# Chunks Dataset for CBSE

### Generation
- The data was generated using LLMSherpa with the following parameters
  - Use_OCR = False
  - strategy = 'sections'
- After the sectioned generation was done, each section was parsed through individual sections to see their size.
- If the size of a section was less than 5000 characters, subsequent sections were chunked together till they start to reach a size more than 5000 characters.


### Format

- All data is stored in the `data` directory
- Every file corresponds to a chapter that has been parsed
- Every file has the following columns ```Board,Class,Subject,ChunkData,Book```
- `Book` tracks the csv file to the book that has been used to create the data
- Some books might be missing because of bad chunking outputs, will use a separate framework to chunk them


### Recommendations
- Recommend using pandas to open the files as dataframe as it has worked seamlessly

### File names
- Each file has been named accoring to Class - Subject - Chapter_PDF_Name