Commit
·
a585869
1
Parent(s):
23cd368
Update README.md
Browse files
README.md
CHANGED
@@ -1,3 +1,56 @@
|
|
1 |
---
|
2 |
license: cc-by-nc-nd-4.0
|
3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
---
|
2 |
license: cc-by-nc-nd-4.0
|
3 |
---
|
4 |
+
|
5 |
+
### Dataset Generation:
|
6 |
+
|
7 |
+
Initially, we select the Amazon Review Dataset as our base data, referenced from Ni et al. (2019)[^1]. We randomly extract 100,000 instances from this dataset. The original labels in this dataset are ratings, scaled from 1 to 5. For our specific task, we categorize them into Positive (rating > 3), Neutral (rating = 3), and Negative (rating < 3), ensuring a balanced number of instances for each label. To generate the synthetic Code-mixed dataset, we apply two distinct methodologies: the Random Code-mixing Algorithm by Krishnan et al. (2021)[^2] and r-CM by Santy et al. (2021)[^3].
|
8 |
+
|
9 |
+
### Class Distribution:
|
10 |
+
|
11 |
+
#### For train.csv:
|
12 |
+
|
13 |
+
| Label | Count | Percentage |
|
14 |
+
|----------|-------|------------|
|
15 |
+
| Negative | 20000 | 33.33% |
|
16 |
+
| Neutral | 20000 | 33.33% |
|
17 |
+
| Positive | 19999 | 33.33% |
|
18 |
+
|
19 |
+
#### For dev.csv:
|
20 |
+
|
21 |
+
| Label | Count | Percentage |
|
22 |
+
|----------|-------|------------|
|
23 |
+
| Neutral | 6667 | 33.34% |
|
24 |
+
| Positive | 6667 | 33.34% |
|
25 |
+
| Negative | 6666 | 33.33% |
|
26 |
+
|
27 |
+
#### For test.csv:
|
28 |
+
|
29 |
+
| Label | Count | Percentage |
|
30 |
+
|----------|-------|------------|
|
31 |
+
| Negative | 6667 | 33.34% |
|
32 |
+
| Positive | 6667 | 33.34% |
|
33 |
+
| Neutral | 6666 | 33.33% |
|
34 |
+
|
35 |
+
### Cite our Paper:
|
36 |
+
|
37 |
+
If you utilize this dataset, kindly cite our paper.
|
38 |
+
|
39 |
+
```bibtex
|
40 |
+
@article{raihan2023mixed,
|
41 |
+
title={Mixed-Distil-BERT: Code-mixed Language Modeling for Bangla, English, and Hindi},
|
42 |
+
author={Raihan, Md Nishat and Goswami, Dhiman and Mahmud, Antara},
|
43 |
+
journal={arXiv preprint arXiv:2309.10272},
|
44 |
+
year={2023}
|
45 |
+
}
|
46 |
+
```
|
47 |
+
|
48 |
+
### References
|
49 |
+
|
50 |
+
[^1]: Ni, J., Li, J., & McAuley, J. (2019). Justifying recommendations using distantly-labeled reviews and fine-grained aspects. In Proceedings of the 2019 conference on empirical methods in natural language processing and the 9th international joint conference on natural language processing (EMNLP-IJCNLP) (pp. 188-197).
|
51 |
+
|
52 |
+
[^2]: Krishnan, J., Anastasopoulos, A., Purohit, H., & Rangwala, H. (2021). Multilingual code-switching for zero-shot cross-lingual intent prediction and slot filling. arXiv preprint arXiv:2103.07792.
|
53 |
+
|
54 |
+
[^3]: Santy, S., Srinivasan, A., & Choudhury, M. (2021). BERTologiCoMix: How does code-mixing interact with multilingual BERT? In Proceedings of the Second Workshop on Domain Adaptation for NLP (pp. 111-121).
|
55 |
+
|
56 |
+
---
|