updated readme
Browse files
README.md
CHANGED
@@ -33,17 +33,24 @@ Old terminology to standardized terminology translation:
|
|
33 |
|
34 |
Original data: https://www.dropbox.com/scl/fo/wwu0ifghw4sco09g67frb/h?rlkey=6ddg3yab9la3zeddvmnsfktxq&e=1&dl=0
|
35 |
|
36 |
-
Working doc: https://docs.google.com/document/d/1ZTiR5M7NBpsfYbJdlP6qcIldBJVzJI1Hm8srgI8hfnI/edit?usp=sharing
|
37 |
|
38 |
## Preprocessing
|
39 |
|
40 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
41 |
|
42 |
-
[Internal Kaggle](https://www.kaggle.com/competitions/ar-xive-project-baselines/overview) code: [clean_data_for_hugging_face.ipynb](https://huggingface.co/datasets/mlcore/arxiv-classifier/blob/main/clean_data_for_hugging_face.ipynb)
|
43 |
|
44 |
NOTE: We only include v1 of papers (to avoid any data leakage between train and test splits). There might be a better way of deduplicating (TODO).
|
45 |
|
46 |
-
## Todo
|
47 |
|
48 |
- Can we get the full text for the 2023 corpus/all papers?
|
49 |
-
- Can we incorporate additional metadata (e.g., authors)?
|
|
|
33 |
|
34 |
Original data: https://www.dropbox.com/scl/fo/wwu0ifghw4sco09g67frb/h?rlkey=6ddg3yab9la3zeddvmnsfktxq&e=1&dl=0
|
35 |
|
36 |
+
<!-- Working doc: https://docs.google.com/document/d/1ZTiR5M7NBpsfYbJdlP6qcIldBJVzJI1Hm8srgI8hfnI/edit?usp=sharing -->
|
37 |
|
38 |
## Preprocessing
|
39 |
|
40 |
+
Transform raw data into JSON format:
|
41 |
+
```
|
42 |
+
# major/minor cats data
|
43 |
+
python preprocess_major_minor.py -d <DATASET NAME> -s <SPLIT> -op <PATH TO SAVE PREPROCESSED DATA>
|
44 |
+
# all 2023 corpus (abstract only)
|
45 |
+
python preprocess_all2023.py
|
46 |
+
```
|
47 |
+
<!-- [4782 Kaggle](https://www.kaggle.com/competitions/cs-4782-2024/overview) code: [4782_preprocess_data.ipynb](4782_preprocess_data.ipynb)
|
48 |
|
49 |
+
[Internal Kaggle](https://www.kaggle.com/competitions/ar-xive-project-baselines/overview) code: [clean_data_for_hugging_face.ipynb](https://huggingface.co/datasets/mlcore/arxiv-classifier/blob/main/clean_data_for_hugging_face.ipynb) -->
|
50 |
|
51 |
NOTE: We only include v1 of papers (to avoid any data leakage between train and test splits). There might be a better way of deduplicating (TODO).
|
52 |
|
53 |
+
<!-- ## Todo
|
54 |
|
55 |
- Can we get the full text for the 2023 corpus/all papers?
|
56 |
+
- Can we incorporate additional metadata (e.g., authors)? -->
|