisely.}\n\\label{tab:precision}\n\\end{table}",
+    "extracted_content": [
+      [
+        "",
+        "Static Match Rate",
+        null,
+        null,
+        "Runtime Precise Rate",
+        null
+      ],
+      [
+        "Benchmark",
+        "Region",
+        "Shape",
+        "Injectivity",
+        "Shape",
+        "Injectivity"
+      ],
+      [
+        "power\nbh",
+        "100%\n100%",
+        "100%\n90%",
+        "100%\n87%",
+        "100%\n100%",
+        "100%\n100%"
+      ],
+      [
+        "db\nraytracer",
+        "100%\n80%",
+        "100%\n85%",
+        "81%\n83%",
+        "100%\n89%",
+        "100%\n98%"
+      ],
+      [
+        "luindex\nlusearch\nrunabs",
+        "95%\n93%\n97%",
+        "95%\n90%\n98%",
+        "82%\n84%\n87%",
+        "100%\n96%\n94%",
+        "91%\n89%\n90%"
+      ]
+    ],
+    "similarity_score": 0.38287752675386444,
+    "table_image": "images/1201.1277v1/table_0.png",
+    "page_image": "pages/1201.1277v1/page_17.png"
+  }
+]

images/1211.6177v6/gt.json

@@ -0,0 +1,67 @@
+[
+  {
+    "id": 0,
+    "page": 9,
+    "bounding_box": [
+      63.57499885559082,
+      123.8790283203125,
+      495.898868560791,
+      257.27899169921875
+    ],
+    "latex_content": "\\begin{table}\n\\centering \n\\begin{tabular}{|m{0.12\\textwidth}|m{0.2\\textwidth}|m{0.11\\textwidth}|m{0.2\\textwidth}|m{0.2\\textwidth}|}\n\\hline\n\\textbf{Identifier index}& \\textbf{Identifier (name of annotation)}&\\textbf{Number of regions}\n&\\textbf{Hierarchical?}& \\textbf{Number of annotated voxels}\\\\ \\hline\n1 & {\\ttfamily{\u2019standard\u2019}} & 209 & Yes &  49,742\\\\ \\hline\n2 & {\\ttfamily{\u2019cortex\u2019}} & 40 &  Yes & 11,862\\\\ \\hline\n3 & {\\ttfamily{\u2019standard+cortex\u2019}} & 242  & Yes &  49,742\\\\ \\hline\n4 & {\\ttfamily{\u2019fine\u2019}} & 94 &  No  & 22,882\\\\ \\hline \n5 & {\\ttfamily{\u2019big12\u2019}} & 13 &  No &  25,155 \\\\ \\hline\n6 & {\\ttfamily{\u2019cortexLayers\u2019}} & 8 & No & 11,862 \\\\ \\hline\n\\end{tabular}\n\\caption{Systems of annotations of the adult mouse brain in the digital version\nof the Allen Reference Atlas, at a resolution of 200 microns.}\n\\label{annotationSystems}\n\\end{table}",
+    "extracted_content": [
+      [
+        "Identifier\nindex",
+        "Identifier\n(name of anno-\ntation)",
+        "Number\nof re-\ngions",
+        "Hierarchical?",
+        "Number of an-\nnotated voxels"
+      ],
+      [
+        "1",
+        "standard",
+        "209",
+        "Yes",
+        "49,742"
+      ],
+      [
+        "2",
+        "cortex",
+        "40",
+        "Yes",
+        "11,862"
+      ],
+      [
+        "3",
+        "standard+cortex",
+        "242",
+        "Yes",
+        "49,742"
+      ],
+      [
+        "4",
+        "fine",
+        "94",
+        "No",
+        "22,882"
+      ],
+      [
+        "5",
+        "big12",
+        "13",
+        "No",
+        "25,155"
+      ],
+      [
+        "6",
+        "cortexLayers",
+        "8",
+        "No",
+        "11,862"
+      ]
+    ],
+    "similarity_score": 0.6530014641288433,
+    "table_image": "images/1211.6177v6/table_0.png",
+    "page_image": "pages/1211.6177v6/page_9.png"
+  }
+]

images/1303.4767v1/gt.json

@@ -0,0 +1,132 @@
+[
+  {
+    "id": 0,
+    "page": 7,
+    "bounding_box": [
+      125.7979965209961,
+      137.85302734375,
+      488.1130065917969,
+      251.427001953125
+    ],
+    "latex_content": "\\begin{table}[htbp]\n\\caption{Summary of cell features. }\n\\centering\n\\begin{tabular}{| c | p{2cm} | p{7.6cm} | c |}\n\\hline\n& Categories    &  Details                                  & \\# of Fea. \\\\\n\\hline\\hline\n1  & Intensity &  Average, Std., Average $log_{10}$, Minimum, Maximum, the 25\\%, 50\\%, 75\\% quantiles of cell pixel intensity  \n&8 \\\\\n\\hline \n2  &  Shape \\& Size  & Perimeter, Area, Non-convexity, Length-Width Ratio, Radius Std.\n& 5\\\\\n\\hline\n3  &  Local Density  & Cell densities in 5 square moving windows with different sizes  & 5\\\\\n\\hline\n4  &  Cell Orient.  & Cell angle, Angle difference with nearest neighbors, The 25\\%, 50\\%, 75\\% quantiles of angle differences in 4 square moving windows with different sizes  \n&14\\\\\n\\hline\n\\end{tabular}\n\\label{TAB:summaryCellFea}\n\\end{table}",
+    "extracted_content": [
+      [
+        "",
+        "Categories",
+        "Details",
+        "# of Fea."
+      ],
+      [
+        "1",
+        "Intensity",
+        "Average, Std., Average log , Minimum, Maximum,\n10\nthe 25%, 50%, 75% quantiles of cell pixel intensity",
+        "8"
+      ],
+      [
+        "2",
+        "Shape & Size",
+        "Perimeter, Area, Non-convexity, Length-Width Ratio,\nRadius Std.",
+        "5"
+      ],
+      [
+        "3",
+        "Local Density",
+        "Cell densities in 5 square moving windows with differ-\nent sizes",
+        "5"
+      ],
+      [
+        "4",
+        "Cell Orient.",
+        "Cell angle, Angle difference with nearest neighbors,\nThe 25%, 50%, 75% quantiles of angle differences in 4\nsquare moving windows with different sizes",
+        "14"
+      ]
+    ],
+    "similarity_score": 0.8133472367049009,
+    "table_image": "images/1303.4767v1/table_0.png",
+    "page_image": "pages/1303.4767v1/page_7.png"
+  },
+  {
+    "id": 1,
+    "page": 7,
+    "bounding_box": [
+      125.7979965209961,
+      288.5880126953125,
+      488.1130065917969,
+      335.6109924316406
+    ],
+    "latex_content": "\\begin{table}[htbp]\n\\caption{Summary of additional entire-well features.  }\n\\centering\n\\begin{threeparttable} \n\\begin{tabular}{| c | p{2cm} | p{7.6cm} | c |}\n\\hline\n& Categories    &  Details                                  & \\# of Fea. \\\\\n\\hline\\hline\n1  & Cell Number &     Number of identified cells in an image    &1 \\\\\n\\hline\n2  &  Cell Gap  &  Summaries* of gap intensity \n & 6\\\\\n \\cline{3-4}\n& & Summaries* \nof the size of circular gaps**& 6\\\\\n\\hline\n\\end{tabular}\n\\begin{tablenotes}\n  \\item[*] Standard deviation, min., max. and the 25\\%, 50\\%, 75\\% quantiles are used as summaries.\n  \\item[**] These features are extracted by performing distance transformation \\cite{distImage} on the IPLab segmented image. Statistical summaries of the intensity of the resulting distance image are used as a description of the size of the circular gaps among cells.\n \\end{tablenotes}\n\\end{threeparttable}\n\\label{TAB:summaryWellFea}\n\\end{table}",
+    "extracted_content": [
+      [
+        "",
+        "Categories",
+        "Details",
+        "# of Fea."
+      ],
+      [
+        "1",
+        "Cell Number",
+        "Number of identified cells in an image",
+        "1"
+      ],
+      [
+        "2",
+        "Cell Gap",
+        "Summaries* of gap intensity",
+        "6"
+      ],
+      [
+        null,
+        null,
+        "Summaries* of the size of circular gaps**",
+        "6"
+      ]
+    ],
+    "similarity_score": 0.46301369863013697,
+    "table_image": "images/1303.4767v1/table_1.png",
+    "page_image": "pages/1303.4767v1/page_7.png"
+  },
+  {
+    "id": 2,
+    "page": 19,
+    "bounding_box": [
+      125.7979965209961,
+      283.0880126953125,
+      502.77099609375,
+      332.37298583984375
+    ],
+    "latex_content": "\\begin{table}[htbp]\n\\caption{Simulation results. }\n\\begin{threeparttable}\n\\begin{tabular}{| p{2.7cm}  |p{1.8cm} | p{1.8cm} | p{2.5cm} | p{2.39cm}|}\n\\hline\nData Objects & Wells-Alone   &  Wells-Alone    & Cell-Well Unions  & Cell-Well Unions \\\\\n\\hline\nCell Analyses & Not done & Std$^{[1]}$ & PCA \\& Std$^{[1]}$ & PLS \\& Std$^{[1]}$ \\\\\n\\hline\\hline\n Uncertainty$^{[2]}$ &   $1.414 \\pm 0.051$ &   $1.390 \\pm 0.055$ & $0.471 \\pm 0.088$ &$0.464 \\pm 0.078$  \\\\\n\\hline\nDWD Error Rate$^{[2]}$ &  $0.212 \\pm 0.011$ & $0.132 \\pm 0.009$  & $0.105 \\pm 0.009$ & $0.104 \\pm 0.009$\\\\\n\\hline\n\\end{tabular}\n\\begin{tablenotes}\n  \\item[1]   Standardize the cell data (or the PC/PLS scores) for each well by their standard deviations.\n  \\item[2]   The 95\\% confidence intervals from 500 simulations are shown.\n   \\end{tablenotes}\n\\end{threeparttable}\n\\label{TAB:simulations}\n\\end{table}",
+    "extracted_content": [
+      [
+        "Data Objects",
+        "Wells-Alone",
+        "Wells-Alone",
+        "Cell-Well Unions",
+        "Cell-Well Unions"
+      ],
+      [
+        "Cell Analyses",
+        "Not done",
+        "Std[1]",
+        "PCA & Std[1]",
+        "PLS & Std[1]"
+      ],
+      [
+        "Uncertainty[2]",
+        "1.414\u00b10.051",
+        "1.390\u00b10.055",
+        "0.471 \u00b1 0.088",
+        "0.464 \u00b1 0.078"
+      ],
+      [
+        "DWD Error Rate[2]",
+        "0.212\u00b10.011",
+        "0.132\u00b10.009",
+        "0.105 \u00b1 0.009",
+        "0.104 \u00b1 0.009"
+      ]
+    ],
+    "similarity_score": 0.5994550408719346,
+    "table_image": "images/1303.4767v1/table_2.png",
+    "page_image": "pages/1303.4767v1/page_19.png"
+  }
+]

images/1303.4986v1/gt.json

@@ -0,0 +1,28 @@
+[
+  {
+    "id": 0,
+    "page": 13,
+    "bounding_box": [
+      202.28900146484375,
+      559.405029296875,
+      408.9580078125,
+      631.9329833984375
+    ],
+    "latex_content": "\\begin{table}[t!] %\\scriptsize\n%\\begin{center}\n%\\begin{tabular}{|l|c|r|}\n%%\\hline\n%%  &   \\small{Work} &  \\small{Leisure}  &  \\small{Coauthor}  & \\small{Lunch} &  \\small{FB} \\\\\n%\\hline\n% $\\mathcal{R}$ & \\small{Covering combination of layers} & \\small{Prob.} \\\\\n%\\hline\n% Coauthor  & Work, Leisure, FB  & 0.95 \\\\\n% Coauthor  & Work, Leisure & 0.90 \\\\\n% Coauthor  & Work & 0.86 \\\\\n% Leisure &  Work, Coauthor, Lunch, FB   & 0.89 \\\\\n% Leisure &  Work, Lunch   & 0.78 \\\\\n% Lunch &  Work, Leisure, Coauthor, FB & 0.70\\\\\n% Work &  Leisure, Coauthor, Lunch, FB & 0.66\\\\  \n% FB &  Work, Leisure, Coauthor, Lunch & 0.64\\\\  \n%\\hline\n%\\end{tabular}\n%\\caption{Best covering combinations of layers for each single network. This was computed as conditional probability that there is an edge in the combination of layers in case that there is an edge in the single network} \n%\\label{tabProb}\n%\\end{center}\n%\\end{table}",
+    "extracted_content": [
+      [
+        "R",
+        "Similar combination of layers",
+        "Jaccard"
+      ],
+      [
+        "Lunch\nWork\nLeisure\nFB\nCoauthor",
+        "Work, Leisure\nCoauthor, Lunch\nCoauthor, Lunch\nWork, Leisure, Lunch\nLeisure, FB",
+        "0.39\n0.36\n0.27\n0.23\n0.07"
+      ]
+    ],
+    "similarity_score": 0.3800277392510402,
+    "table_image": "images/1303.4986v1/table_0.png",
+    "page_image": "pages/1303.4986v1/page_13.png"
+  }
+]

images/1304.6146v1/gt.json

@@ -0,0 +1,84 @@
+[
+  {
+    "id": 0,
+    "page": 11,
+    "bounding_box": [
+      311.9729919433594,
+      77.24798583984375,
+      566.1939697265625,
+      140.80902099609375
+    ],
+    "latex_content": "\\begin{table}\n\\mycaption{\\label{tbl:ss_controller_comparison} Results from 2420 trials in software simulation.}\n\\begin{center}\n\\begin{tabular} {|l|c|c|c|c|c|}\n\\hline\n& Estimated & MPC (up to   & MPC & Baseline   \\\\  %& BiRRT  \\\\\n& Optimal   & 6 Reaches) & (Single Reach)  & Controller \\\\  %& with Margin \\\\\n\\hline\nSuccess rate    & 98.2\\% & 91.1\\% & 78.6\\% & 30.5\\% \\\\  % & 79.1\\% \\\\\nAvg. max. &    \\multirow{2}{*}{-}   & \\multirow{2}{*}{20.1N}  & \\multirow{2}{*}{13.3N}\n& \\multirow{2}{*}{72.0N}  \\\\  % & - \\\\\ncontact force        & & & & \\\\\nAvg. contact & \\multirow{2}{*}{-}   & \\multirow{2}{*}{3.76N}  & \\multirow{2}{*}{5.9N}\n& \\multirow{2}{*}{28.6N}  \\\\  % & - \\\\\nforce        & & & & \\\\\n\\hline\n\\end{tabular}\n\\end{center}\n\\vspace{-0.3cm}\n\\end{table}",
+    "extracted_content": [
+      [
+        "",
+        "Estimated\nOptimal",
+        "MPC (up to\n6 Reaches)",
+        "MPC\n(Single Reach)",
+        "Baseline\nController"
+      ],
+      [
+        "Success rate\nAvg. max.\ncontact force\nAvg. contact\nforce",
+        "98.2%\n-\n-",
+        "91.1%\n20.1N\n3.76N",
+        "78.6%\n13.3N\n5.9N",
+        "30.5%\n72.0N\n28.6N"
+      ]
+    ],
+    "similarity_score": 0.5863636363636363,
+    "table_image": "images/1304.6146v1/table_0.png",
+    "page_image": "pages/1304.6146v1/page_11.png"
+  },
+  {
+    "id": 1,
+    "page": 13,
+    "bounding_box": [
+      75.79900360107422,
+      256.81201171875,
+      273.1780090332031,
+      303.6919860839844
+    ],
+    "latex_content": "\\begin{table}[t!]\n\\mycaption{Model predictive controller vs the\nbaseline controller in the hardware-in-the-loop testbed.\n\\label{tbl:hil_comparison}}\n\\begin{center}\n\\begin{tabular} {|l|c|c|}\n\\hline\n& MPC &\nBaseline Controller \\\\\n\\hline\nSuccess rate & 5/5 & 3/5 \\\\\nAvg. max. contact force & 5.6N & 17.7N \\\\\nAvg. contact force above & \\multirow{2}{*}{5.5N} &\n\\multirow{2}{*}{14.3N} \\\\\n$f_{c_i}^{thresh}$ (5N) & & \\\\\n%Avg. contact force & 3.3N & 8.1N \\\\\n\\hline\n\\end{tabular}\n\\end{center}\n\\end{table}",
+    "extracted_content": [
+      [
+        "",
+        "MPC",
+        "Baseline Controller"
+      ],
+      [
+        "Success rate\nAvg. max. contact force\nAvg. contact force above\nfthresh (5N)\nci",
+        "5/5\n5.6N\n5.5N",
+        "3/5\n17.7N\n14.3N"
+      ]
+    ],
+    "similarity_score": 0.7210884353741497,
+    "table_image": "images/1304.6146v1/table_1.png",
+    "page_image": "pages/1304.6146v1/page_13.png"
+  },
+  {
+    "id": 2,
+    "page": 14,
+    "bounding_box": [
+      312.5440368652344,
+      488.3999938964844,
+      562.4610595703125,
+      544.2459716796875
+    ],
+    "latex_content": "\\begin{table} [t]\n\\mycaption{Model predictive controller vs baseline controller\nin foliage.\\label{tbl:foliage_comparison}}\n\\vspace{-0.2cm}\n\\begin{center}\n\\begin{tabular} {|l|c|c|}\n\\hline\n& MPC &\nBaseline Controller \\\\\n\\hline\nSuccess rate & 3/5 & 1/5 \\\\\nExceeded safety threshold (15N) & 0/20 attempts & 19/20 attempts \\\\\nAvg. max. contact force & 5.5N & 14.5N \\\\\nAvg. contact force above & \\multirow{2}{*}{5.2N} & \\multirow{2}{*}{9.2N} \\\\\n$f_{c_i}^{thresh}$ (5N) & & \\\\\n%don't care threshold (5N) & & \\\\\n\\hline\n\\end{tabular}\n\\end{center}\n\\vspace{-0.3cm}\n\\end{table}",
+    "extracted_content": [
+      [
+        "",
+        "MPC",
+        "Baseline Controller"
+      ],
+      [
+        "Success rate\nExceeded safety threshold (15N)\nAvg. max. contact force\nAvg. contact force above\nfthresh (5N)\nci",
+        "3/5\n0/20 attempts\n5.5N\n5.2N",
+        "1/5\n19/20 attempts\n14.5N\n9.2N"
+      ]
+    ],
+    "similarity_score": 0.6602870813397129,
+    "table_image": "images/1304.6146v1/table_2.png",
+    "page_image": "pages/1304.6146v1/page_14.png"
+  }
+]

images/1305.2505v1/gt.json

@@ -0,0 +1,58 @@
+[
+  {
+    "id": 0,
+    "page": 18,
+    "bounding_box": [
+      323.3190002441406,
+      67.260009765625,
+      525.5609893798828,
+      117.072998046875
+    ],
+    "latex_content": "\\begin{table}[t]\n\t\\centering\n\t\\begin{tabular}{|c|c|}\n\t\t\\hline\n\t\tHypothesis class & Rademacher Complexity\\\\\\hline\n\t\t$\\B_q(\\norm{\\W}_q)$ & $2\\norm{\\X}_p\\norm{\\W}_q\\sqrt{\\frac{p - 1}{n}}$\\\\\\hline\n\t\t$\\B_1(\\norm{\\W}_1)$& $2\\norm{\\X}_\\infty \\norm{\\W}_1\\sqrt{\\frac{e\\log d}{n}}$\\\\\\hline\n\t\\end{tabular}\n\t\\caption{Rademacher complexity bounds for AUC maximization. We have $1/p+1/q = 1$ and $q > 1$.}\n\t\\label{tab:rad-bounds-auc}\n\\end{table}",
+    "extracted_content": [
+      [
+        "Hypothesis class",
+        "Rademacher Complexity"
+      ],
+      [
+        "( )\nBq \u2225W\u2225q",
+        "q\n2 p\u22121\n\u2225X\u2225p \u2225W\u2225q n"
+      ],
+      [
+        "( )\nB1 \u2225W\u22251",
+        "q\n2 e log d\n\u2225X\u2225 \u221e\u2225W\u22251 n"
+      ]
+    ],
+    "similarity_score": 0.4713804713804714,
+    "table_image": "images/1305.2505v1/table_0.png",
+    "page_image": "pages/1305.2505v1/page_18.png"
+  },
+  {
+    "id": 1,
+    "page": 20,
+    "bounding_box": [
+      68.20600128173828,
+      67.260009765625,
+      276.67498779296875,
+      111.15899658203125
+    ],
+    "latex_content": "\\begin{table}[t]\n\t\\centering\n\t\\begin{tabular}{|c|c|}\n\t\t\\hline\n\t\tHypothesis Class & Rademacher Avg. Bound\\\\\\hline\n\t\t$\\S_2(1)$ & $\\kappa^2\\sqrt{\\frac{p}{n}}$\\\\\\hline\n\t\t$\\Delta(1)$ & $\\kappa^2\\sqrt{\\frac{e\\log p}{n}}$\\\\\\hline\n\t\\end{tabular}\n\t\\caption{Rademacher complexity bounds for Multiple kernel learning}\n\t\\label{tab:mkl-rad-bounds}\n\\end{table}",
+    "extracted_content": [
+      [
+        "Hypothesis Class",
+        "Rademacher Avg. Bound"
+      ],
+      [
+        "(1)\nS2",
+        "\u03ba2pp\nn"
+      ],
+      [
+        "\u2206(1)",
+        "q\n\u03ba2 e log p\nn"
+      ]
+    ],
+    "similarity_score": 0.5064377682403434,
+    "table_image": "images/1305.2505v1/table_1.png",
+    "page_image": "pages/1305.2505v1/page_20.png"
+  }
+]

images/1306.2584v2/gt.json

@@ -0,0 +1,98 @@
+[
+  {
+    "id": 0,
+    "page": 5,
+    "bounding_box": [
+      72.0,
+      82.7750244140625,
+      575.864990234375,
+      667.2822723388672
+    ],
+    "latex_content": "\\begin{table}[!hp]\n\\setcounter{table}{0}\n\\makeatletter \n\\renewcommand{\\thetable}{\\@arabic\\c@table} \n\\makeatother\n\n\\setlength{\\LTcapwidth}{\\textwidth}\n\\renewcommand{\\arraystretch}{1.5}\n\\begin{longtable}{|p{1in}|p{4in}|p{1.5in}|}\n\\caption{Listing of the 18 attractor signatures}\\\\\n\\hline\n\\textbf{Name} & \\textbf{Top members} & \\textbf{Comments}\\\\\n\\hline\n\\endfirsthead\n\n\\multicolumn{3}{c}\n{{\\bfseries \\tablename \\thetable{} -- continued from previous page}} \\\\\n\\hline \n\\textbf{Name} & \\textbf{Top members} & \\textbf{Comments}\\\\ \n\\hline \n\\endhead\n\n\\hline \\multicolumn{3}{r}{{Continued on next page...}} \\\\\n\\endfoot\n\n\\hline \\hline\n\\endlastfoot\n\n\\hline\n\\multicolumn{3}{|l|}{\\textbf{mRNA}}\\\\\n\\hline\n\\textbf{LYM}\t&\t\\textit{SASH3, CD53, NCKAP1L, LCP2, IL10RA, PTPRC, EVI2B, BIN2, WAS, HAVCR2}\t&\tlymphocyte infiltration\\\\\n\\hline\n\\textbf{CIN}\t&\t\\textit{TPX2, KIF4A, KIFC1, NCAPG, BUB1, NCAPH, CDCA5, KIF2C, PLK1, CENPA}\t&\tmitotic chromosomal instability\\\\\n\\hline\n\\textbf{MES}\t&\t\\textit{COL3A1, COL5A2, COL1A2, THBS2, COL5A1, VCAN, COL6A3, SPARC, AEBP1, FBN1}\t&\tmesenchymal transition \\\\\n\\hline\n\\textbf{END}\t&\t\\textit{CDH5, ROBO4, CXorf36, CD34, CLEC14A, ARHGEF15, CD93, LDB2, ELTD1, MYCT1} \t&\tendothelial markers\\\\\n\\hline\n\\textbf{``\\textit{AHSA2}''}\t&\t\\textit{AHSA2, LOC91316, PILRB, ZNF767, TTLL3, CCNL2, PABPC1L, LENG8, CHKB CPT1B, SEC31B}\t&\\\\\n\\hline\n\\textbf{IFIT}\t&\t\\textit{IFIT3, MX1, OAS2, RSAD2, CMPK2, IFIT1, IFI44L, IFI44, IFI6, OAS1}\t&\tinterferon-induced\\\\\n\\hline\n\\textbf{``\\textit{WDR38}''}\t&\t\\textit{WDR38, YSK4, ROPN1L, C1orf194, MORN5, WDR16, RSPH4A, FAM183A, ZMYND10, DNAI1}\t&\t\\\\\n\\hline\n\\multicolumn{3}{|l|}{\\textbf{Genomically co-localized mRNA}} \\\\\n\\hline\n\\textbf{MHC Class II}\t&\t\\textit{HLA-DPA1, HLA-DRA, HLA-DPB1, HLA-DRB1, HLA-DMA, HLA-DMB, HLA-DOA, HLA-DQA1, HLA-DRB5}\t&\tstrongly associated with LYM\\\\\n\\hline\n\\textbf{GIMAP cluster}\t&\t\\textit{GIMAP4, GIMAP7, GIMAP6, GIMAP5, GIMAP8, GIMAP1}\t&\tstrongly associated with LYM\\\\\n\\hline\n\\textbf{Chr8q24.3 amplicon}\t&\t\\textit{SHARPIN, HSF1, TIGD5, GPR172A, ZC3H3, EXOSC4, SCRIB, CYHR1, MAF1, PUF60} \t&\tmost prominent Pan-Cancer amplicon\\\\\n\\hline\n\\multicolumn{3}{|l|}{\\textbf{microRNA}} \\\\\n\\hline\n\\textbf{\\textit{DLK1}-\\textit{DIO3} RNA cluster}\t&\tmir-127, mir-134, mir-379, mir-409, mir-382, mir-758, mir-381, mir-370, mir-654, mir-431\t&\tincludes \\textit{MEG3} long noncoding RNA; associated with MES\\\\\n\\hline\n\\textbf{``mir-509''}\t&\tmir-509, mir-514, mir-508\t&\t\\\\\n\\hline\n\\textbf{``mir-144''}\t&\tmir-144, mir-451, mir-486\t&\tassociated with erythropoiesis\\\\\n\\hline\n\\multicolumn{3}{|l|}{\\textbf{Methylation}}\\\\\n\\hline\n\\textbf{``RMND1''}\t&\tRMND1-6-151814639, MAP3K7-6-91353911, DNAAF1-16-82735714, PTRH2-17-55139429, ZNF143-11-9439170, cg03627896\n, TAMM41-3-11863582, CDK5-7-150385869, OTUB1-11-63510174, AATF-17-32380976\t&\t\\\\\n\\hline\n\\textbf{M+}\t&\tcg13928306, MTMR11-1-148175405, cg27324619, TNKS1BP1-11-56846646, C11orf52-11-111294703, IL17RC-3-9934128, cg24765079, ERBB3-12-54759072, IL22RA1-1-24342151, C11orf52-11-111294903\t&\tmethylated in infiltrating lymphocytes\\\\\n\\hline\n\\textbf{M-}\t&\tBIN2-12-50003941, PTPRCAP-11-66961771, TNFAIP8L2-1-149395922, IGFLR1-19-40925164, FAM113B-12-45896487, CD6-11-60495754, KLHL6-3-184755939, PTPN7-1-200396189, FAM78A-9-133141340, ACAP1-17-7180947\t&\tUnmethylated in infiltrating lymphocytes, may be causal to the expression of some of the genes of the LYM signature\\\\\n\\hline\n\\multicolumn{3}{|l|}{\\textbf{Protein activity}} \\\\\n\\hline\n\\textbf{``c-Met''}\t&\tc-Met, Snail, PARP\\_cleaved, Caspase-8, ERCC1, Rb\t&\tRelated to apoptosis\\\\\n\\hline\n\\textbf{``Akt''}\t&\tAkt, Tuberin, STAT5A\t&\\\\\n\\end{longtable}\n\n\\begin{figure}[!p]\n\\fbox{\n\\begin{minipage}{6.5in}\n\\includegraphics[width=\\textwidth]{figure2.png}\n\\caption{\nScatter plots connecting the LYM, M+ and M- metagenes in 12 cancer types. Each dot represents a cancer sample. The horizontal and vertical axes measure the average methylation values of the two methylation signatures, M- and M+, while the value of the expression of the LYM metagene is color-coded. In all three cases, the metagene is defined by the average of the top ranked genes as described in \\textbf{Table ~\\ref{tab:tabS1}}. \n}\n\\label{fig:fig2}\n\\end{minipage}\n}\n\\end{figure}\n\n\\subsection{Lymphocyte infiltration: LYM mRNA signature; M+\\hspace{1pt} methylation signature; M-\\hspace{1pt}\u00a0methylation signature}\nThese three signatures are related to tumor infiltration by lymphocytes. We list \nthem together because they are strongly interrelated (\\textbf{Fig. ~\\ref{fig:fig2}}) even though each \nof the three was independently derived using an unsupervised computational method. \nThe presence of LYM is accompanied by the presence of M+ and the absence of M- in \nall solid cancer types, suggesting that the three signatures reflect the same \nbiomolecular event, which appears to be the infiltration of immune cells in tumor \ntissue. Indeed, there is remarkable similarity (\\textbf{Fig. ~\\ref{fig:fig3}}) between the LYM signature \nand the ``immune score'' of the ESTIMATE tumor purity computational tool (\\url{http://ibl.mdanderson.org/estimate}). \nThe values of the M+ methylation signature are also remarkably similar to those of the methylation-based \n``leukocyte percentage'' estimation \\cite{huiNature} (available under Synapse ID syn1809222).\n\n\\begin{figure}[!p]\n\\fbox{\n\\begin{minipage}{6.5in}\n\\includegraphics[width=\\textwidth]{figure3.png}\n\\caption{\nScatter plots demonstrating the pan-cancer similarity of the value of the LYM metagene with the immune score of the ESTIMATE tumor purity computational tool (http://ibl.mdanderson.org/estimate) measuring immune cell infiltration. Each dot represents a cancer sample. The horizontal axis measures the expression value of the LYM metagene and the vertical axis measures the ESTIMATE immune score of infiltration. Note that the ESTIMATE did not provide scores for rectum cancer, and the estimation of immune cell infiltration is not applicable in leukemia.\n}\n\\label{fig:fig3}\n\\end{minipage}\n}\n\\end{figure}\n\nWe had previously found\\cite{billCancerInfo} all three LYM, M+ and M- signatures from their association with \nthe expression of miR-142. We have now confirmed this association with miR-142 in the pancan12 data sets, \nand we found that miR-150 and miR-155 are also strongly associated with the LYM signature. \nWe had also previously independently identified the LYM signature as an attractor metagene\\cite{mePLoS}, and \nused it in the winning model of the Sage Bionetworks Breast Cancer Prognosis Challenge\\cite{meSTM}. \nSpecifically the LYM signature is strongly associated with improved prognosis in ER-negative breast cancers, \nand this fact also provides an explanation for the relatively better prognosis in medullary, \ncompared with other types of high-grade breast cancers. \n\nThe interrelationship of the LYM, M+ and M- signatures, as shown in \\textbf{Fig. ~\\ref{fig:fig2}}, appears to be a consequence \nof the presence of different subclasses of cells (as opposed to being a methylation switch inside \nthe same cell), consistent with their assumed role of measuring the extent of lymphocyte infiltration \nin the tumor. In other words, the M+ methylation sites, normally unmethylated, are largely methylated \nin the infiltrating leukocytes; and the M- methylation sites, normally methylated, are largely \nunmethylated in the infiltrating leukocytes. Consistently, many of the genes methylated by the M- \nsignature are identical to those of LYM (six among the 27 genes of the M- signature (\\textit{BIN2}, \\textit{TNFAIP8L2}, \n\\textit{ACAP1}, \\textit{NCKAP1L}, \\textit{FAM78A}, \\textit{PTPN7}) listed in Table S1 are also among the 168 \ngenes listed in the LYM attractor metagene \n($P < 9.21\\times10^{-7}$ based on Fisher\u2019s exact test). The observed significant overlap in the gene sets and \nthe negative association between gene expression in LYM and DNA methylation in M- are consistent with the notion \nthat the absence of DNA methylation is permissive for gene expression, suggesting that the expression of the LYM\nsignature in the infiltrating lymphocytes may be facilitated in part by the hypomethylation of the M- signature.\n\n\n\nThe sharp definition of the LYM signature (being a Pan-Cancer attractor signature pointing to few genes \nat the core of coexpression) provides strong hints about the precise nature of this leukocyte infiltration. \nSpecifically, the membership of the top-ranked genes (\\textit{SASH3}, \\textit{CD53}, \\textit{NCKAP1L}, \\textit{LCP2}, \n\\textit{IL10RA}, \\textit{PTPRC}, \\textit{EVI2B}, \\textit{BIN2}, \\textit{WAS}, \\textit{HAVCR2}, \\ldots) \npoint to a specific type of lymphocytes. We have speculated\\cite{mePLoS} that these infiltrating lymphocytes \nare T cells having undergone a particular type of co-stimulation providing hypotheses for related adoptive transfer therapy.\n\nTwo proteins strongly associated with the LYM signature are two tyrosine kinases: Lck (lymphocyte-specific protein tyrosine kinase) \nand Syk (spleen tyrosine kinase).\n\n\\subsection{CIN (mitotic chromosomal instability) mRNA signature}\nThis signature is related to mitotic chromosomal instability. It is similar to numerous known ``proliferation'' signatures, \nbut its sharp definition as an attractor metagene specifically points to the kinetochore-microtubule interface and \nassociated kinesins. Comparison with similar mitotic signatures in normal cells may help pinpoint driver genes for \nmalignant chromosomal instability. The signature is strongly associated with tumor grade as well as poor prognosis in many, \nif not all, cancer types.\n\nTwo proteins strongly associated with the CIN signature are Cyclin B1 and CDK1. Consistently, it is known that the cyclin \nB1-Cdk1 complex of cyclin-dependent kinase 1 is involved in the early events of mitosis, and that nuclear cyclin B1 protein \nmay induce chromosomal instability and enhance the aggressiveness of the carcinoma cells\\cite{bibi7}. \n\n\\subsection{MES (mesenchymal transition) mRNA signature}\nThis signature is related to mesenchymal transition and invasiveness of cancer cells. It is similar to numerous \n``stromal'' or ``mesenchymal'' signatures; however there is evidence\\cite{dimitrisBMCCancer} that many among the genes of the \nsignature are largely produced by transdifferentiated cancer cells. We hypothesize that such cells, known to \nassume the duties of cancer-associated fibroblasts in some tumors\\cite{hanahan2011}, may  have become indistinguishable, \neven using laser capture microdissection, from stromal fibroblasts. We had originally identified the MES signature \nfrom its association with tumor stage\\cite{hoonBMCMedGenomics}; specifically the signature appears only after a particular cancer \ntype-specific tumor stage threshold has been reached.\n\nThe values of the MES signature are remarkably similar to the ``stromal score'' of the ESTIMATE tumor purity computational tool \n(\\url{http://ibl.mdanderson.org/estimate}) measuring fibroblast infiltration. Based on our previous reasoning, however, \nwe believe that this interpretation may not be fully accurate, and that it will be important to find out to what extent \nsome of the cells expressing some of these mesenchymal markers may actually be transdifferentiated cancer cells, and \nwhether the estimated tumor purity may be affected by other types of normal cells instead of stromal fibroblasts.\n\nThe co-regulated microRNAs most strongly associated with the MES signature are miR-199a, miR-199b, and miR-214. The \\textit{DLK1}-\n\\textit{DIO3} RNA cluster attractor signature, described later, is also strongly associated with MES.\n\nThe protein most strongly associated with the MES signature is Fibronectin.\n\n\\subsection{END (endothelial marker) mRNA signature}\nThis is a novel angiogenesis-associated attractor signature. Nearly all the top-ranked genes (\\textbf{Table 1}) are \nendothelial markers.  The top gene, \\textit{CDH5}, codes for VE-cadherin, which is known to be involved in a pathway suppressing \nangiogenic sprouting\\cite{bib9}. The second gene, \\textit{ROBO4}, is known to inhibit VEGF-induced pathologic angiogenesis and endothelial \nhyperpermeability\\cite{bib10}. Consistently, the END attractor metagene appears to be protective and anti-angiogenic, stabilizing the \nvascular network. For example, 22 out of the 27 genes of the END attractor are among the 265 genes included in File S2 of a recent \nstudy\\cite{bib11} of renal cell carcinoma ($P < 8.4\\times10^{-38}$ based on Fisher\u2019s exact test) as most associated with patients\u2019 \nsurvival. These good-prognosis genes were intermixed in the same file with many poor-prognosis genes of the CIN attractor, \nsuggesting that the CIN and END attractor metagenes are two of the most prognostic features in renal cell carcinoma. \n\nInterestingly, the MES and END attractor metagenes are positively associated with each other (\\textbf{Fig. ~\\ref{fig:fig4}}), in the sense \nthat overexpression of the END signature tends to imply overexpression of the MES signature and vice-versa. This is consistent \nwith mutual exclusivity between angiogenesis and invasiveness and with related findings\\cite{bib12} that VEGF inhibits tumor cell \ninvasion and mesenchymal transition, while antiangiogenic therapy is associated with increased invasiveness\\cite{bib13}. It may also \nexplain the paradoxical protective nature of signatures related to the MES attractor metagene in invasive breast cancers\\cite{bib14}.\n\n\\begin{figure}[!p]\n\\fbox{\n\\begin{minipage}{6.5in}\n\\includegraphics[width=\\textwidth]{figure4.png}\n\\caption{\nScatter plots demonstrating the association between MES and END attractor metagenes. The horizontal and vertical axes measure the values of the MES and END signatures. The two signatures have positive correlation, although this association is not sufficiently strong to merge the two attractors into one. This association suggests that the invasive MES signature and the antiangiogenic END signature tend to be present simultaneously.\n}\n\\label{fig:fig4}\n\\end{minipage}\n}\n\\end{figure}\n\n\n\\subsection{``AHSA2'' mRNA signature}\nWe do not yet know what this signature represents. We observed that several noncoding RNAs (e.g. NCRNA00105, NCRNA00201) \nare in relatively high-ranked positions among its members.\n\n\\subsection{IFIT (interferon-induced) mRNA signature}\nThe members of this signature are interferon-induced. For example, we observed large enrichment of the genes of the \nsignature among those upregulated by IFN-$\\alpha$ in the side population (SP) of ovarian cancer cells\\cite{bib15} from the \nlist provided in Supplementary Table S4 of that paper, in which the authors concluded that tumors bearing large SP numbers \ncould be particularly sensitive to IFN-$\\alpha$ treatment. \n\n\\subsection{``WDR38'' mRNA signature}\nWe do not know what this signature represents, except that we had found one of its key members, gene \\textit{ZMYND10}, to be \nprotective and associated with estrogen receptor expression in breast cancer.\n\n\\subsection{MHC Class II genomically co-localized mRNA signature}\nWe found this signature using the genomically co-localized version of the algorithm. It is very highly correlated with LYM.\n\n\\subsection{GIMAP genomically co-localized mRNA signature}\nAs above, we found this signature using the genomically co-localized version of the algorithm. It is also very highly correlated with LYM.\n\n\\subsection{Chr8q24.3 amplicon mRNA signature}\nThis is the strongest pan-cancer amplicon signature. It was previously found predictive of early relapse in ER-positive breast cancers\\cite{bib16}.\n\n\\subsection{``RMND1'' methylation signature}\nWe do not yet know what the comethylation of the sites of this signature signifies.\n\n\\begin{figure}[!t]\n\\fbox{\n\\begin{minipage}{6.5in}\n\\includegraphics[width=\\textwidth]{figure5.png}\n\\caption{\nThe \\textit{DLK1}-\\textit{DIO3} cluster of noncoding RNAs. Shown is a screen capture from the UCSC Genome Browser (\\url{http://genome.ucsc.edu}). The cluster of imprinted genes delineated by the \\textit{DLK1} and \\textit{DIO3} genes (outside the shown region) is located on chromosome 14. We found that the corresponding pan-cancer attractor signature does not contain any paternally inherited protein-coding genes. It does contain the numerous noncoding RNA genes expressed from the maternally inherited homolog, including the \\textit{MEG3} long noncoding RNA gene.\n}\n\\label{fig:fig5}\n\\end{minipage}\n}\n\\end{figure}\n\n\\subsection{\\textit{DLK1}-\\textit{DIO3} RNA cluster signature}\nThis is the strongest pan-cancer multi-microRNA coexpression signature. It consists of numerous noncoding RNAs within the \n\\textit{DLK1}-\\textit{DIO3} imprinted genomic region of chr14q32. \\textbf{Fig. ~\\ref{fig:fig5}} shows a screen capture of the genomic region \nfrom the UCSC Genome Browser (\\url{http://genome.ucsc.edu/}). We confirmed that the coexpression signature also includes the \n\\textit{MEG3} long noncoding RNA located at the upstream end of the region. It may also include numerous small nuclear \nRNAs at the central region, but there were no associated probe sets to confirm the coexpression. We found that this ncRNA \nsignature is associated with the MES (mesenchymal transition) mRNA signature. For example, the ranked list of mRNAs most \nassociated with the \\textit{DLK1}-\\textit{DIO3} ncRNA signature starts from \\textit{POSTN}, \\textit{PCOLCE}, \\textit{COL5A2}, \\textit{COL1A2}, \n\\textit{GLT8D2}, \\textit{COL5A1}, \\textit{SFRP2}, and \\textit{FAP}.\n\nExpression of the imprinted \\textit{DLK1}-\\textit{DIO3} ncRNA cluster is believed to be vital for the development potential of \nembryonic stem cells\\cite{bib17}, consistent with the hypothesis\\cite{bib18} that mesenchymal transition in cancer reactivates \nembryonic developmental programs and makes cancer cells invasive and stem-like. The \\textit{DLK1}-\\textit{DIO3} ncRNA signature \nwas also found to define a stem-like subtype of hepatocellular carcinoma associated with poor survival\\cite{bib19}. The details \nof the regulation mechanism for this ncRNA cluster coexpression in the \\textit{DLK1}-\\textit{DIO3} region are unclear.\n\n\\subsection{``miR-509/miR-514/miR-508'' microRNA signature}\nThese three microRNAs are co-localized at chrXq27.3. We do not know what this signature signifies.\n\n\\subsection{``miR-144/miR-451/miR-486'' microRNA signature}\nThis is a three-microRNA signature related to erythropoiesis. The first two genes are located in the bicistronic microRNA \nlocus miR-144/451, highly expressed during erythrocyte development\\cite{bib20}. The mRNAs most associated to this microRNA \nsignature are hemoglobin-related: \\textit{HBB}, \\textit{HBA1}, \\textit{HBA2} and \\textit{ALAS2}. The protein most associated \nwith this signature is HER3. These three microRNAs were identified as promising biomarkers for detection of esophageal cancer.\n\n\\subsection{c-Met/Snail/PARP\\_cleaved/Caspase-8/ERCC1/Rb protein activity signature}\nThis protein coexpression signature appears to combine the contribution of several pathways and we hope that a plausible \nand useful biological ``story'' will be developed based on the simultaneous activity of all these six proteins in some \ncancer samples. We note that each of these proteins\\cite{bib21, bib22, bib23, bib24, bib25, bib26} has been related in various ways with resistance to chemotherapy \nor apoptosis.\n\n\\subsection{Akt/Tuberin/STAT5A protein activity signature}\nWe do not know what the coexpression of Akt, Tuberin, STAT5A proteins represents in cancer. It is known, however, that \nlow levels of STAT5A protein in breast cancer are associated with tumor progression and unfavorable clinical outcomes\\cite{bib27}.\n\n\\section*{DISCUSSION}\nThe Pan-Cancer nature (\\textbf{Fig. ~\\ref{fig:figS1}}) of each of the signatures described in this paper suggests that they represent \nimportant biomolecular events.  A reasonable concern is whether some of these ``pan-cancer'' signatures may instead reflect \nfundamental normal ``pan-tissue'' biological mechanisms. Even if this is true for some of these signatures, this does not \nexclude the possibility that they are aberrant and play important roles in some cancer samples.  Furthermore, this provides \nthe opportunity to compare similar signatures in normal vs. malignant tissues to pinpoint potential cancer-specific genes. \n\nBecause of its exhaustive search starting from all potential ``seeds'' in all data sets from twelve different cancer types, \nour iterative data mining algorithm is guaranteed to have identified all pan-cancer molecular signatures involving simultaneous \npresence of a large number of coordinately expressed genes, proteins, or comethylated sites. We hope that these signatures are \nfurther scrutinized by the medical research community for the purpose of developing potential diagnostic, predictive, and \neventually therapeutic products applicable in multiple cancers.  \n\n\\section*{ACKNOWLEDGEMENTS}\nWe are thankful to Hanina Hibshoosh, Chris Miller and Gordon Mills for helpful discussions, \nwhich contributed to improved interpretation of the signatures disclosed in this paper.\n\n\\section*{Accessibility} All figures in this paper, including supplementary figures and tables, as well as the files of generated attractor molecular signatures, are available in Synapse under ID syn1686966. \n\n\\section*{Data description and availability} The data sets of TCGA pancan12 freeze 4.7 used to derive the results of this paper are described and are available under Synapse ID syn300013 with doi:10.7303/syn300013. The twelve cancer types are bladder urothelial carcinoma (BLCA), breast invasive carcinoma (BRCA), colon adenocarcinoma (COAD), glioblastoma multiforme (GBM), head and neck squamous cell carcinoma (HNSC), kidney renal clear cell carcinoma (KIRC), acute myeloid leukemia (LAML), lung adenocarcinoma (LUAD) , lung squamous cell carcinoma (LUSC), ovarian serous cystadenocarcinoma (OV), rectum adenocarcinoma (READ), and uterine corpus endometrioid carcinoma (UCEC).\n\n\\pagebreak\n%% Put the bibliography here, most people will use BiBTeX in\n%% which case the environment below should be replaced with\n%% the \\bibliography{} command.\n\n\\begin{thebibliography}{1}\n\\bibitem{hanahan2000} Hanahan, D. \\& Weinberg, R.A. The hallmarks of cancer. Cell 100, 57-70 (2000).\n\\bibitem{hanahan2011} Hanahan, D. \\& Weinberg, R.A. Hallmarks of cancer: the next generation. Cell 144, 646-74 (2011). \n\\bibitem{mePLoS} Cheng, W.Y., Ou Yang, T.H. \\& Anastassiou, D. Biomolecular events in cancer revealed by attractor metagenes. PLoS Comput Biol 9, e1002920 (2013).\n\\bibitem{meSTM} Cheng, W.Y., Ou Yang, T.H. \\& Anastassiou, D. Development of a prognostic model for breast cancer survival in an open challenge environment. Sci Transl Med 5, 181ra50 (2013).\n\\bibitem{bib5} McCarthy, N. Prognostic models: Rising to the challenge. Nat Rev Cancer, doi:10.1038/nrc3530 (2013).\n\\bibitem{huiNature} Shen, H. \\textit{et al.} Comprehensive Cross-Cancer Comparison of DNA Methylation Profiles, submitted. (2013).\n\\bibitem{billCancerInfo} Andreopoulos, B. \\& Anastassiou, D. Integrated Analysis Reveals hsa-miR-142 as a Representative of a Lymphocyte-Specific Gene Expression and Methylation Signature. Cancer Inform 11, 61-75 (2012).\n\\bibitem{bibi7} Suzuki, T. \\textit{et al.} Nuclear cyclin B1 in human breast carcinoma as a potent prognostic factor. Cancer Sci 98, 644-51 (2007).\n\\bibitem{dimitrisBMCCancer} Anastassiou, D. \\textit{et al.} Human cancer cells express Slug-based epithelial-mesenchymal transition gene expression signature obtained in vivo. BMC Cancer 11, 529 (2011).\n\\bibitem{hoonBMCMedGenomics} Kim, H., Watkinson, J., Varadan, V. \\& Anastassiou, D. Multi-cancer computational analysis reveals invasion-associated variant of desmoplastic reaction involving INHBA, THBS2 and COL11A1. BMC Med Genomics 3, 51 (2010).\n\\bibitem{bib9} Abraham, S. \\textit{et al.} VE-Cadherin-mediated cell-cell interaction suppresses sprouting via signaling to MLC2 phosphorylation. Curr Biol 19, 668-74 (2009).\n\\bibitem{bib10} Jones, C.A. \\textit{et al.} Robo4 stabilizes the vascular network by inhibiting pathologic angiogenesis and endothelial hyperpermeability. Nat Med 14, 448-53 (2008).\n\\bibitem{bib11} Wozniak, M.B. \\textit{et al.} Integrative genome-wide gene expression profiling of clear cell renal cell carcinoma in Czech Republic and in the United States. PLoS One 8, e57886 (2013).\n\\bibitem{bib12} Lu, K.V. \\textit{et al.} VEGF inhibits tumor cell invasion and mesenchymal transition through a MET/VEGFR2 complex. Cancer Cell 22, 21-35 (2012).\n\\bibitem{bib13} Paez-Ribes, M. \\textit{et al}. Antiangiogenic therapy elicits malignant progression of tumors to increased local invasion and distant metastasis. Cancer Cell 15, 220-31 (2009).\n\\bibitem{bib14} Beck, A.H., Espinosa, I., Gilks, C.B., van de Rijn, M. \\& West, R.B. The fibromatosis signature defines a robust stromal response in breast carcinoma. Lab Invest 88, 591-601 (2008).\n\\bibitem{bib15} Moserle, L. \\textit{et al}. The side population of ovarian cancer cells is a primary target of IFN-alpha antitumor effects. Cancer Res 68, 5658-68 (2008).\n\\bibitem{bib16} Bilal, E. \\textit{et al}. Amplified Loci on Chromosomes 8 and 17 Predict Early Relapse in ER-Positive Breast Cancers. PLoS One 7, e38575 (2012).\n\\bibitem{bib17} Stadtfeld, M. \\textit{et al}. Aberrant silencing of imprinted genes on chromosome 12qF1 in mouse induced pluripotent stem cells. Nature 465, 175-81 (2010).\n\\bibitem{bib18} Mani, S.A. \\textit{et al}. The epithelial-mesenchymal transition generates cells with properties of stem cells. Cell 133, 704-15 (2008).\n\\bibitem{bib19} Luk, J.M. \\textit{et al}. DLK1-DIO3 genomic imprinted microRNA cluster at 14q32.2 defines a stemlike subtype of hepatocellular carcinoma associated with poor survival. J Biol Chem 286, 30706-13 (2011).\n\\bibitem{bib20} Yu, D. \\textit{et al}. miR-451 protects against erythroid oxidant stress by repressing 14-3-3zeta. Genes Dev 24, 1620-33 (2010).\n\\bibitem{bib21} Tang, M.K., Zhou, H.Y., Yam, J.W. \\& Wong, A.S. c-Met overexpression contributes to the acquired apoptotic resistance of nonadherent ovarian cancer cells through a cross talk mediated by phosphatidylinositol 3-kinase and extracellular signal-regulated kinase 1/2. Neoplasia 12, 128-38 (2010).\n\\bibitem{bib22} Haslehurst, A.M. \\textit{et al}. EMT transcription factors snail and slug directly contribute to cisplatin resistance in ovarian cancer. BMC Cancer 12, 91 (2012).\n\\bibitem{bib23} D\u2019Amours, D., Sallmann, F.R., Dixit, V.M. \\& Poirier, G.G. Gain-of-function of poly(ADP-ribose) polymerase-1 upon cleavage by apoptotic proteases: implications for apoptosis. J Cell Sci 114, 3771-8 (2001).\n\\bibitem{bib24} Kim, P.K., Mahidhara, R. \\& Seol, D.W. The role of caspase-8 in resistance to cancer chemotherapy. Drug Resist Updat 4, 293-6 (2001).\n\\bibitem{bib25} Olaussen, K.A. \\textit{et al}. DNA repair by ERCC1 in non-small-cell lung cancer and cisplatin-based adjuvant chemotherapy. N Engl J Med 355, 983-91 (2006).\n\\bibitem{bib26} Collard, T.J. \\textit{et al}. The retinoblastoma protein (Rb) as an anti-apoptotic factor: expression of Rb is required for the anti-apoptotic function of BAG-1 protein in colorectal tumour cells. Cell Death Dis 3, e408 (2012).\n\\bibitem{bib27} Peck, A.R. \\textit{et al}. Low levels of Stat5a protein in breast cancer are associated with tumor progression and unfavorable clinical outcomes. Breast Cancer Res 14, R130 (2012).\n\n\n\\end{thebibliography}\n\n\\pagebreak\n\\begin{methods}\n\\subsection{Data normalization}\nThe data platform for each cancer types and its corresponding Synapse ID is given below. \n\n\\begin{table}[!hf]\n\\renewcommand{\\arraystretch}{1}\n\\hspace*{-0.5in}\n\\begin{threeparttable}\n\\begin{tabular}{ |l|c|c|c|c|}\n\\hline\n\\textbf{Molecular profile} & mRNA & Protein & miRNA & DNA methylation \\\\ \n\\hline\n\\multirow{3}{*}{\\textbf{Platform}} &                      & Reverse phase protein    &                 & Infinium \\\\\n                                   &    Illumina HiSeq    & lysate microarray        & Illumina HiSeq  & HumanMethylation27 \\\\\n                                   &                      & (RPPA)                   &                 & BeadChip \\\\\n\\hline\n\\textbf{Cancer type} & \\multicolumn{4}{|c|}{\\textbf{Synapse ID}} \\\\\n\\hline\nBLCA&syn1571504&syn1681048&syn1571494&syn1889358\\tnote{*}\\\\ \nBRCA&syn417812&syn1571267&syn395575&syn411485\\\\ \nCOAD&syn1446197&syn416772&syn464211&syn411993\\\\ \nGBM&syn1446214&syn416777&NA&syn412284\\\\ \nHNSC&syn1571420&syn1571409&syn1571411&syn1889356\\tnote{*}\\\\ \nKIRC&syn417925&syn416783&syn395617&syn412701\\\\ \nLAML&syn1681084&NA&syn1571533&syn1571536\\\\ \nLUAD&syn1571468&syn1571446&syn1571453&syn1571458\\\\ \nLUSC&syn418033&syn1367036&syn395691&syn415758\\\\ \nOV&syn1446264&syn416789&syn1356544&syn415945\\\\ \nREAD&syn1446276&syn416795&syn464222&syn416194\\\\ \nUCEC&syn1446289&syn416800&syn395720&syn416204\\\\ \n\\hline\n\\end{tabular}\n\\begin{tablenotes}\n  \\item[*] The data sets were extracted from HumanMethylation450 BeadChip.\n\\end{tablenotes}\n\\end{threeparttable}\n\\end{table}",
+    "extracted_content": [
+      [
+        "Name",
+        "Top members",
+        "Comments"
+      ],
+      [
+        "END",
+        "CDH5, ROBO4, CXorf36, CD34, CLEC14A, ARHGEF15,\nCD93, LDB2, ELTD1, MYCT1",
+        "endothelial markers"
+      ],
+      [
+        "\u201cAHSA2\u201d",
+        "AHSA2, LOC91316, PILRB, ZNF767, TTLL3, CCNL2,\nPABPC1L, LENG8, CHKB CPT1B, SEC31B",
+        ""
+      ],
+      [
+        "IFIT",
+        "IFIT3, MX1, OAS2, RSAD2, CMPK2, IFIT1, IFI44L, IFI44,\nIFI6, OAS1",
+        "interferon-induced"
+      ],
+      [
+        "\u201cWDR38\u201d",
+        "WDR38, YSK4, ROPN1L, C1orf194, MORN5, WDR16,\nRSPH4A, FAM183A, ZMYND10, DNAI1",
+        ""
+      ],
+      [
+        "Genomically co-localized mRNA",
+        null,
+        null
+      ],
+      [
+        "MHC Class II",
+        "HLA-DPA1, HLA-DRA, HLA-DPB1, HLA-DRB1, HLA-\nDMA, HLA-DMB, HLA-DOA, HLA-DQA1, HLA-DRB5",
+        "strongly associated\nwith LYM"
+      ],
+      [
+        "GIMAP clus-\nter",
+        "GIMAP4, GIMAP7, GIMAP6, GIMAP5, GIMAP8, GIMAP1",
+        "strongly associated\nwith LYM"
+      ],
+      [
+        "Chr8q24.3\namplicon",
+        "SHARPIN, HSF1, TIGD5, GPR172A, ZC3H3, EXOSC4,\nSCRIB, CYHR1, MAF1, PUF60",
+        "most prominent Pan-\nCancer amplicon"
+      ],
+      [
+        "microRNA",
+        null,
+        null
+      ],
+      [
+        "DLK1-DIO3\nRNA cluster",
+        "mir-127, mir-134, mir-379, mir-409, mir-382, mir-758, mir-\n381, mir-370, mir-654, mir-431",
+        "includes MEG3 long\nnoncoding RNA; asso-\nciated with MES"
+      ],
+      [
+        "\u201cmir-509\u201d",
+        "mir-509, mir-514, mir-508",
+        ""
+      ],
+      [
+        "\u201cmir-144\u201d",
+        "mir-144, mir-451, mir-486",
+        "associated with ery-\nthropoiesis"
+      ],
+      [
+        "Methylation",
+        null,
+        null
+      ],
+      [
+        "\u201cRMND1\u201d",
+        "RMND1-6-151814639, MAP3K7-6-91353911, DNAAF1-\n16-82735714, PTRH2-17-55139429, ZNF143-11-9439170,\ncg03627896 , TAMM41-3-11863582, CDK5-7-150385869,\nOTUB1-11-63510174, AATF-17-32380976",
+        ""
+      ],
+      [
+        "M+",
+        "cg13928306, MTMR11-1-148175405, cg27324619,\nTNKS1BP1-11-56846646, C11orf52-11-111294703,\nIL17RC-3-9934128, cg24765079, ERBB3-12-54759072,\nIL22RA1-1-24342151, C11orf52-11-111294903",
+        "methylated in infiltrat-\ning lymphocytes"
+      ]
+    ],
+    "similarity_score": 0.5887573964497042,
+    "table_image": "images/1306.2584v2/table_0.png",
+    "page_image": "pages/1306.2584v2/page_5.png"
+  }
+]

images/1307.0781v1/gt.json

@@ -0,0 +1,220 @@
+[
+  {
+    "id": 0,
+    "page": 14,
+    "bounding_box": [
+      175.91109924316407,
+      53.33697509765625,
+      429.4549987792969,
+      120.38897705078125
+    ],
+    "latex_content": "\\begin{table}[t]\n\\centering\n{\\fontsize{8}{6}\\selectfont\n\\setlength{\\tabcolsep}{.1em}\n\\begin{tabular}{|l|c|c|c|c|c|}\n\\hline\nLearner & 1 &2 &3 & 4\\\\\n\\hline\nClassification  & Naive Bayes,  &Always $1$, & RBF Network, & Random Tree,  \\\\\nFunction (S1)  & Logistic & Voted Perceptron &J48 & Always $0$\\\\\n\\hline\n Error  & 47,  & 53,  & 47, & 47,  \\\\\npercentage (S1) & 3 & 4 &  47 & 47\\\\\n\\hline\n Classification  & Naive Bayes,  &Always $1$,  & RBF Network,  & Random Tree,  \\\\\nFunction (S2) & Random & Random & J48 & Always $0$ \\\\\n\\hline\n Error & 47,  & 53,  & 47,  & 47,  \\\\ \npercentage (S2) & 50 & 50 & 47 & 47 \\\\\n\\hline\n\\end{tabular}\n}\n\\add{\\vspace{-0.1in}}\n\\caption{Simulation setup}\n\\vspace{-0.25in}\n\\label{tab:sim_setup}\n\\end{table}",
+    "extracted_content": [
+      [
+        "Learner",
+        "1",
+        "2",
+        "3",
+        "4"
+      ],
+      [
+        "Classification\nFunction (S1)",
+        "Naive Bayes,\nLogistic",
+        "Always 1,\nVoted Perceptron",
+        "RBF Network,\nJ48",
+        "Random Tree,\nAlways 0"
+      ],
+      [
+        "Error\npercentage (S1)",
+        "47,\n3",
+        "53,\n4",
+        "47,\n47",
+        "47,\n47"
+      ],
+      [
+        "Classification\nFunction (S2)",
+        "Naive Bayes,\nRandom",
+        "Always 1,\nRandom",
+        "RBF Network,\nJ48",
+        "Random Tree,\nAlways 0"
+      ],
+      [
+        "Error\npercentage (S2)",
+        "47,\n50",
+        "53,\n50",
+        "47,\n47",
+        "47,\n47"
+      ]
+    ],
+    "similarity_score": 0.5048543689320388,
+    "table_image": "images/1307.0781v1/table_0.png",
+    "page_image": "pages/1307.0781v1/page_14.png"
+  },
+  {
+    "id": 1,
+    "page": 14,
+    "bounding_box": [
+      212.12024688720703,
+      151.96697998046875,
+      393.2457580566406,
+      179.71002197265625
+    ],
+    "latex_content": "\\begin{table}[t]\n\\centering\n{\\fontsize{8}{6}\\selectfont\n\\setlength{\\tabcolsep}{.3em}\n\\begin{tabular}{|l|c|c|c|c|}\n\\hline\n & $D_1(t)$ & $D_2(t)$ & $D_3(t)$ & $m_T$  \\\\\n\\hline\n(C1) CoS & $t^{1/8} \\log t$ & $2 t^{1/8} \\log t$ & $t^{1/8} \\log t$ & $\\lceil T \\rceil^{1/4}$  \\\\\n\\hline\n%(C1) DCZA & $t^{1/8} \\log t$ & $2 t^{1/8} \\log t$ & $t^{1/8} \\log t$ &  & $1$ & $4$ \\\\\n%\\hline\n(C2) CoS & $t^{1/2} \\log t$ & $2 t^{1/2} \\log t$ & $t^{1/2} \\log t$ & $\\lceil T \\rceil^{1/4}$  \\\\\n\\hline\n%(C2) DCZA & $t^{2/p} \\log t$ & $2 t^{2/p} \\log t$ & $t^{2/p} \\log t$ &  & $1$ & $(3+\\sqrt{17})/2$ \\\\\n%\\hline\n\\end{tabular}\n}\n\\add{\\vspace{-0.05in}}\n\\caption{Parameters for CoS}\n\\label{tab:par_setup}\n\\add{\\vspace{-0.4in}}\n\\end{table}",
+    "extracted_content": [
+      [
+        "",
+        "D1(t)",
+        "D2(t)",
+        "D3(t)",
+        "mT"
+      ],
+      [
+        "(C1) CoS",
+        "t1/8 log t",
+        "2t1/8 log t",
+        "t1/8 log t",
+        "\u2308T\u23091/4"
+      ],
+      [
+        "(C2) CoS",
+        "t1/2 log t",
+        "2t1/2 log t",
+        "t1/2 log t",
+        "\u2308T\u23091/4"
+      ]
+    ],
+    "similarity_score": 0.5446009389671361,
+    "table_image": "images/1307.0781v1/table_1.png",
+    "page_image": "pages/1307.0781v1/page_14.png"
+  },
+  {
+    "id": 2,
+    "page": 3,
+    "bounding_box": [
+      176.27973225911458,
+      53.33697509765625,
+      429.19012451171875,
+      155.47601318359375
+    ],
+    "latex_content": "\\begin{table}[t]\n\\centering\n{\\renewcommand{\\arraystretch}{0.6}\n{\\fontsize{8}{6}\\selectfont\n\\setlength{\\tabcolsep}{.1em}\n\\begin{tabular}{|l|c|c|c|c|c|}\n\\hline\n&  \\cite{breiman1996bagging, buhlmann2003boosting, lazarevic2001distributed, chen2004channel, perlich2011cross} & \\cite{mateos2010distributed, kargupta1999collective} &  \\cite{zheng2011attribute} & This work \\\\\n\\hline\nAggregation & non-cooperative & cooperative & cooperative & \\rev{no} \\\\\n\\hline\nMessage  & none & data & training  & data and label \\\\\nexchange & & & residual & only if improves  \\\\\n& & & &   performance \\\\\n\\hline\nLearning  & offline/online & offline & offline & Non-bayesian \\\\\napproach&&&& online\\\\\n\\hline\nCorrelation & N/A & no & no & yes\\\\\nexploitation & & & &\\\\\n\\hline\nInformation from  & no & all & all & only if improves  \\\\\nother learners & & & &  accuracy \\\\\n\\hline\nData partition & horizontal & horizontal & vertical & horizontal \\\\\n\\hline\nBound on regret,  & no &no &no &yes - sublinear\\\\\nconvergence rate &&&&\\\\\n\\hline\n\\end{tabular}\n}\n}\n\\caption{Comparison with related work in distributed data mining}\n\\label{tab:comparison1}\n\\add{\\vspace{-0.1in}}\n\\end{table}",
+    "extracted_content": [
+      [
+        "",
+        "[3], [8], [13]\u2013[15]",
+        "[7], [9]",
+        "[5]",
+        "This work"
+      ],
+      [
+        "Aggregation",
+        "non-cooperative",
+        "cooperative",
+        "cooperative",
+        "no"
+      ],
+      [
+        "Message\nexchange",
+        "none",
+        "data",
+        "training\nresidual",
+        "data and label\nonly if improves\nperformance"
+      ],
+      [
+        "Learning\napproach",
+        "offline/online",
+        "offline",
+        "offline",
+        "Non-bayesian\nonline"
+      ],
+      [
+        "Correlation\nexploitation",
+        "N/A",
+        "no",
+        "no",
+        "yes"
+      ],
+      [
+        "Information from\nother learners",
+        "no",
+        "all",
+        "all",
+        "only if improves\naccuracy"
+      ],
+      [
+        "Data partition",
+        "horizontal",
+        "horizontal",
+        "vertical",
+        "horizontal"
+      ],
+      [
+        "Bound on regret,\nconvergence rate",
+        "no",
+        "no",
+        "no",
+        "yes - sublinear"
+      ]
+    ],
+    "similarity_score": 0.47578589634664403,
+    "table_image": "images/1307.0781v1/table_2.png",
+    "page_image": "pages/1307.0781v1/page_3.png"
+  },
+  {
+    "id": 3,
+    "page": 3,
+    "bounding_box": [
+      179.8161277770996,
+      191.3740234375,
+      429.19012451171875,
+      243.114990234375
+    ],
+    "latex_content": "\\begin{table}[t]\n\\centering\n{\\fontsize{8}{6}\\selectfont\n\\setlength{\\tabcolsep}{.25em}\n\\vspace{-0.2in}\n\\begin{tabular}{|l|c|c|c|c|c|}\n\\hline\n&\\cite{slivkins2009contextual, dudik2011efficient, langford2007epoch, chu2011contextual} &  \\cite{hliu1, anandkumar, tekin2012sequencing} & \\cite{tekin4} & This work \\\\\n\\hline\nMulti-user & no & yes & yes & yes \\\\\n\\hline\nCooperative & N/A & yes & no & yes \\\\\n\\hline\nContextual & yes & no & no & yes \\\\\n\\hline\nData arrival  & arbitrary & i.i.d. or Markovian & i.i.d. & i.i.d or arbitrary \\\\\nprocess& & & & \\\\\n\\hline\nRegret & sublinear & logarithmic & may be linear & sublinear \\\\\n\\hline\n\\end{tabular}\n}\n\\caption{Comparison with related work in multi-armed bandits}\n\\vspace{-0.35in}\n\\label{tab:comparison2}\n\\end{table}",
+    "extracted_content": [
+      [
+        "",
+        "[16]\u2013[19]",
+        "[23]\u2013[25]",
+        "[26]",
+        "This work"
+      ],
+      [
+        "Multi-user",
+        "no",
+        "yes",
+        "yes",
+        "yes"
+      ],
+      [
+        "Cooperative",
+        "N/A",
+        "yes",
+        "no",
+        "yes"
+      ],
+      [
+        "Contextual",
+        "yes",
+        "no",
+        "no",
+        "yes"
+      ],
+      [
+        "Data arrival\nprocess",
+        "arbitrary",
+        "i.i.d. or Markovian",
+        "i.i.d.",
+        "i.i.d or arbitrary"
+      ],
+      [
+        "Regret",
+        "sublinear",
+        "logarithmic",
+        "may be linear",
+        "sublinear"
+      ]
+    ],
+    "similarity_score": 0.6435045317220544,
+    "table_image": "images/1307.0781v1/table_3.png",
+    "page_image": "pages/1307.0781v1/page_3.png"
+  }
+]

images/1307.8013v1/gt.json

@@ -0,0 +1,182 @@
+[
+  {
+    "id": 0,
+    "page": 4,
+    "bounding_box": [
+      54.0,
+      258.2130126953125,
+      302.4639892578125,
+      521.8250122070312
+    ],
+    "latex_content": "\\begin{table}[hbtp]\n\\caption{Scenarios of Data Analysis.}\\label{algorithm_scenarios}\n\\centering\n\\begin{tabular}{|p{2cm}|p{3.0cm}|p{2.5cm}|} \\hline\n  Name& Domain &Scenarios \\\\ \\hline\n          &search engine     &  \tLog analysis \\\\ \\cline{3-3}\n \tGrep    & social network  &Web information extraction \\\\ \\cline{3-3}\n \t        & electronic commerce  &Fuzzy search  \\\\ \\hline\n  Bayes & social network  & Spam recognition\\\\ \\cline{3-3}\n             & electronic commerce\t& Web page classification \\\\ \\hline\n           & social network      & Image Processing \\\\ \\cline{3-3}\n \tSVM       & electronic commerce   & Data Mining \\\\ \\cline{3-3}\n \t          &      & Text Categorization \\\\ \\hline\n  PageRank &search engine & Compute the page rank \\\\ \\hline\n  Fuzzy  &search engine    & Image processing \\\\ \\cline{3-3}\n  K-means& social network &High-resolution landform\\\\\n  K-means        & electronic commerce & classification \\\\ \\hline %\\cline{2-2}\n         % &     & Speech recognition \\\\ \\hline\n         & social network   &  \tSpeech recognition \\\\ \\cline{3-3}\n   HMM   & search engine  & \tWord Segmentation \\\\ \\cline{3-3}\n          &  & \tHandwriting recognition \\\\ \\hline\n             &search engine   & Word frequency count \\\\ \\cline{3-3}\n  WordCount & social network   & Calculating the TF-IDF value \\\\ \\cline{3-3}\n \t             & electronic commerce &Obtaining the user operations count\\\\ \\hline\n%  Vector calculation & Similarity calculation \\\\ \\cline{2-2}\n%    & Bioinformatics \\\\ \\hline\n  Sort & electronic commerce & Document sorting\\\\ \\cline{3-3}\n\t     & search engine & Pages sorting \\\\\n       & social network &  \\\\ \\hline\n %  & computer vision \\\\ \\cline{2-2}\n%dimension reduction &pattern recognition \\\\ \\cline{2-2}\n%  &face recognition \\\\ \\hline\n\\end{tabular}\n\\end{table}",
+    "extracted_content": [
+      [
+        "Name",
+        "Domain",
+        "Scenarios"
+      ],
+      [
+        "Grep",
+        "search engine\nsocial network\nelectronic commerce",
+        "Log analysis"
+      ],
+      [
+        null,
+        null,
+        "Web information ex-\ntraction"
+      ],
+      [
+        null,
+        null,
+        "Fuzzy search"
+      ],
+      [
+        "Bayes",
+        "social network\nelectronic commerce",
+        "Spam recognition"
+      ],
+      [
+        null,
+        null,
+        "Web page classifica-\ntion"
+      ],
+      [
+        "SVM",
+        "social network\nelectronic commerce",
+        "Image Processing"
+      ],
+      [
+        null,
+        null,
+        "Data Mining"
+      ],
+      [
+        null,
+        null,
+        "Text Categorization"
+      ],
+      [
+        "PageRank",
+        "search engine",
+        "Compute the page\nrank"
+      ],
+      [
+        "Fuzzy\nK-means\nK-means",
+        "search engine\nsocial network\nelectronic commerce",
+        "Image processing"
+      ],
+      [
+        null,
+        null,
+        "High-resolution land-\nform\nclassification"
+      ],
+      [
+        "HMM",
+        "social network\nsearch engine",
+        "Speech recognition"
+      ],
+      [
+        null,
+        null,
+        "Word Segmentation"
+      ],
+      [
+        null,
+        null,
+        "Handwriting recogni-\ntion"
+      ],
+      [
+        "WordCount",
+        "search engine\nsocial network\nelectronic commerce",
+        "Word frequency count"
+      ],
+      [
+        null,
+        null,
+        "Calculating the TF-\nIDF value"
+      ],
+      [
+        null,
+        null,
+        "Obtaining the user\noperations count"
+      ],
+      [
+        "Sort",
+        "electronic commerce\nsearch engine\nsocial network",
+        "Document sorting"
+      ],
+      [
+        null,
+        null,
+        "Pages sorting"
+      ]
+    ],
+    "similarity_score": 0.3920750782064651,
+    "table_image": "images/1307.8013v1/table_0.png",
+    "page_image": "pages/1307.8013v1/page_4.png"
+  },
+  {
+    "id": 1,
+    "page": 4,
+    "bounding_box": [
+      339.5639953613281,
+      264.19097900390625,
+      533.4359741210938,
+      377.96434529622394
+    ],
+    "latex_content": "\\begin{table}\n\\caption{Details of Hardware Configurations.}\\label{hwconfigeration}\n\\center\n\\begin{tabular}{|c|c|}\n  \\hline\n  %\\multicolumn{2}{|c|}{CPU Type} & \\multicolumn{2}{|c|}{Intel CPU Core} \\\\ \\hline\n  %\\multicolumn{2}{|c|}{Intel \\textregistered Xeon}  &\\multicolumn{2}{|c|}{4cores@1.6G} \\\\  \\hline\n  CPU Type & Intel \\textregistered Xeon E5645\\\\ \\hline\n  \\# Cores & 6 cores@2.4G \\\\ \\hline\n  \\# threads& 12 threads \\\\ \\hline\n\t\\#Sockets & 2 \\\\ \\hline\n  \\hline\n  ITLB & 4-way set associative, 64 entries \\\\ \\hline\n  DTLB & 4-way set associative, 64 entries \\\\ \\hline\n  L2 TLB& 4-way associative, 512 entries \\\\ \\hline\n  L1 DCache & 32KB, 8-way associative, 64 byte/line \\\\ \\hline\n  L1 ICache & 32KB, 4-way associative, 64 byte/line \\\\ \\hline\n  L2 Cache & 256 KB, 8-way associative, 64 byte/line \\\\ \\hline\n  L3 Cache &  12 MB, 16-way associative, 64 byte/line \\\\ \\hline\n  Memory & 32 GB , DDR3 \\\\  \\hline\n\\end{tabular}\n\\end{table}",
+    "extracted_content": [
+      [
+        "CPU Type",
+        "Intel\u20ddRXeon E5645"
+      ],
+      [
+        "# Cores",
+        "6 cores@2.4G"
+      ],
+      [
+        "# threads",
+        "12 threads"
+      ],
+      [
+        "#Sockets",
+        "2"
+      ],
+      [
+        "ITLB",
+        "4-way set associative, 64 entries"
+      ],
+      [
+        "DTLB",
+        "4-way set associative, 64 entries"
+      ],
+      [
+        "L2 TLB",
+        "4-way associative, 512 entries"
+      ],
+      [
+        "L1 DCache",
+        "32KB, 8-way associative, 64 byte/line"
+      ],
+      [
+        "L1 ICache",
+        "32KB, 4-way associative, 64 byte/line"
+      ],
+      [
+        "L2 Cache",
+        "256 KB, 8-way associative, 64 byte/line"
+      ],
+      [
+        "L3 Cache",
+        "12 MB, 16-way associative, 64 byte/line"
+      ],
+      [
+        "Memory",
+        "32 GB , DDR3"
+      ]
+    ],
+    "similarity_score": 0.8945054945054945,
+    "table_image": "images/1307.8013v1/table_1.png",
+    "page_image": "pages/1307.8013v1/page_4.png"
+  }
+]

images/1308.4565v2/gt.json

@@ -0,0 +1,508 @@
+[
+  {
+    "id": 0,
+    "page": 11,
+    "bounding_box": [
+      312.14712142944336,
+      53.33697509765625,
+      564.7523803710938,
+      120.38897705078125
+    ],
+    "latex_content": "\\begin{table}[t]\n\\centering\n{\\fontsize{8}{6}\\selectfont\n\\setlength{\\tabcolsep}{.1em}\n\\begin{tabular}{|l|c|c|c|c|c|}\n\\hline\nLearner & 1 &2 &3 & 4\\\\\n\\hline\nClassification  & Naive Bayes,  &Always $1$, & RBF Network, & Random Tree,  \\\\\nFunction (S1)  & Logistic & Voted Perceptron &J48 & Always $0$\\\\\n\\hline\n Error  & 47,  & 53,  & 47, & 47,  \\\\\npercentage (S1) & 3 & 4 &  47 & 47\\\\\n\\hline\n Classification  & Naive Bayes,  &Always $1$,  & RBF Network,  & Random Tree,  \\\\\nFunction (S2) & Random & Random & J48 & Always $0$ \\\\\n\\hline\n Error & 47,  & 53,  & 47,  & 47,  \\\\ \npercentage (S2) & 50 & 50 & 47 & 47 \\\\\n\\hline\n\\end{tabular}\n}\n\\add{\\vspace{-0.1in}}\n\\caption{Base classification functions used by the learners and their error percentages on the test data.}\n\\vspace{-0.25in}\n\\label{tab:sim_setup}\n\\end{table}",
+    "extracted_content": [
+      [
+        "Learner",
+        "1",
+        "2",
+        "3",
+        "4"
+      ],
+      [
+        "Classification\nFunction (S1)",
+        "Naive Bayes,\nLogistic",
+        "Always 1,\nVoted Perceptron",
+        "RBF Network,\nJ48",
+        "Random Tree,\nAlways 0"
+      ],
+      [
+        "Error\npercentage (S1)",
+        "47,\n3",
+        "53,\n4",
+        "47,\n47",
+        "47,\n47"
+      ],
+      [
+        "Classification\nFunction (S2)",
+        "Naive Bayes,\nRandom",
+        "Always 1,\nRandom",
+        "RBF Network,\nJ48",
+        "Random Tree,\nAlways 0"
+      ],
+      [
+        "Error\npercentage (S2)",
+        "47,\n50",
+        "53,\n50",
+        "47,\n47",
+        "47,\n47"
+      ]
+    ],
+    "similarity_score": 0.4544319600499376,
+    "table_image": "images/1308.4565v2/table_0.png",
+    "page_image": "pages/1308.4565v2/page_11.png"
+  },
+  {
+    "id": 1,
+    "page": 11,
+    "bounding_box": [
+      312.14712142944336,
+      152.52099609375,
+      564.7523803710938,
+      199.4580078125
+    ],
+    "latex_content": "\\begin{table}[t]\n\\centering\n{\\fontsize{8}{6}\\selectfont\n\\setlength{\\tabcolsep}{.3em}\n\\begin{tabular}{|l|c|c|c|c|c|c|}\n\\hline\n & $D_1(t)$ & $D_2(t)$ & $D_3(t)$ & $m_T$ & $A$ & $p$ \\\\\n\\hline\n(Z1) CoS & $t^{1/8} \\log t$ & $2 t^{1/8} \\log t$ & $t^{1/8} \\log t$ & $\\lceil T \\rceil^{1/4}$ & & \\\\\n\\hline\n(Z1) DCZA & $t^{1/8} \\log t$ & $2 t^{1/8} \\log t$ & $t^{1/8} \\log t$ &  & $1$ & $4$ \\\\\n\\hline\n(Z2) CoS & $t^{1/2} \\log t$ & $2 t^{1/2} \\log t$ & $t^{1/2} \\log t$ & $\\lceil T \\rceil^{1/4}$ & & \\\\\n\\hline\n(Z2) DCZA & $t^{2/p} \\log t$ & $2 t^{2/p} \\log t$ & $t^{2/p} \\log t$ &  & $1$ & $(3+\\sqrt{17})/2$ \\\\\n\\hline\n\\end{tabular}\n}\n\\add{\\vspace{-0.05in}}\n\\caption{Input parameters for CoS and DCZA for two different parameter sets Z1 and Z2.}\n\\label{tab:par_setup}\n\\add{\\vspace{-0.4in}}\n\\end{table}",
+    "extracted_content": [
+      [
+        "",
+        "D1(t)",
+        "D2(t)",
+        "D3(t)",
+        "mT",
+        "A",
+        "p"
+      ],
+      [
+        "(Z1) CoS",
+        "t1/8 log t",
+        "2t1/8 log t",
+        "t1/8 log t",
+        "\u2308T\u23091/4",
+        "",
+        ""
+      ],
+      [
+        "(Z1) DCZA",
+        "t1/8 log t",
+        "2t1/8 log t",
+        "t1/8 log t",
+        "",
+        "1",
+        "4"
+      ],
+      [
+        "(Z2) CoS",
+        "t1/2 log t",
+        "2t1/2 log t",
+        "t1/2 log t",
+        "\u2308T\u23091/4",
+        "",
+        ""
+      ],
+      [
+        "(Z2) DCZA",
+        "t2/p log t",
+        "2t2/p log t",
+        "t2/p log t",
+        "",
+        "1",
+        "\u221a\n(3 + 17)/2"
+      ]
+    ],
+    "similarity_score": 0.52,
+    "table_image": "images/1308.4565v2/table_1.png",
+    "page_image": "pages/1308.4565v2/page_11.png"
+  },
+  {
+    "id": 2,
+    "page": 12,
+    "bounding_box": [
+      56.45712375640869,
+      53.33697509765625,
+      286.24386978149414,
+      107.27099609375
+    ],
+    "latex_content": "\\begin{table}[t]\n\\centering\n{\\fontsize{8}{6}\\selectfont\n\\setlength{\\tabcolsep}{.3em}\n\\begin{tabular}{|l|c|c|}\n\\hline\n(Parameters) Algorithm & (S1) Error $\\%$  & (S2) Error $\\%$ \\\\\n\\hline\n(Z1) CoS (previous label as context) & 0.7 & 0.9  \\\\\n\\hline\n(Z1) DCZA (previous label as context) & 1.4 & 1.9  \\\\\n\\hline\nAdaBoost & 4.8 & 53  \\\\\n\\hline\n($w=100$) SWA  & 2.4 & 2.7 \\\\\n\\hline\n($w=1000$) SWA  & 11 & 11  \\\\\n\\hline\n(Z1) CoS (no-context) & 5.2 & 49.8  \\\\\n\\hline \n\\end{tabular}\n}\n\\add{\\vspace{-0.05in}}\n\\caption{Comparison of error percentages of CoS, DCZA, AdaBoost, SWA and CoS with no context.}\n\\label{tab:error_comp}\n\\vspace{-0.2in}\n%\\vspace{-0.4in}\n\\end{table}",
+    "extracted_content": [
+      [
+        "(Parameters) Algorithm",
+        "(S1) Error %",
+        "(S2) Error %"
+      ],
+      [
+        "(Z1) CoS (previous label as context)",
+        "0.7",
+        "0.9"
+      ],
+      [
+        "(Z1) DCZA (previous label as context)",
+        "1.4",
+        "1.9"
+      ],
+      [
+        "AdaBoost",
+        "4.8",
+        "53"
+      ],
+      [
+        "(w = 100) SWA",
+        "2.4",
+        "2.7"
+      ],
+      [
+        "(w = 1000) SWA",
+        "11",
+        "11"
+      ],
+      [
+        "(Z1) CoS (no-context)",
+        "5.2",
+        "49.8"
+      ]
+    ],
+    "similarity_score": 0.7660311958405546,
+    "table_image": "images/1308.4565v2/table_2.png",
+    "page_image": "pages/1308.4565v2/page_12.png"
+  },
+  {
+    "id": 3,
+    "page": 12,
+    "bounding_box": [
+      49.136778089735245,
+      146.8626708984375,
+      302.0871107313368,
+      208.0789794921875
+    ],
+    "latex_content": "\\begin{table}[t]\n\\centering\n{\\fontsize{8}{6}\\selectfont\n\\setlength{\\tabcolsep}{.3em}\n\\begin{tabular}{|l|c|c|c|}\n\\hline\n(Setting) & Error $\\%$ & Training $\\%$ & Exploration $\\%$  \\\\\nAlgorithm  & context=A1,A2,A3 & context=A1,A2,A3 & context=A1,A2,A3 \\\\\n\\hline\n(Z1,S1) CoS & 0.7, 4.6, 4.8 & 0.3, 3, 2.8 & 1.4, 6.3, 8.5 \\\\\n\\hline\n(Z1,S1) DCZA & 1.4, 3.5, 3.2 & 0.4, 1.3, 0.9 & 4, 5.9, 7 \\\\\n\\hline\n(Z1,S2) CoS & 0.9, 39, 10 & 0.3, 3, 2.8 & 1.5, 6.5, 8.6 \\\\\n\\hline\n(Z1,S2) DCZA & 1.9, 38, 4.8 & 0.4, 1.3, 1 & 4, 6, 7 \\\\\n\\hline\n(Z2,S1) CoS & 16, 14, 41 & 8.5, 16, 79 & 55 27 20\\\\\n\\hline\n(Z2,S1) DCZA & 31, 29, 29 & 33 19 87 & 66 66 12 \\\\\n\\hline \n\\end{tabular}\n}\n\\add{\\vspace{-0.05in}}\n\\caption{Error, training and exploration percentages of CoS and DCZA under different simulation and parameter settings. (A1) context as the previous label, (A2) context as srcbytes feature, (A3) context as time.}\n\\label{tab:sim_results}\n%\\vspace{-0.4in}\n\\vspace{-0.2in}\n\\end{table}",
+    "extracted_content": [
+      [
+        "(Setting)\nAlgorithm",
+        "Error %\ncontext=A1,A2,A3",
+        "Training %\ncontext=A1,A2,A3",
+        "Exploration %\ncontext=A1,A2,A3"
+      ],
+      [
+        "(Z1,S1) CoS",
+        "0.7, 4.6, 4.8",
+        "0.3, 3, 2.8",
+        "1.4, 6.3, 8.5"
+      ],
+      [
+        "(Z1,S1) DCZA",
+        "1.4, 3.5, 3.2",
+        "0.4, 1.3, 0.9",
+        "4, 5.9, 7"
+      ],
+      [
+        "(Z1,S2) CoS",
+        "0.9, 39, 10",
+        "0.3, 3, 2.8",
+        "1.5, 6.5, 8.6"
+      ],
+      [
+        "(Z1,S2) DCZA",
+        "1.9, 38, 4.8",
+        "0.4, 1.3, 1",
+        "4, 6, 7"
+      ],
+      [
+        "(Z2,S1) CoS",
+        "16, 14, 41",
+        "8.5, 16, 79",
+        "55 27 20"
+      ],
+      [
+        "(Z2,S1) DCZA",
+        "31, 29, 29",
+        "33 19 87",
+        "66 66 12"
+      ]
+    ],
+    "similarity_score": 0.6673228346456693,
+    "table_image": "images/1308.4565v2/table_3.png",
+    "page_image": "pages/1308.4565v2/page_12.png"
+  },
+  {
+    "id": 4,
+    "page": 12,
+    "bounding_box": [
+      315.0172058105469,
+      53.33697509765625,
+      557.1978149414062,
+      84.12298583984375
+    ],
+    "latex_content": "\\begin{table}[t]\n\\centering\n{\\fontsize{8}{6}\\selectfont\n\\setlength{\\tabcolsep}{.3em}\n\\begin{tabular}{|l|c|c|c|}\n\\hline\n(Setting) & Error $\\%$ & Training $\\%$ & Exploration $\\%$  \\\\\nAlgorithm & context=A1,A2,A3 & context=A1,A2,A3 & context=A1,A2,A3 \\\\\n\\hline\n(Z1,S1) CoS & 1.8, 4.1, 6.7 & 2, 9.2, 10.3 & 1.4, 3.6, 8.5 \\\\\n\\hline\n(Z1,S2) CoS & 24.6, 44.3, 31.3 & 2, 9.2, 10.3 & 1.4, 3.6, 8.5  \\\\\n\\hline\n\\end{tabular}\n}\n\\caption{Error, training and exploration percentages of CoS for  worst-case correlation between the learners for three different context types.}\n\\label{tab:sim_results2}\n\\vspace{-0.2in}\n%\\vspace{-0.4in}\n\\end{table}",
+    "extracted_content": [
+      [
+        "(Setting)\nAlgorithm",
+        "Error %\ncontext=A1,A2,A3",
+        "Training %\ncontext=A1,A2,A3",
+        "Exploration %\ncontext=A1,A2,A3"
+      ],
+      [
+        "(Z1,S1) CoS",
+        "1.8, 4.1, 6.7",
+        "2, 9.2, 10.3",
+        "1.4, 3.6, 8.5"
+      ],
+      [
+        "(Z1,S2) CoS",
+        "24.6, 44.3, 31.3",
+        "2, 9.2, 10.3",
+        "1.4, 3.6, 8.5"
+      ]
+    ],
+    "similarity_score": 0.5601374570446735,
+    "table_image": "images/1308.4565v2/table_4.png",
+    "page_image": "pages/1308.4565v2/page_12.png"
+  },
+  {
+    "id": 5,
+    "page": 12,
+    "bounding_box": [
+      315.0172058105469,
+      131.89801025390625,
+      557.1978149414062,
+      162.83367919921875
+    ],
+    "latex_content": "\\begin{table}[t]\n\\centering\n{\\fontsize{8}{6}\\selectfont\n\\setlength{\\tabcolsep}{.3em}\n\\begin{tabular}{|l|c|c|}\n\\hline\nensemble CoS & context-dependent weights & context-indep weights \\\\\nParameters: Z1 & S1, S2 & S1, S2 \\\\\n\\hline\ntotal error $\\%$ & 5.9, 10.2 & 3.8, 4.94  \\\\\n\\hline\nexploitation error $\\%$ & 2.9, 6.8 & 1.76, 2.17  \\\\\n\\hline\n\\end{tabular}\n}\n\\caption{Total error percentage, and error percentage of the errors made in exploitation steps for CoS with ensemble learner.}\n\\label{tab:weights}\n%\\vspace{-0.4in}\n\\vspace{-0.2in}\n\\end{table}",
+    "extracted_content": [
+      [
+        "ensemble CoS\nParameters: Z1",
+        "context-dependent weights\nS1, S2",
+        "context-indep weights\nS1, S2"
+      ],
+      [
+        "total error %",
+        "5.9, 10.2",
+        "3.8, 4.94"
+      ],
+      [
+        "exploitation error %",
+        "2.9, 6.8",
+        "1.76, 2.17"
+      ]
+    ],
+    "similarity_score": 0.5910064239828694,
+    "table_image": "images/1308.4565v2/table_5.png",
+    "page_image": "pages/1308.4565v2/page_12.png"
+  },
+  {
+    "id": 6,
+    "page": 13,
+    "bounding_box": [
+      75.31724739074707,
+      53.33697509765625,
+      270.87100982666016,
+      76.99033610026042
+    ],
+    "latex_content": "\\begin{table}[t]\n\\centering\n{\\fontsize{8}{6}\\selectfont\n\\setlength{\\tabcolsep}{.1em}\n\\begin{tabular}{|l|c|c|c|c|c|}\n\\hline\n(Setting) Algorithm /$p_r$ & 1 & 0.5 & 0.1 & 0.01 \\\\\n\\hline\n(Z1,S2) CoS (context is time) error $\\%$ & 10 & 13.9 & 36.4 & 47.1 \\\\\n\\hline\n(Z1,S2) DCZA (context is time) error $\\%$ & 4.8 & 4.8 & 16.3 & 56.6 \\\\\n\\hline\n\\end{tabular}\n}\n\\caption{Error percentages of CoS and DCZA as a function of $p_r$ (probability of receiving the label at each time slot) when context is time.}\n\\vspace{-0.2in}\n\\label{tab:errorperc}\n\\end{table}",
+    "extracted_content": [
+      [
+        "(Setting) Algorithm /pr",
+        "1",
+        "0.5",
+        "0.1",
+        "0.01"
+      ],
+      [
+        "(Z1,S2) CoS (context is time) error %",
+        "10",
+        "13.9",
+        "36.4",
+        "47.1"
+      ],
+      [
+        "(Z1,S2) DCZA (context is time) error %",
+        "4.8",
+        "4.8",
+        "16.3",
+        "56.6"
+      ]
+    ],
+    "similarity_score": 0.6533333333333333,
+    "table_image": "images/1308.4565v2/table_6.png",
+    "page_image": "pages/1308.4565v2/page_13.png"
+  },
+  {
+    "id": 7,
+    "page": 13,
+    "bounding_box": [
+      115.25924873352051,
+      127.00900268554688,
+      230.9290008544922,
+      150.31732177734375
+    ],
+    "latex_content": "\\begin{table}[t]\n\\centering\n{\\fontsize{8}{6}\\selectfont\n\\setlength{\\tabcolsep}{.1em}\n\\begin{tabular}{|l|c|c|c|c|c|}\n\\hline\n$\\#$ of learners & 1 & 2 & 3 & 4 \\\\\n\\hline\nCoS error $\\%$ & 49.8 & 49.7 & 50.2 & 22.3 \\\\\n\\hline\nDCZA error $\\%$ & 49.8 & 49.8 & 49.8 & 22.7 \\\\\n\\hline\n\\end{tabular}\n}\n\\caption{Error percentages of CoS and DCZA for learner 1, as a function of the number of learners present in the system.}\n\\vspace{-0.2in}\n\\label{tab:nlearn}\n\\end{table}",
+    "extracted_content": [
+      [
+        "# of learners",
+        "1",
+        "2",
+        "3",
+        "4"
+      ],
+      [
+        "CoS error %",
+        "49.8",
+        "49.7",
+        "50.2",
+        "22.3"
+      ],
+      [
+        "DCZA error %",
+        "49.8",
+        "49.8",
+        "49.8",
+        "22.7"
+      ]
+    ],
+    "similarity_score": 0.5503355704697986,
+    "table_image": "images/1308.4565v2/table_7.png",
+    "page_image": "pages/1308.4565v2/page_13.png"
+  },
+  {
+    "id": 8,
+    "page": 13,
+    "bounding_box": [
+      321.277837117513,
+      53.33697509765625,
+      550.9371643066406,
+      91.61798095703125
+    ],
+    "latex_content": "\\begin{table}[t]\n\\centering\n{\\fontsize{8}{6}\\selectfont\n\\setlength{\\tabcolsep}{.1em}\n\\begin{tabular}{|l|c|c|c|c|c|}\n\\hline\nd & error $\\%$ & training $\\%$  & selection (except training/exploration) $\\%$  \\\\\n& & of learners 2,3,4 & of learners 1,2,3,4  \\\\ \n\\hline\n0 & 0.9 & 0.27, 0.23, 0.16 & 52.9, 47, 0.1, 0\\\\\n\\hline\n0.5 & 1 & 0.27, 0.23, 0.16 & 53, 47, 0, 0 \\\\\n\\hline\n0.7 & 23.7 &0.27, 0.23, 0.16 & 100, 0, 0, 0\\\\\n\\hline\n\\end{tabular}\n}\n\\caption{Error and arm selection percentages as a function of calling cost}\n\\vspace{-0.25in}\n\\label{tab:callcost}\n\\end{table}",
+    "extracted_content": [
+      [
+        "d",
+        "error %",
+        "training %\nof learners 2,3,4",
+        "selection (except training/exploration) %\nof learners 1,2,3,4"
+      ],
+      [
+        "0",
+        "0.9",
+        "0.27, 0.23, 0.16",
+        "52.9, 47, 0.1, 0"
+      ],
+      [
+        "0.5",
+        "1",
+        "0.27, 0.23, 0.16",
+        "53, 47, 0, 0"
+      ],
+      [
+        "0.7",
+        "23.7",
+        "0.27, 0.23, 0.16",
+        "100, 0, 0, 0"
+      ]
+    ],
+    "similarity_score": 0.7475538160469667,
+    "table_image": "images/1308.4565v2/table_8.png",
+    "page_image": "pages/1308.4565v2/page_13.png"
+  },
+  {
+    "id": 9,
+    "page": 13,
+    "bounding_box": [
+      328.08721313476565,
+      127.00900268554688,
+      544.1278076171875,
+      159.4329833984375
+    ],
+    "latex_content": "\\begin{table}[t]\n\\centering\n{\\fontsize{8}{6}\\selectfont\n\\setlength{\\tabcolsep}{.1em}\n\\begin{tabular}{|l|c|c|c|c|}\n\\hline\n(Setting) Algorithm & previous label (A1) & srcbytes (A2) & time   \\\\\n& is context & is context & is context \\\\\n\\hline\n(Z1,S1) CoS error $\\%$ & 2.68 & 3.64 & 6.43  \\\\\n\\hline\n(Z1,S2) CoS error $\\%$ & 23.8 & 42.6 & 29 \\\\\n\\hline\n\\end{tabular}\n}\n\\caption{Error percentages of CoS for learner 1, when learner 1 only sends its context information to the other learners.}\n\\vspace{-0.3in}\n\\label{tab:onlycontext}\n\\end{table}",
+    "extracted_content": [
+      [
+        "(Setting) Algorithm",
+        "previous label (A1)\nis context",
+        "srcbytes (A2)\nis context",
+        "time\nis context"
+      ],
+      [
+        "(Z1,S1) CoS error %",
+        "2.68",
+        "3.64",
+        "6.43"
+      ],
+      [
+        "(Z1,S2) CoS error %",
+        "23.8",
+        "42.6",
+        "29"
+      ]
+    ],
+    "similarity_score": 0.6311111111111111,
+    "table_image": "images/1308.4565v2/table_9.png",
+    "page_image": "pages/1308.4565v2/page_13.png"
+  },
+  {
+    "id": 10,
+    "page": 3,
+    "bounding_box": [
+      49.1465003490448,
+      53.33697509765625,
+      305.9354419708252,
+      162.17901611328125
+    ],
+    "latex_content": "\\begin{table}[t]\n\\centering\n{\\renewcommand{\\arraystretch}{0.6}\n{\\fontsize{8}{7}\\selectfont\n\\setlength{\\tabcolsep}{.1em}\n\\begin{tabular}{|l|c|c|c|c|c|}\n\\hline\n&  \\cite{breiman1996bagging, buhlmann2003boosting, lazarevic2001distributed, chen2004channel, perlich2011cross} & \\cite{mateos2010distributed, kargupta1999collective} &  \\cite{zheng2011attribute} & This work \\\\\n\\hline\nAggregation & non-cooperative & cooperative & cooperative & \\rev{no} \\\\\n\\hline\nMessage  & none & data & training  & data and label \\\\\nexchange & & & residual & only if improves  \\\\\n& & & &   performance \\\\\n\\hline\nLearning  & offline/online & offline & offline & Non-bayesian \\\\\napproach&&&& online\\\\\n\\hline\nCorrelation & N/A & no & no & yes\\\\\nexploitation & & & &\\\\\n\\hline\nInformation from  & no & all & all & only if improves  \\\\\nother learners & & & &  accuracy \\\\\n\\hline\nData partition & horizontal & horizontal & vertical & horizontal \\\\\n&&&& and vertical \\\\\n\\hline\nBound on regret,  & no &no &no &yes - sublinear\\\\\nconvergence rate &&&&\\\\\n\\hline\n\\end{tabular}\n}\n}\n\\caption{Comparison with related work in distributed data mining.}\n\\label{tab:comparison1}\n\\vspace{-0.2in}\n\\end{table}",
+    "extracted_content": [
+      [
+        "",
+        "[6], [11], [16]\u2013[18]",
+        "[10], [12]",
+        "[8]",
+        "This work"
+      ],
+      [
+        "Aggregation",
+        "non-cooperative",
+        "cooperative",
+        "cooperative",
+        "no"
+      ],
+      [
+        "Message\nexchange",
+        "none",
+        "data",
+        "training\nresidual",
+        "data and label\nonly if improves\nperformance"
+      ],
+      [
+        "Learning\napproach",
+        "offline/online",
+        "offline",
+        "offline",
+        "Non-bayesian\nonline"
+      ],
+      [
+        "Correlation\nexploitation",
+        "N/A",
+        "no",
+        "no",
+        "yes"
+      ],
+      [
+        "Information from\nother learners",
+        "no",
+        "all",
+        "all",
+        "only if improves\naccuracy"
+      ],
+      [
+        "Data partition",
+        "horizontal",
+        "horizontal",
+        "vertical",
+        "horizontal\nand vertical"
+      ],
+      [
+        "Bound on regret,\nconvergence rate",
+        "no",
+        "no",
+        "no",
+        "yes - sublinear"
+      ]
+    ],
+    "similarity_score": 0.43297252289758537,
+    "table_image": "images/1308.4565v2/table_10.png",
+    "page_image": "pages/1308.4565v2/page_3.png"
+  }
+]