Datasets:

Modalities:
Text
Languages:
English
Libraries:
Datasets
License:
boudinfl commited on
Commit
830aaef
·
1 Parent(s): d79f45a

update files with spacy tokenizer

Browse files
Files changed (5) hide show
  1. README.md +21 -11
  2. dev.jsonl +0 -0
  3. prmu.py +99 -0
  4. test.jsonl +0 -0
  5. train.jsonl +0 -0
README.md CHANGED
@@ -5,25 +5,24 @@
5
  Inspec is a dataset for benchmarking keyphrase extraction and generation models.
6
  The dataset is composed of 2,000 abstracts of scientific papers collected from the [Inspec database](https://www.theiet.org/resources/inspec/).
7
  Keyphrases were annotated by professional indexers in an uncontrolled setting (that is, not limited to thesaurus entries).
8
- Details about the inspec dataset can be found in the original paper:
9
- - Anette Hulth. 2003.
10
- [Improved automatic keyword extraction given more linguistic knowledge](https://aclanthology.org/W03-1028).
11
- In Proceedings of the 2003 Conference on Empirical Methods in Natural Language Processing, pages 216-223.
12
-
13
- Reference (indexer-assigned) keyphrases are also categorized under the PRMU (<u>P</u>resent-<u>R</u>eordered-<u>M</u>ixed-<u>U</u>nseen) scheme as proposed in the following paper:
14
- - Florian Boudin and Ygor Gallina. 2021.
15
- [Redefining Absent Keyphrases and their Effect on Retrieval Effectiveness](https://aclanthology.org/2021.naacl-main.330/).
16
- In Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pages 4185–4193, Online. Association for Computational Linguistics.
17
 
18
- ## Content
 
 
 
 
 
 
 
19
 
20
  The dataset is divided into the following three splits:
21
 
22
  | Split | # documents | # keyphrases | % Present | % Reordered | % Mixed | % Unseen |
23
  | :--------- | ----------: | -----------: | --------: | ----------: | ------: | -------: |
24
  | Train | 1,000 | 9.79 | 77.83 | 9.90 | 6.30 | 5.98 |
25
- | Test | 500 | 9.83 | 78.49 | 9.82 | 6.76 | 4.92 |
26
  | Validation | 500 | 9.15 | 77.90 | 9.82 | 6.74 | 5.54 |
 
27
 
28
  The following data fields are available :
29
 
@@ -33,3 +32,14 @@ The following data fields are available :
33
  - **keyphrases**: list of reference keyphrases.
34
  - **prmu**: list of <u>P</u>resent-<u>R</u>eordered-<u>M</u>ixed-<u>U</u>nseen categories for reference keyphrases.
35
 
 
 
 
 
 
 
 
 
 
 
 
 
5
  Inspec is a dataset for benchmarking keyphrase extraction and generation models.
6
  The dataset is composed of 2,000 abstracts of scientific papers collected from the [Inspec database](https://www.theiet.org/resources/inspec/).
7
  Keyphrases were annotated by professional indexers in an uncontrolled setting (that is, not limited to thesaurus entries).
8
+ Details about the inspec dataset can be found in the original paper [(Hulth, 2003)][hulth-2003].
 
 
 
 
 
 
 
 
9
 
10
+
11
+ Reference (indexer-assigned) keyphrases are also categorized under the PRMU (<u>P</u>resent-<u>R</u>eordered-<u>M</u>ixed-<u>U</u>nseen) scheme as proposed in [(Boudin and Gallina, 2021)][boudin-2021].
12
+
13
+ Text pre-processing (tokenization) is carried out using `spacy` (`en_core_web_sm` model) with a special rule to avoid splitting words with hyphens (e.g. graph-based is kept as one token).
14
+ Stemming (Porter's stemmer implementation provided in `nltk`) is performed before reference keyphrases are matched against the source text.
15
+ Details about the process can be found in `prmu.py`.
16
+
17
+ ## Content and statistics
18
 
19
  The dataset is divided into the following three splits:
20
 
21
  | Split | # documents | # keyphrases | % Present | % Reordered | % Mixed | % Unseen |
22
  | :--------- | ----------: | -----------: | --------: | ----------: | ------: | -------: |
23
  | Train | 1,000 | 9.79 | 77.83 | 9.90 | 6.30 | 5.98 |
 
24
  | Validation | 500 | 9.15 | 77.90 | 9.82 | 6.74 | 5.54 |
25
+ | Test | 500 | 9.83 | 78.49 | 9.82 | 6.76 | 4.92 |
26
 
27
  The following data fields are available :
28
 
 
32
  - **keyphrases**: list of reference keyphrases.
33
  - **prmu**: list of <u>P</u>resent-<u>R</u>eordered-<u>M</u>ixed-<u>U</u>nseen categories for reference keyphrases.
34
 
35
+ ## References
36
+
37
+ - (Hulth, 2003) Anette Hulth. 2003.
38
+ [Improved automatic keyword extraction given more linguistic knowledge](https://aclanthology.org/W03-1028).
39
+ In Proceedings of the 2003 Conference on Empirical Methods in Natural Language Processing, pages 216-223.
40
+ - (Boudin and Gallina, 2021) Florian Boudin and Ygor Gallina. 2021.
41
+ [Redefining Absent Keyphrases and their Effect on Retrieval Effectiveness](https://aclanthology.org/2021.naacl-main.330/).
42
+ In Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pages 4185–4193, Online. Association for Computational Linguistics.
43
+
44
+ [hulth-2003]: https://aclanthology.org/W03-1028/
45
+ [boudin-2021]: https://aclanthology.org/2021.naacl-main.330/
dev.jsonl CHANGED
The diff for this file is too large to render. See raw diff
 
prmu.py ADDED
@@ -0,0 +1,99 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # -*- coding: utf-8 -*-
2
+
3
+ import sys
4
+ import json
5
+ import spacy
6
+
7
+ from nltk.stem.snowball import SnowballStemmer as Stemmer
8
+
9
+ nlp = spacy.load("en_core_web_sm")
10
+
11
+ # https://spacy.io/usage/linguistic-features#native-tokenizer-additions
12
+
13
+ from spacy.lang.char_classes import ALPHA, ALPHA_LOWER, ALPHA_UPPER
14
+ from spacy.lang.char_classes import CONCAT_QUOTES, LIST_ELLIPSES, LIST_ICONS
15
+ from spacy.util import compile_infix_regex
16
+
17
+ # Modify tokenizer infix patterns
18
+ infixes = (
19
+ LIST_ELLIPSES
20
+ + LIST_ICONS
21
+ + [
22
+ r"(?<=[0-9])[+\-\*^](?=[0-9-])",
23
+ r"(?<=[{al}{q}])\.(?=[{au}{q}])".format(
24
+ al=ALPHA_LOWER, au=ALPHA_UPPER, q=CONCAT_QUOTES
25
+ ),
26
+ r"(?<=[{a}]),(?=[{a}])".format(a=ALPHA),
27
+ # ✅ Commented out regex that splits on hyphens between letters:
28
+ # r"(?<=[{a}])(?:{h})(?=[{a}])".format(a=ALPHA, h=HYPHENS),
29
+ r"(?<=[{a}0-9])[:<>=/](?=[{a}])".format(a=ALPHA),
30
+ ]
31
+ )
32
+
33
+ infix_re = compile_infix_regex(infixes)
34
+ nlp.tokenizer.infix_finditer = infix_re.finditer
35
+
36
+
37
+ def contains(subseq, inseq):
38
+ return any(inseq[pos:pos + len(subseq)] == subseq for pos in range(0, len(inseq) - len(subseq) + 1))
39
+
40
+
41
+ def find_pmru(tok_title, tok_text, tok_kp):
42
+ """Find PRMU category of a given keyphrase."""
43
+
44
+ # if kp is present
45
+ if contains(tok_kp, tok_title) or contains(tok_kp, tok_text):
46
+ return "P"
47
+
48
+ # if kp is considered as absent
49
+ else:
50
+
51
+ # find present and absent words
52
+ present_words = [w for w in tok_kp if w in tok_title or w in tok_text]
53
+
54
+ # if "all" words are present
55
+ if len(present_words) == len(tok_kp):
56
+ return "R"
57
+ # if "some" words are present
58
+ elif len(present_words) > 0:
59
+ return "M"
60
+ # if "no" words are present
61
+ else:
62
+ return "U"
63
+
64
+
65
+ if __name__ == '__main__':
66
+
67
+ data = []
68
+
69
+ # read the dataset
70
+ with open(sys.argv[1], 'r') as f:
71
+ # loop through the documents
72
+ for line in f:
73
+ doc = json.loads(line.strip())
74
+
75
+ title_spacy = nlp(doc['title'])
76
+ abstract_spacy = nlp(doc['abstract'])
77
+
78
+ title_tokens = [token.text for token in title_spacy]
79
+ abstract_tokens = [token.text for token in abstract_spacy]
80
+
81
+ title_stems = [Stemmer('porter').stem(w.lower()) for w in title_tokens]
82
+ abstract_stems = [Stemmer('porter').stem(w.lower()) for w in abstract_tokens]
83
+
84
+ keyphrases_stems = []
85
+ for keyphrase in doc['keyphrases']:
86
+ keyphrase_spacy = nlp(keyphrase)
87
+ keyphrase_tokens = [token.text for token in keyphrase_spacy]
88
+ keyphrase_stems = [Stemmer('porter').stem(w.lower()) for w in keyphrase_tokens]
89
+ keyphrases_stems.append(keyphrase_stems)
90
+
91
+ prmu = [find_pmru(title_stems, abstract_stems, kp) for kp in keyphrases_stems]
92
+ doc['prmu'] = prmu
93
+
94
+ data.append(json.dumps(doc))
95
+ print(doc['id'])
96
+
97
+ # write the json
98
+ with open(sys.argv[2], 'w') as o:
99
+ o.write("\n".join(data))
test.jsonl CHANGED
The diff for this file is too large to render. See raw diff
 
train.jsonl CHANGED
The diff for this file is too large to render. See raw diff