Datasets:

Languages:
English
Multilinguality:
monolingual
Size Categories:
unknown
Language Creators:
machine-generated
Annotations Creators:
expert-generated
Source Datasets:
original
License:
ruanchaves commited on
Commit
8b2dd6b
1 Parent(s): ac5d75e

Upload stan_large.py

Browse files
Files changed (1) hide show
  1. stan_large.py +98 -0
stan_large.py ADDED
@@ -0,0 +1,98 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ """STAN large dataset as corrected by Maddela et al.."""
2
+
3
+ import datasets
4
+ import pandas as pd
5
+ import pickle
6
+
7
+ _CITATION = """
8
+ @inproceedings{maddela-etal-2019-multi,
9
+ title = "Multi-task Pairwise Neural Ranking for Hashtag Segmentation",
10
+ author = "Maddela, Mounica and
11
+ Xu, Wei and
12
+ Preo{\c{t}}iuc-Pietro, Daniel",
13
+ booktitle = "Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics",
14
+ month = jul,
15
+ year = "2019",
16
+ address = "Florence, Italy",
17
+ publisher = "Association for Computational Linguistics",
18
+ url = "https://aclanthology.org/P19-1242",
19
+ doi = "10.18653/v1/P19-1242",
20
+ pages = "2538--2549",
21
+ abstract = "Hashtags are often employed on social media and beyond to add metadata to a textual utterance with the goal of increasing discoverability, aiding search, or providing additional semantics. However, the semantic content of hashtags is not straightforward to infer as these represent ad-hoc conventions which frequently include multiple words joined together and can include abbreviations and unorthodox spellings. We build a dataset of 12,594 hashtags split into individual segments and propose a set of approaches for hashtag segmentation by framing it as a pairwise ranking problem between candidate segmentations. Our novel neural approaches demonstrate 24.6{\%} error reduction in hashtag segmentation accuracy compared to the current state-of-the-art method. Finally, we demonstrate that a deeper understanding of hashtag semantics obtained through segmentation is useful for downstream applications such as sentiment analysis, for which we achieved a 2.6{\%} increase in average recall on the SemEval 2017 sentiment analysis dataset.",
22
+ }
23
+ """
24
+
25
+ _DESCRIPTION = """
26
+ The description below was taken from the paper "Multi-task Pairwise Neural Ranking for Hashtag Segmentation"
27
+ by Maddela et al..
28
+
29
+ "STAN large, our new expert curated dataset, which includes all 12,594 unique English hashtags and their
30
+ associated tweets from the same Stanford dataset.
31
+
32
+ STAN small is the most commonly used dataset in previous work. However, after reexamination, we found annotation
33
+ errors in 6.8% of the hashtags in this dataset, which is significant given that the error rate of the state-of-the art
34
+ models is only around 10%. Most of the errors were related to named entities. For example, #lionhead,
35
+ which refers to the “Lionhead” video game company, was labeled as “lion head”.
36
+
37
+ We therefore constructed the STAN large dataset of 12,594 hashtags with additional quality control for human annotations."
38
+ """
39
+ _URLS = {
40
+ "train": "https://github.com/prashantkodali/HashSet/raw/master/datasets/stan-large-maddela_et_al_train.pkl",
41
+ "dev": "https://github.com/prashantkodali/HashSet/raw/master/datasets/stan-large-maddela_et_al_dev.pkl",
42
+ "test": "https://github.com/prashantkodali/HashSet/raw/master/datasets/stan-large-maddela_et_al_test.pkl"
43
+ }
44
+
45
+ class StanLarge(datasets.GeneratorBasedBuilder):
46
+
47
+ VERSION = datasets.Version("1.0.0")
48
+
49
+ def _info(self):
50
+ return datasets.DatasetInfo(
51
+ description=_DESCRIPTION,
52
+ features=datasets.Features(
53
+ {
54
+ "index": datasets.Value("int32"),
55
+ "hashtag": datasets.Value("string"),
56
+ "segmentation": datasets.Value("string"),
57
+ "alternatives": datasets.Sequence(
58
+ {
59
+ "segmentation": datasets.Value("string")
60
+ }
61
+ )
62
+ }
63
+ ),
64
+ supervised_keys=None,
65
+ homepage="https://github.com/mounicam/hashtag_master",
66
+ citation=_CITATION,
67
+ )
68
+
69
+ def _split_generators(self, dl_manager):
70
+ downloaded_files = dl_manager.download(_URLS)
71
+ return [
72
+ datasets.SplitGenerator(name=datasets.Split.TRAIN, gen_kwargs={"filepath": downloaded_files["train"] }),
73
+ datasets.SplitGenerator(name=datasets.Split.VALIDATION, gen_kwargs={"filepath": downloaded_files["dev"] }),
74
+ datasets.SplitGenerator(name=datasets.Split.TEST, gen_kwargs={"filepath": downloaded_files["test"] }),
75
+ ]
76
+
77
+ def _generate_examples(self, filepath):
78
+
79
+ def get_segmentation(row):
80
+ return row["goldtruths"][0]
81
+
82
+ def get_alternatives(row):
83
+ segmentations = [{
84
+ "segmentation": x
85
+ } for x in row["goldtruths"]]
86
+
87
+ return segmentations[1:]
88
+
89
+ with open(filepath, 'rb') as f:
90
+ records = pickle.load(f)
91
+ records = records.to_dict("records")
92
+ for idx, row in enumerate(records):
93
+ yield idx, {
94
+ "index": idx,
95
+ "hashtag": row["hashtags"],
96
+ "segmentation": get_segmentation(row),
97
+ "alternatives": get_alternatives(row)
98
+ }