holylovenia
commited on
Commit
•
34d1875
1
Parent(s):
7f18583
Upload README.md with huggingface_hub
Browse files
README.md
ADDED
@@ -0,0 +1,78 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
license: mit
|
3 |
+
tags:
|
4 |
+
- coreference-resolution
|
5 |
+
language:
|
6 |
+
- ind
|
7 |
+
---
|
8 |
+
|
9 |
+
# indocoref
|
10 |
+
|
11 |
+
Dataset contains articles from Wikipedia Bahasa Indonesia which fulfill these conditions:
|
12 |
+
|
13 |
+
- The pages contain many noun phrases, which the authors subjectively pick: (i) fictional plots, e.g., subtitles for films,
|
14 |
+
|
15 |
+
TV show episodes, and novel stories; (ii) biographies (incl. fictional characters); and (iii) historical events or important events.
|
16 |
+
|
17 |
+
- The pages contain significant variation of pronoun and named-entity. We count the number of first, second, third person pronouns,
|
18 |
+
|
19 |
+
and clitic pronouns in the document by applying string matching.We examine the number
|
20 |
+
|
21 |
+
of named-entity using the Stanford CoreNLP
|
22 |
+
|
23 |
+
NER Tagger (Manning et al., 2014) with a
|
24 |
+
|
25 |
+
model trained from the Indonesian corpus
|
26 |
+
|
27 |
+
taken from Alfina et al. (2016).
|
28 |
+
|
29 |
+
The Wikipedia texts have length of 500 to
|
30 |
+
|
31 |
+
2000 words.
|
32 |
+
|
33 |
+
We sample 201 of pages from subset of filtered
|
34 |
+
|
35 |
+
Wikipedia pages. We hire five annotators who are
|
36 |
+
|
37 |
+
undergraduate student in Linguistics department.
|
38 |
+
|
39 |
+
They are native in Indonesian. Annotation is carried out using the Script d’Annotation des Chanes
|
40 |
+
|
41 |
+
de Rfrence (SACR), a web-based Coreference resolution annotation tool developed by Oberle (2018).
|
42 |
+
|
43 |
+
From the 201 texts, there are 16,460 mentions
|
44 |
+
|
45 |
+
tagged by the annotators
|
46 |
+
|
47 |
+
## Dataset Usage
|
48 |
+
|
49 |
+
Run `pip install nusacrowd` before loading the dataset through HuggingFace's `load_dataset`.
|
50 |
+
|
51 |
+
## Citation
|
52 |
+
|
53 |
+
```
|
54 |
+
@inproceedings{artari-etal-2021-multi,
|
55 |
+
title = {A Multi-Pass Sieve Coreference Resolution for {I}ndonesian},
|
56 |
+
author = {Artari, Valentina Kania Prameswara and Mahendra, Rahmad and Jiwanggi, Meganingrum Arista and Anggraito, Adityo and Budi, Indra},
|
57 |
+
year = 2021,
|
58 |
+
month = sep,
|
59 |
+
booktitle = {Proceedings of the International Conference on Recent Advances in Natural Language Processing (RANLP 2021)},
|
60 |
+
publisher = {INCOMA Ltd.},
|
61 |
+
address = {Held Online},
|
62 |
+
pages = {79--85},
|
63 |
+
url = {https://aclanthology.org/2021.ranlp-1.10},
|
64 |
+
abstract = {Coreference resolution is an NLP task to find out whether the set of referring expressions belong to the same concept in discourse. A multi-pass sieve is a deterministic coreference model that implements several layers of sieves, where each sieve takes a pair of correlated mentions from a collection of non-coherent mentions. The multi-pass sieve is based on the principle of high precision, followed by increased recall in each sieve. In this work, we examine the portability of the multi-pass sieve coreference resolution model to the Indonesian language. We conduct the experiment on 201 Wikipedia documents and the multi-pass sieve system yields 72.74{\%} of MUC F-measure and 52.18{\%} of BCUBED F-measure.}
|
65 |
+
}
|
66 |
+
```
|
67 |
+
|
68 |
+
## License
|
69 |
+
|
70 |
+
MIT
|
71 |
+
|
72 |
+
## Homepage
|
73 |
+
|
74 |
+
[https://github.com/valentinakania/indocoref/](https://github.com/valentinakania/indocoref/)
|
75 |
+
|
76 |
+
### NusaCatalogue
|
77 |
+
|
78 |
+
For easy indexing and metadata: [https://indonlp.github.io/nusa-catalogue](https://indonlp.github.io/nusa-catalogue)
|