{ "paper_id": "O12-1008", "header": { "generated_with": "S2ORC 1.0.0", "date_generated": "2023-01-19T08:03:16.754602Z" }, "title": "", "authors": [], "year": "", "venue": null, "identifiers": {}, "abstract": "", "pdf_parse": { "paper_id": "O12-1008", "_pdf_hash": "", "abstract": [], "body_text": [ { "text": "EQUATION", "cite_spans": [], "ref_spans": [], "eq_spans": [ { "start": 0, "end": 8, "text": "EQUATION", "ref_id": "EQREF", "raw_str": "h one-of-N i w l N i w l 1 0 i j k (1) x i (t) t i i 1 N N \u23aa \u23a9 \u23aa \u23a8 \u23a7 \u2208 = 0 if 1 ) ( i t x i (1) h (t=3) 3N x x V V V (2) j v ji j i \u03b8 j j net j (t) j y j (t) j ) ( ) ( ) ( ) ( j j i j i ji j net f t y t x v t net = + = \u2211 \u03b8 (2) f(net j ) (Activation Function) 0 1 (Sigmoid Function)", "eq_num": "(3)" } ], "section": "", "sec_num": null }, { "text": "x e x f ) (", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "", "sec_num": null }, { "text": "EQUATION", "cite_spans": [], "ref_spans": [], "eq_spans": [ { "start": 0, "end": 8, "text": "EQUATION", "ref_id": "EQREF", "raw_str": "\u2212 + = 1 1 ) ( (3) W (4) ) ( ) ( ) ( ) ( k k j k j kj k net g t y t y w t net = + = \u2211 \u03b8 (4) w kj k j \u03b8 k k net k (t) k y k (t) k 1 g(net k ) (Softmax Activation Function) (Transfer Function)", "eq_num": "(5" } ], "section": "", "sec_num": null }, { "text": "EQUATION", "cite_spans": [], "ref_spans": [], "eq_spans": [ { "start": 0, "end": 8, "text": "EQUATION", "ref_id": "EQREF", "raw_str": "N l h i w ) | ( h l w P i = 1990", "eq_num": "(5)" } ], "section": "", "sec_num": null }, { "text": "Elman (Elman Networks) [3] (Jordan Networks) [4] (Bi-directional RNN) [ ", "cite_spans": [ { "start": 6, "end": 22, "text": "(Elman Networks)", "ref_id": null }, { "start": 23, "end": 26, "text": "[3]", "ref_id": "BIBREF2" }, { "start": 45, "end": 48, "text": "[4]", "ref_id": "BIBREF3" }, { "start": 70, "end": 71, "text": "[", "ref_id": null } ], "ref_spans": [], "eq_spans": [], "section": "", "sec_num": null }, { "text": "( ) N (Relevance Information) w 1 w 3 w 2 S 1 1 0 0 0 0 1 0 1 0 1 0 1 0 0 0 0 0 0 0 1 1 0 1 0 1 1 0 1 0 1 1 w R 3 w R 2 w R 1 w 3 w 2 w w 1 w 3 w 5 R 1 w 1 w 3 w 2 S 1 1 0 0 0 0 1 0 1 0 0 0 0 0 0 0 1 1 1 0 0 1 0 1 0 1 1 w R 3 w R 2 w R 1 w 3 w 2 w w 1 w 3 w 4 1 w R w 1 w 2 w 3 3 w R w 1 w 3 w 5 2 w R (Sentence Relevance Information) (Word Relevance Information) 1 S 1 S 1 w 3 w 2 w 1 R 1 S 1 w R 3 w R 2 w R 1 w 3 w 1 0 1 1 0 1 0 (6) \u23aa \u23a9 \u23aa \u23a8 \u23a7 \u22c5 + \u22c5 \u2212 = \u2264 < = = \u2212 t t t t R R R L t if R R if \u03b1 \u03b1 1 ' 0 ' ) 1 ( , 0 0, t (6) t R t ' t R L \u03b1 0 \u03b1 0.6 1 S w ( ) w 1 w 2 w 3 S 1 w 4 w 5 4 3 2 1 4 6 . 0 24 . 0 096 . 0 064 . 0 ' w w w w w R R R R R + + + =", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "", "sec_num": null }, { "text": "Proceedings of the Twenty-Fourth Conference on Computational Linguistics and Speech Processing (ROCLING 2012) ", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "", "sec_num": null }, { "text": "EQUATION", "cite_spans": [], "ref_spans": [], "eq_spans": [ { "start": 0, "end": 8, "text": "EQUATION", "ref_id": "EQREF", "raw_str": "(Unigram Word Vector) (Clustering) K (K-means)[7] S S (Mean Vector) (7) s L k k s s L v v s \u2211 = = 1 , (7) k s v , S k s v S s L s (Similarity) M 2 2 ) , cos( s k s k s k v u v u v U \u22c5 = (8) U k k k u k M s v s (1) (9) ) , cos( max arg s k s U v U RNNLM k = (9) k U RNNLM (9) P ) ( ) ( k RNNLM k U P U P K U \u2248 (10) k U (10) (2) S p k , \u03b3 \u2211 = = S s s k s k s k v u v u 1 ' ' , ) , cos( ) , cos( \u03b3 (11) ) ( ) ( 1 , k RNNLM S s s k k U P U P s \u2211 = \u22c5 = \u03b3 (12) ) ( k U P ) ( k RNNLMs U P s (3) (Uniform) s k, \u03b2 S s k 1 , = \u03b2 (13) S", "eq_num": "(14)" } ], "section": "", "sec_num": null }, { "text": "Proceedings of the Twenty-Fourth Conference on Computational Linguistics and Speech Processing(ROCLING 2012)", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "", "sec_num": null }, { "text": "w Proceedings of the Twenty-Fourth Conference on Computational Linguistics and Speech Processing (ROCLING 2012)", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "", "sec_num": null } ], "back_matter": [], "bib_entries": { "BIBREF0": { "ref_id": "b0", "title": "The problem of learning long-term dependencies in recurrent networks", "authors": [ { "first": "Y", "middle": [], "last": "Bengio", "suffix": "" }, { "first": "P", "middle": [], "last": "Frasconi", "suffix": "" }, { "first": "P", "middle": [], "last": "Simard", "suffix": "" } ], "year": 1993, "venue": "Proc. IEEE International Conference on Neural Networks", "volume": "3", "issue": "", "pages": "1183--1188", "other_ids": {}, "num": null, "urls": [], "raw_text": "Y. Bengio, P. Frasconi, and P. Simard, \"The problem of learning long-term dependencies in recurrent networks,\" in Proc. IEEE International Conference on Neural Networks, Vol. 3, pp. 1183-1188, 1993.", "links": null }, "BIBREF1": { "ref_id": "b1", "title": "A neural probabilistic language model", "authors": [ { "first": "Y", "middle": [], "last": "Bengio", "suffix": "" }, { "first": "R", "middle": [], "last": "Ducharme", "suffix": "" }, { "first": "P", "middle": [], "last": "Vincent", "suffix": "" }, { "first": "C", "middle": [], "last": "Jauvin", "suffix": "" }, { "first": "J", "middle": [ "K" ], "last": "", "suffix": "" }, { "first": "T", "middle": [], "last": "Hofmann", "suffix": "" }, { "first": "T", "middle": [], "last": "Poggio", "suffix": "" }, { "first": "J", "middle": [], "last": "Shawetaylor", "suffix": "" } ], "year": 2003, "venue": "Journal of Machine Learning Research", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "Y. Bengio, R. Ducharme, P. Vincent, C. Jauvin, J. K, T. Hofmann, T. Poggio, and J. Shawetaylor. A neural probabilistic language model. In Journal of Machine Learning Research, 2003.", "links": null }, "BIBREF2": { "ref_id": "b2", "title": "Finding structure in time", "authors": [ { "first": "J", "middle": [ "L" ], "last": "Elman", "suffix": "" } ], "year": 1990, "venue": "Proceedings of the Twenty-Fourth Conference on Computational Linguistics and Speech Processing", "volume": "14", "issue": "", "pages": "179--211", "other_ids": {}, "num": null, "urls": [], "raw_text": "J. L. Elman, \"Finding structure in time,\" Cognitive Science, Vol. 14, No. 2, pp. 179-211, 1990. Proceedings of the Twenty-Fourth Conference on Computational Linguistics and Speech Processing (ROCLING 2012)", "links": null }, "BIBREF3": { "ref_id": "b3", "title": "Attractor dynamics and parallelism in a connectionist sequential machine", "authors": [ { "first": "M", "middle": [ "L" ], "last": "Jordan", "suffix": "" } ], "year": 1986, "venue": "Proc. the eighth annual conference of the cognitive science society", "volume": "", "issue": "", "pages": "531--546", "other_ids": {}, "num": null, "urls": [], "raw_text": "M. L. Jordan, \"Attractor dynamics and parallelism in a connectionist sequential machine,\" in Proc. the eighth annual conference of the cognitive science society, pp.531-546, 1986", "links": null }, "BIBREF4": { "ref_id": "b4", "title": "Bidirectional recurrent neural networks", "authors": [ { "first": "M", "middle": [], "last": "Schuster", "suffix": "" }, { "first": "K", "middle": [ "K" ], "last": "Paliwal", "suffix": "" } ], "year": 1997, "venue": "IEEE Transactions on Signal Processing", "volume": "45", "issue": "11", "pages": "2673--2681", "other_ids": {}, "num": null, "urls": [], "raw_text": "M. Schuster and K. K. Paliwal, \"Bidirectional recurrent neural networks,\" IEEE Transactions on Signal Processing, Vol. 45, No. 11, pp. 2673-2681, 1997.", "links": null }, "BIBREF5": { "ref_id": "b5", "title": "Learning long-term dependencies with gradient descent is difficult", "authors": [ { "first": "Y", "middle": [], "last": "Bengio", "suffix": "" }, { "first": "P", "middle": [], "last": "Simard", "suffix": "" }, { "first": "P", "middle": [], "last": "Frasconi", "suffix": "" } ], "year": 1994, "venue": "IEEE Transaction on Neural Networks", "volume": "5", "issue": "2", "pages": "157--166", "other_ids": {}, "num": null, "urls": [], "raw_text": "Y. Bengio, P. Simard, and P. Frasconi, Learning long-term dependencies with gradient descent is difficult,\" IEEE Transaction on Neural Networks, Vol. 5, No. 2, pp. 157-166, 1994.", "links": null }, "BIBREF6": { "ref_id": "b6", "title": "Some methods for classification and analysis of multivariate observations", "authors": [ { "first": "J", "middle": [ "B" ], "last": "Macqueen", "suffix": "" } ], "year": null, "venue": "Proc. 5 th Berkeley Symposium on Mathematical Statistics and Probability", "volume": "", "issue": "", "pages": "281--297", "other_ids": {}, "num": null, "urls": [], "raw_text": "J. B. MacQueen, \"Some methods for classification and analysis of multivariate observations,\" in Proc. 5 th Berkeley Symposium on Mathematical Statistics and Probability, pp. 281-297.", "links": null }, "BIBREF7": { "ref_id": "b7", "title": "MATBN: A Mandarin Chinese broadcast news corpus", "authors": [ { "first": "H.-M", "middle": [], "last": "Wang", "suffix": "" }, { "first": "B", "middle": [], "last": "Chen", "suffix": "" }, { "first": "J.-W", "middle": [], "last": "Kuo", "suffix": "" }, { "first": "S.-S", "middle": [], "last": "Cheng", "suffix": "" } ], "year": 2005, "venue": "International Journal of Computational Linguistics & Chinese Language Processing", "volume": "10", "issue": "2", "pages": "219--236", "other_ids": {}, "num": null, "urls": [], "raw_text": "H.-M. Wang, B. Chen, J.-W. Kuo and S.-S. Cheng, \"MATBN: A Mandarin Chinese broadcast news corpus,\" International Journal of Computational Linguistics & Chinese Language Processing, Vol. 10, No. 2, pp. 219-236, 2005.", "links": null }, "BIBREF8": { "ref_id": "b8", "title": "Srilm -an extensible language modeling toolkit", "authors": [ { "first": "Andreas", "middle": [], "last": "Stolcke", "suffix": "" } ], "year": 2002, "venue": "Proceedings of the International Conference on Spoken Language Processing", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "Stolcke, Andreas. Srilm -an extensible language modeling toolkit. In Proceedings of the International Conference on Spoken Language Processing, Denver, Colorado, September 2002.", "links": null }, "BIBREF9": { "ref_id": "b9", "title": "Estimation of probabilities from sparse data for the language model component of a speech recognizer", "authors": [ { "first": "S", "middle": [ "M" ], "last": "Katz", "suffix": "" } ], "year": 1987, "venue": "Proc. IEEE Transactions on Acoustics, Speech, and Signal Processing", "volume": "35", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "S. M. Katz, \"Estimation of probabilities from sparse data for the language model component of a speech recognizer,\" in Proc. IEEE Transactions on Acoustics, Speech, and Signal Processing, Vol. ASSP-35, No. 3, pp. 400, 1987.", "links": null }, "BIBREF10": { "ref_id": "b10", "title": "RNNLM -Recurrent neural network language modeling toolkit", "authors": [ { "first": "T", "middle": [], "last": "Mikolov", "suffix": "" }, { "first": "S", "middle": [], "last": "Kombrink", "suffix": "" }, { "first": "A", "middle": [], "last": "Deoras", "suffix": "" }, { "first": "L", "middle": [], "last": "Burget", "suffix": "" }, { "first": "J", "middle": [], "last": "\u010cernock\u00fd", "suffix": "" } ], "year": 2011, "venue": "Proc. IEEE workshop on Automatic Speech Recognition and Understanding", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "T. Mikolov, S. Kombrink, A. Deoras, L. Burget and J. \u010cernock\u00fd, \"RNNLM -Recurrent neural network language modeling toolkit,\" in Proc. IEEE workshop on Automatic Speech Recognition and Understanding, 2011", "links": null } }, "ref_entries": { "TABREF0": { "html": null, "content": "
(Feature Extraction)(Input Layer)(Hidden Layer)(Output Layer)
(Projection Layer)
(Acoustic Model)(Language
Model)
N(Linguistic
Decoding)
(Synapse)N
(Neural Network Language Models, NNLM)
(Recurrent Neural Network Language Models, RNNLM)1994
[1]
( )
(Neural Networks)(Artificial Intelligence)
(Artificial Neural Networks, ANN)1940(Neuron)
(Perceptron)
( )
[2]
", "type_str": "table", "num": null, "text": "Proceedings of the Twenty-Fourth Conference on Computational Linguistics and Speech Processing(ROCLING 2012)" }, "TABREF2": { "html": null, "content": "
Proceedings of the Twenty-Fourth Conference on Computational Linguistics and Speech Processing (ROCLING 2012)
( )1()
30,600 Oracle(23100
)1,99893.22%31.5
1,9971.5
1.2%1.56%
9.52%
( ) RNN (Global) RNNLM(3) (RNN+BG) 232.31(4)(%) 85.67 236.97 100(M=100) 236.97(%) 85.17% 85.17(%) RNN -(%) -
( )0 0.1230.0584.29 234.93 85.6585.8884.29 85.36 85.55(Edit distance) 84.29 0.19 85.591.26
0.2230.13234.75 85.73Mikolov 85.83[11] 85.34 85.73Recurrent Neural 0.17 1.16 0.24% 85.71
Network Language Modeling Toolkit (RNNLM) 0.3 85.8685.8685.78
10.4230.12234.83 85.8785.7785.40 85.810.2385.691.52
0 RNNLM 0.585.7985.7785.60
0.685.6185.5085.52
0 0.7185.34% 84.29 85.3984.29 85.3584.29 85.35
0.1 0.885.64 85.1585.68 85.13100 85.58 85.17% 85.05
0.04% 0.2 0.910085.87 84.7140.08%85.88 84.66185.94 84.630.14%
0.3 185.86 82.7585.92 83.3985.94 83.31
0.4 0.5P) 85.90 ( k U = \u03b2 1 1 S s k \u2211 = 1 85.81,ss 0 rnnlm RNN P \u22c5( 0 U) 85.91 k 85.7885.81 85.74(14) 0.19%
0.17% 0.23% 0.61.26% 1.16% 1.52% 85.70 85.5785.44
0.7185.460(%) (%)(%) 85.43(%)(%)85.29(%) (%)
0.81 85.0585.3384.970.1684.99 1.06
0.9450.93459.06 84.6084.73 85.3183.61 84.55 0.14-84.52 0.94-
( ) (BG)182.5685.3282.620.1582.59 1.02
RNN607.07 85.5623.50(Mandarin Across Taiwan-Broadcast News, 82.31 82.41 -1.2 -7.32
MATBN)[8] RNN+BG232.31 85.452001 236.972003 85.67 (%) (%)85.17 (%) (%)1.56 (%) (%)(SLG) 9.52 (%) (%)
(PTS) Oracle RNN RNN ) 23 ) N197 2002 -236.97 236.97 1,997 ( 1.5 2001 232.31 -85.4 232.31 (RNN) 2001 ) 30,600 93.22 85.67 85.67 (%) 2002 2001 2002 92.66 85.17 85.17 1,998 ( 1.5 (%) 85.25 623.50 223.63 229.01 85.63 85.09 -0.08 --3 -30,600 ( 23 ---(%) ) 30,632 ( (Central News -0.56 230.51 236.45 85.71 85.21 0.04 0.27 85.3 6 85.41 0.24 1.60 85.35 9 229.72 234.35 85.84 85.26 0.09 0.58 85.24 0.07 0.46 12 231.42 236.19 85.83 85.34 0.17 1.14 85.29 0.12 0.82 N
Agency, CNA) (BG) 1 1226.04 85.2 229.98231.19 234.6285.56 85.86 (out-of-vocabulary, OOV) 85.03 85.34-0.14 0.17-0.95 1.16
0 0SRI Language Modeling Toolkit (SRILM)[9]
Katz Back-off[10]
(RNN+BG)
", "type_str": "table", "num": null, "text": "Proceedings of the Twenty-Fourth Conference on Computational Linguistics and Speech Processing(ROCLING 2012)" } } } }