Datasets:

Modalities:
Text
Formats:
json
Languages:
English
ArXiv:
License:
yangwang825 commited on
Commit
d12936a
·
verified ·
1 Parent(s): 13d76d2

Rebuild from the original release with character offsets

Browse files
README.md CHANGED
@@ -1,32 +1,94 @@
1
  ---
 
 
 
 
2
  dataset_info:
3
- config_name: default
4
  features:
5
- - name: input
6
- dtype: string
7
- - name: output
8
- dtype: json
9
- - name: schema
10
- struct:
11
- - name: entities
12
- list:
13
- - name: label
14
- dtype: string
15
- - name: description
16
- dtype: string
17
- - name: relations
18
- list:
19
- - name: label
20
- dtype: string
21
- - name: description
22
- dtype: string
23
  configs:
24
- - config_name: default
25
- data_files:
26
- - split: train
27
- path: data/train.jsonl
28
- - split: validation
29
- path: data/validation.jsonl
30
- - split: test
31
- path: data/test.jsonl
 
 
 
 
 
 
 
 
32
  ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  ---
2
+ pretty_name: BioRED
3
+ license: other
4
+ language:
5
+ - en
6
  dataset_info:
7
+ - config_name: ner
8
  features:
9
+ - name: input
10
+ dtype: string
11
+ - name: output
12
+ dtype: json
13
+ - name: schema
14
+ dtype: json
15
+ - config_name: re
16
+ features:
17
+ - name: input
18
+ dtype: string
19
+ - name: output
20
+ dtype: json
21
+ - name: schema
22
+ dtype: json
 
 
 
 
23
  configs:
24
+ - config_name: ner
25
+ data_files:
26
+ - split: train
27
+ path: data/ner/train.jsonl
28
+ - split: validation
29
+ path: data/ner/validation.jsonl
30
+ - split: test
31
+ path: data/ner/test.jsonl
32
+ - config_name: re
33
+ data_files:
34
+ - split: train
35
+ path: data/re/train.jsonl
36
+ - split: validation
37
+ path: data/re/validation.jsonl
38
+ - split: test
39
+ path: data/re/test.jsonl
40
  ---
41
+
42
+ # BioRED
43
+
44
+ BioRED (named entity recognition and relation extraction), rebuilt for MNEB from the original release with character-level offsets.
45
+
46
+ **Source:** https://ftp.ncbi.nlm.nih.gov/pub/lu/BioRED/BIORED.zip (PubTator files)
47
+
48
+ | config | split | records |
49
+ | --- | --- | --- |
50
+ | ner | train | 751 |
51
+ | ner | validation | 191 |
52
+ | ner | test | 190 |
53
+ | re | train | 751 |
54
+ | re | validation | 191 |
55
+ | re | test | 190 |
56
+
57
+ ## Labels
58
+
59
+ **entities:** `gene or gene product`, `chemical entity`, `disease or phenotypic feature`, `sequence variant`, `organism taxon`, `cell line`
60
+
61
+ ## Format
62
+
63
+ Each record has `input` (the text), `output` (gold annotations grouped by label) and `schema` (the label set with descriptions). Offsets are character offsets into `input`, half-open `[start, end)`.
64
+
65
+ ```
66
+ "output": {
67
+ "entities": {"<label>": [{"text", "start", "end"}]},
68
+ "relations": {"<label>": [{"head": {"text", "start", "end"},
69
+ "tail": {"text", "start", "end"}}]},
70
+ "events": {"<label>": [{"trigger": {"text", "start", "end"},
71
+ "arguments": [{"role", "text", "start", "end"}]}]}
72
+ }
73
+ ```
74
+
75
+ Each schema item has a `description` and a `description_source`: `official` when adapted from the dataset's paper or guideline, `mneb` when written for MNEB. Relation datasets also give `setting` (`end_to_end` or `gold_entity`) and `negative_label`; relation items give `symmetric`. Event items list the `roles` valid for that event type.
76
+
77
+ Within a record, identical annotations are kept once and outer whitespace is trimmed from spans; source annotations are otherwise kept as annotated.
78
+
79
+ ## Notes
80
+
81
+ - Licence: public domain (US Government work, NCBI).
82
+ - Config `ner`: every mention. Config `re`: relations between given mentions, one representative mention per concept and piece (as in CalOn).
83
+ - Documents longer than 1,200 characters are cut at whitespace outside every annotated span (CalOn, arXiv:2610.17113); annotations whose parts fall in different pieces are dropped and counted.
84
+ - All relation types are non-directional (symmetric). Novelty labels are not kept.
85
+
86
+ ## Build
87
+
88
+ Built by `scripts/datasets/biored.py` in the MNEB repository.
89
+
90
+ ## Citation
91
+
92
+ Luo et al. (2022), BioRED: a rich biomedical relation extraction dataset. Briefings in Bioinformatics.
93
+
94
+ License: other
data/ner/test.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
data/ner/train.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
data/ner/validation.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
data/re/test.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
data/re/train.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
data/re/validation.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
data/test.jsonl DELETED
The diff for this file is too large to render. See raw diff
 
data/train.jsonl DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:34de47b2869d84a5befa344fd09cba30ca1053852eeecdf1e70f3dff8de36d79
3
- size 11587950
 
 
 
 
data/validation.jsonl DELETED
The diff for this file is too large to render. See raw diff
 
schema.json DELETED
@@ -1,62 +0,0 @@
1
- {
2
- "entities": [
3
- {
4
- "label": "gene or gene product",
5
- "description": "A gene, protein, mRNA, or other gene product mentioned in biomedical text."
6
- },
7
- {
8
- "label": "disease or phenotypic feature",
9
- "description": "A disease, symptom, syndrome, or disease-related phenotypic feature."
10
- },
11
- {
12
- "label": "chemical entity",
13
- "description": "A chemical substance, drug, or other chemical compound."
14
- },
15
- {
16
- "label": "organism taxon",
17
- "description": "An organism or taxonomic species."
18
- },
19
- {
20
- "label": "sequence variant",
21
- "description": "A genomic or protein sequence variant, including substitutions, insertions, deletions, and other mutations."
22
- },
23
- {
24
- "label": "cell line",
25
- "description": "An established cell line used in biomedical research."
26
- }
27
- ],
28
- "relations": [
29
- {
30
- "label": "association",
31
- "description": "A general biological association that cannot be classified into a more specific relation type."
32
- },
33
- {
34
- "label": "positive correlation",
35
- "description": "A positive association in which one entity promotes, induces, increases, or is positively correlated with the other."
36
- },
37
- {
38
- "label": "negative correlation",
39
- "description": "A negative association in which one entity inhibits, suppresses, treats, decreases, or protects against the other."
40
- },
41
- {
42
- "label": "bind",
43
- "description": "A direct physical binding or molecular interaction between two entities."
44
- },
45
- {
46
- "label": "comparison",
47
- "description": "An explicit comparison between two entities without asserting another specific relation."
48
- },
49
- {
50
- "label": "cotreatment",
51
- "description": "A relation indicating that two chemicals or drugs are administered together as part of the same treatment."
52
- },
53
- {
54
- "label": "drug interaction",
55
- "description": "A pharmacodynamic interaction between two chemicals or drugs."
56
- },
57
- {
58
- "label": "conversion",
59
- "description": "A relation indicating that one chemical is converted or transformed into another chemical."
60
- }
61
- ]
62
- }