Datasets:

Languages:
English
ArXiv:
License:
dfki-nlp commited on
Commit
955acf1
1 Parent(s): 0b3a56a

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +15 -15
README.md CHANGED
@@ -10,7 +10,7 @@ license:
10
  - other
11
  multilinguality:
12
  - monolingual
13
- pretty_name: The TAC Relation Extraction Dataset
14
  size_categories:
15
  - 100K<n<1M
16
  source_datasets:
@@ -50,7 +50,7 @@ task_ids:
50
  ## Dataset Description
51
  - **Homepage:** [https://nlp.stanford.edu/projects/tacred](https://nlp.stanford.edu/projects/tacred)
52
  - **Paper:** [Position-aware Attention and Supervised Data Improve Slot Filling](https://aclanthology.org/D17-1004/)
53
- - **Point of Contact:** See [https://nlp.stanford.edu/projects/tacred/](https://nlp.stanford.edu/projects/tacred/)
54
  - **Size of downloaded dataset files:** 62.3 MB
55
  - **Size of the generated dataset:** 139.2 MB
56
  - **Total amount of disk used:** 201.5 MB
@@ -84,8 +84,8 @@ An example of 'train' looks as follows:
84
  "relation": "org:founded_by",
85
  "token": ["Tom", "Thabane", "resigned", "in", "October", "last", "year", "to", "form", "the", "All", "Basotho", "Convention", "-LRB-", "ABC", "-RRB-", ",", "crossing", "the", "floor", "with", "17", "members", "of", "parliament", ",", "causing", "constitutional", "monarch", "King", "Letsie", "III", "to", "dissolve", "parliament", "and", "call", "the", "snap", "election", "."],
86
  "subj_start": 10,
87
- "subj_end": 13,
88
- "obj_start": 0,
89
  "obj_end": 2,
90
  "subj_type": "ORGANIZATION",
91
  "obj_type": "PERSON",
@@ -100,7 +100,7 @@ The data fields are the same among all splits.
100
 
101
  - `id`: the instance id of this sentence, a `string` feature.
102
  - `docid`: the TAC KBP document id of this sentence, a `string` feature.
103
- - `tokens`: the list of tokens of this sentence, obtained with the StanfordNLP toolkit, a `list` of `string` features.
104
  - `relation`: the relation label of this instance, a `string` classification label.
105
  - `subj_start`: the 0-based index of the start token of the relation subject mention, an `ìnt` feature.
106
  - `subj_end`: the 0-based index of the end token of the relation subject mention, exclusive, an `ìnt` feature.
@@ -108,8 +108,8 @@ The data fields are the same among all splits.
108
  - `obj_start`: the 0-based index of the start token of the relation object mention, an `ìnt` feature.
109
  - `obj_end`: the 0-based index of the end token of the relation object mention, exclusive, an `ìnt` feature.
110
  - `obj_type`: the NER type of the object mention, among 23 fine-grained types used in the [Stanford NER system](https://stanfordnlp.github.io/CoreNLP/ner.html), a `string` feature.
111
- - `pos_tags`: the part-of-speech tag per token. the NER type of the subject mention, among 23 fine-grained types used in the [Stanford NER system](https://stanfordnlp.github.io/CoreNLP/ner.html), a `list` of `string` features.
112
- - `ner_tags`: the NER tags of tokens (IO-Scheme), among 23 fine-grained types used in the [Stanford NER system](https://stanfordnlp.github.io/CoreNLP/ner.html), a `list` of `string` features.
113
  - `stanford_deprel`: the Stanford dependency relation tag per token, a `list` of `string` features.
114
  - `stanford_head`: the head (source) token index (0-based) for the dependency relation per token. The root token has a head index of -1, a `list` of `int` features.
115
  ### Data Splits
@@ -127,11 +127,11 @@ To miminize dataset bias, TACRED is stratified across years in which the TAC KBP
127
  [More Information Needed]
128
  ### Annotations
129
  #### Annotation process
130
- See the Stanford paper and the Tacred Revisited paper, plus their appendices.
131
 
132
- To ensure that models trained on TACRED are not biased towards predicting false positives on real-world text,
133
- all sampled sentences where no relation was found between the mention pairs were fully annotated to be negative examples. As a result, 79.5% of the examples
134
- are labeled as no_relation.
135
  #### Who are the annotators?
136
  [More Information Needed]
137
  ### Personal and Sensitive Information
@@ -147,9 +147,9 @@ are labeled as no_relation.
147
  ### Dataset Curators
148
  [More Information Needed]
149
  ### Licensing Information
150
- To respect the copyright of the underlying TAC KBP corpus, TACRED is released via the
151
- Linguistic Data Consortium ([LDC License](https://catalog.ldc.upenn.edu/license/ldc-non-members-agreement.pdf)).
152
- You can download TACRED from the [LDC TACRED webpage](https://catalog.ldc.upenn.edu/LDC2018T24).
153
  If you are an LDC member, the access will be free; otherwise, an access fee of $25 is needed.
154
  ### Citation Information
155
  The original dataset:
@@ -182,4 +182,4 @@ For the revised version, please also cite:
182
  }
183
  ```
184
  ### Contributions
185
- #Thanks to [@github-username](https://github.com/<github-username>) for adding this dataset.
 
10
  - other
11
  multilinguality:
12
  - monolingual
13
+ pretty_name: tacred
14
  size_categories:
15
  - 100K<n<1M
16
  source_datasets:
 
50
  ## Dataset Description
51
  - **Homepage:** [https://nlp.stanford.edu/projects/tacred](https://nlp.stanford.edu/projects/tacred)
52
  - **Paper:** [Position-aware Attention and Supervised Data Improve Slot Filling](https://aclanthology.org/D17-1004/)
53
+ - **Point of Contact:** See [https://nlp.stanford.edu/projects/tacred/](https://nlp.stanford.edu/projects/tacred/)
54
  - **Size of downloaded dataset files:** 62.3 MB
55
  - **Size of the generated dataset:** 139.2 MB
56
  - **Total amount of disk used:** 201.5 MB
 
84
  "relation": "org:founded_by",
85
  "token": ["Tom", "Thabane", "resigned", "in", "October", "last", "year", "to", "form", "the", "All", "Basotho", "Convention", "-LRB-", "ABC", "-RRB-", ",", "crossing", "the", "floor", "with", "17", "members", "of", "parliament", ",", "causing", "constitutional", "monarch", "King", "Letsie", "III", "to", "dissolve", "parliament", "and", "call", "the", "snap", "election", "."],
86
  "subj_start": 10,
87
+ "subj_end": 13,
88
+ "obj_start": 0,
89
  "obj_end": 2,
90
  "subj_type": "ORGANIZATION",
91
  "obj_type": "PERSON",
 
100
 
101
  - `id`: the instance id of this sentence, a `string` feature.
102
  - `docid`: the TAC KBP document id of this sentence, a `string` feature.
103
+ - `token`: the list of tokens of this sentence, obtained with the StanfordNLP toolkit, a `list` of `string` features.
104
  - `relation`: the relation label of this instance, a `string` classification label.
105
  - `subj_start`: the 0-based index of the start token of the relation subject mention, an `ìnt` feature.
106
  - `subj_end`: the 0-based index of the end token of the relation subject mention, exclusive, an `ìnt` feature.
 
108
  - `obj_start`: the 0-based index of the start token of the relation object mention, an `ìnt` feature.
109
  - `obj_end`: the 0-based index of the end token of the relation object mention, exclusive, an `ìnt` feature.
110
  - `obj_type`: the NER type of the object mention, among 23 fine-grained types used in the [Stanford NER system](https://stanfordnlp.github.io/CoreNLP/ner.html), a `string` feature.
111
+ - `stanford_pos`: the part-of-speech tag per token. the NER type of the subject mention, among 23 fine-grained types used in the [Stanford NER system](https://stanfordnlp.github.io/CoreNLP/ner.html), a `list` of `string` features.
112
+ - `stanford_ner`: the NER tags of tokens (IO-Scheme), among 23 fine-grained types used in the [Stanford NER system](https://stanfordnlp.github.io/CoreNLP/ner.html), a `list` of `string` features.
113
  - `stanford_deprel`: the Stanford dependency relation tag per token, a `list` of `string` features.
114
  - `stanford_head`: the head (source) token index (0-based) for the dependency relation per token. The root token has a head index of -1, a `list` of `int` features.
115
  ### Data Splits
 
127
  [More Information Needed]
128
  ### Annotations
129
  #### Annotation process
130
+ See the Stanford paper and the Tacred Revisited paper, plus their appendices.
131
 
132
+ To ensure that models trained on TACRED are not biased towards predicting false positives on real-world text,
133
+ all sampled sentences where no relation was found between the mention pairs were fully annotated to be negative examples. As a result, 79.5% of the examples
134
+ are labeled as no_relation.
135
  #### Who are the annotators?
136
  [More Information Needed]
137
  ### Personal and Sensitive Information
 
147
  ### Dataset Curators
148
  [More Information Needed]
149
  ### Licensing Information
150
+ To respect the copyright of the underlying TAC KBP corpus, TACRED is released via the
151
+ Linguistic Data Consortium ([LDC License](https://catalog.ldc.upenn.edu/license/ldc-non-members-agreement.pdf)).
152
+ You can download TACRED from the [LDC TACRED webpage](https://catalog.ldc.upenn.edu/LDC2018T24).
153
  If you are an LDC member, the access will be free; otherwise, an access fee of $25 is needed.
154
  ### Citation Information
155
  The original dataset:
 
182
  }
183
  ```
184
  ### Contributions
185
+ #Thanks to [@github-username](https://github.com/<github-username>) for adding this dataset.