Datasets:
Update README.md
Browse files
README.md
CHANGED
|
@@ -1,3 +1,53 @@
|
|
| 1 |
---
|
| 2 |
license: cc-by-sa-4.0
|
| 3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
---
|
| 2 |
license: cc-by-sa-4.0
|
| 3 |
---
|
| 4 |
+
|
| 5 |
+
This is a multilabel dataset used for Noise Identification purpose in the paper **"A Comparative Analysis of Noise Reduction Methods in Sentiment Analysis on Noisy Bangla Texts"** accepted in *2024 The 9th Workshop on Noisy and User-generated Text (W-NUT) collocated with EACL 2024*.
|
| 6 |
+
|
| 7 |
+
- Annotated by 4 native Bangla speakers with 90% trustworthiness score.
|
| 8 |
+
- Fleiss' Kappa Score: 0.69
|
| 9 |
+
|
| 10 |
+
## Definition of noise categories
|
| 11 |
+
|Type|Definition|
|
| 12 |
+
|-----|---------|
|
| 13 |
+
|**Local Word**|Any regional words even if there is a spelling error.|
|
| 14 |
+
|**Word Misuse**|Wrong use of words or unnecessary repetitions of words.|
|
| 15 |
+
|**Context/Word Missing**|Not enough information or missing words.|
|
| 16 |
+
|**Wrong Serial**|Wrong order of the words.|
|
| 17 |
+
|**Mixed Language**|Words in another language. Foreign words that were adopted into the Bangla language over time are excluded from this type.|
|
| 18 |
+
|**Punctuation Error**|Improper placement or missing punctuation. Sentences ending without "।" were excluded from this type.|
|
| 19 |
+
|**Spacing Error**|Improper use of white space.|
|
| 20 |
+
|**Spelling Error**|Words not following spelling of Bangla Academy Dictionary.|
|
| 21 |
+
|**Coined Word**|Emoji, symbolic emoji, link.|
|
| 22 |
+
|**Others**|Noises that do not fall into categories mentioned above.|
|
| 23 |
+
|
| 24 |
+
|
| 25 |
+
## Statistics of NC-SentNoB per noise class
|
| 26 |
+
|Class|Instances|#Word/Instance|
|
| 27 |
+
|---|---|---|
|
| 28 |
+
|**Local Word**|2,084 (0.136%)|16.05|
|
| 29 |
+
|**Word Misuse**|661 (0.043%)|18.55|
|
| 30 |
+
|**Context/Word Missing**|550 (0.036%)|13.19|
|
| 31 |
+
|**Wrong Serial**|69 (0.005%)|15.30
|
| 32 |
+
|**Mixed Language**|6,267 (0.410%)|17.91
|
| 33 |
+
|**Punctuation Error**|5,988 (0.391%)|17.25|
|
| 34 |
+
|**Spacing Error**|2,456 (0.161%)|18.78|
|
| 35 |
+
|**Spelling Error**|5,817 (0.380%)|17.30|
|
| 36 |
+
|**Coined Words**|549 (0.036%|15.45|
|
| 37 |
+
|**Others**|1,263 (0.083%)|16.52|
|
| 38 |
+
|
| 39 |
+
## Heatmap of correlation coefficient
|
| 40 |
+
<img src="https://huggingface.co/datasets/ktoufiquee/NC-SentNoB/resolve/main/corr_heatmap.png">
|
| 41 |
+
|
| 42 |
+
## Citation
|
| 43 |
+
If you use the datasets, please cite the following paper:
|
| 44 |
+
```
|
| 45 |
+
@misc{elahi2024comparative,
|
| 46 |
+
title={A Comparative Analysis of Noise Reduction Methods in Sentiment Analysis on Noisy Bangla Texts},
|
| 47 |
+
author={Kazi Toufique Elahi and Tasnuva Binte Rahman and Shakil Shahriar and Samir Sarker and Md. Tanvir Rouf Shawon and G. M. Shahariar},
|
| 48 |
+
year={2024},
|
| 49 |
+
eprint={2401.14360},
|
| 50 |
+
archivePrefix={arXiv},
|
| 51 |
+
primaryClass={cs.CL}
|
| 52 |
+
}
|
| 53 |
+
```
|