Update README.md
Browse files
README.md
CHANGED
@@ -1,10 +1,71 @@
|
|
1 |
---
|
2 |
license: cc-by-4.0
|
3 |
configs:
|
4 |
-
|
5 |
-
|
6 |
-
|
7 |
-
|
8 |
-
|
9 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
10 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
---
|
2 |
license: cc-by-4.0
|
3 |
configs:
|
4 |
+
- config_name: default
|
5 |
+
data_files:
|
6 |
+
- split: train
|
7 |
+
path: train/data.jsonl
|
8 |
+
- split: test
|
9 |
+
path: test/data.jsonl
|
10 |
+
task_categories:
|
11 |
+
- text-classification
|
12 |
+
language:
|
13 |
+
- he
|
14 |
+
size_categories:
|
15 |
+
- 10K<n<100K
|
16 |
---
|
17 |
+
|
18 |
+
# HebrewSentiment - A Sentiment-Analysis Dataset in Hebrew
|
19 |
+
|
20 |
+
## Summary
|
21 |
+
HebrewSentiment is a Hebrew dataset for the sentiment analysis task.
|
22 |
+
|
23 |
+
## Introduction
|
24 |
+
|
25 |
+
This dataset was constructed via [To Fill In].
|
26 |
+
|
27 |
+
## Dataset Statistics
|
28 |
+
|
29 |
+
The table below shows the number of examples from each category in each of the splits:
|
30 |
+
|
31 |
+
| split | total | positive | negative | neutral |
|
32 |
+
|-------|----------|----------|----------|---------|
|
33 |
+
| train | 39,135 | 8,968 | 7,669 | 22,498 |
|
34 |
+
| test | 2,170 | 503 | 433 | 1,234 |
|
35 |
+
|
36 |
+
## Dataset Description
|
37 |
+
|
38 |
+
Each row in the dataset contains the following fields:
|
39 |
+
|
40 |
+
- **id**: A unique identifier for that training examples
|
41 |
+
- **text**: The textual content of the input sentence
|
42 |
+
- **tag_ids**: The label of the example (`Neutral`/`Positive`/`Negative`)
|
43 |
+
- **task_name**: [To fill in]
|
44 |
+
- **campaign_id**: [To fill in]
|
45 |
+
- **annotator_agreement_strength**: [To fill in]
|
46 |
+
- **survey_name**: [To fill in]
|
47 |
+
- **industry**: [To fill in]
|
48 |
+
- **type**: [To fill in]
|
49 |
+
|
50 |
+
## Models and Comparisons
|
51 |
+
|
52 |
+
In collaboration with [DICTA](https://dicta.org.il/) we trained a model on this dataset and are happy to release it to the public: [DictaBERT-Sentiment](https://huggingface.co/dicta-il/dictabert-sentiment).
|
53 |
+
|
54 |
+
In addition, we compared the performance of the model to the previous existing sentiment dataset - [Hebrew-Sentiment-Data from OnlpLab](https://github.com/OnlpLab/Hebrew-Sentiment-Data).
|
55 |
+
We fine-tuned [dictabert](https://huggingface.co/dicta-il/dictabert) 3 times - once on the OnlpLab dataset, once on this dataset, and once on both datasets together and the results can be seen in the table below:
|
56 |
+
|
57 |
+
| Training Corpus: | OnlpLab | | | | | HebrewSentiment| | | | |
|
58 |
+
|------------------|------|----------------|------|------|--------|--------------|------|------|---|---|
|
59 |
+
| | Accuracy | Macro F1 | F1 Positive | F1 Negative | F1 Neutral | Accuracy | Macro F1 | F1 Positive | F1 Negative | F1 Neutral |
|
60 |
+
| OnlpLab+HebrewSentiment | 87 | 61.7 | 93.2 | 74.6 | 17.4 | 83.9 | 82.7 | 79.8 | 81.8 | 86.4 |
|
61 |
+
| OnlpLab | 88.2 | 63.3 | 93.8 | 72.1 | 24 | 41.3 | 42.2 | 48.1 | 56.3 | 22.2 |
|
62 |
+
| HebrewSentiment | 69.9 | 51.7 | 82.2 | 62.9 | 10.2 | 84.4 | 83.2 | 81 | 82.1 | 86.6 |
|
63 |
+
|
64 |
+
## Contributors
|
65 |
+
|
66 |
+
[To fill in]
|
67 |
+
|
68 |
+
Contributors: [To fill in]
|
69 |
+
|
70 |
+
## Acknowledgments
|
71 |
+
We would like to express our gratitude to [To fill in]
|