license: cc-by-sa-4.0
task_categories:
- text-classification
- table-question-answering
language:
- en
pretty_name: Indian Contracts in Adjudicated Texts
size_categories:
- n<1K
Indian Contracts in Adjudicated Texts or "ICAT" is a dataset generated with the help of an automated pipeline involving text segregation and classification.
Version 1 of this dataset released at schematise/ICAT-version1
About version 1:
- Expert annotated.
- Data sources validated by PDFs from Court websites.
- PDFs for all judgments shared alongside for data originality from truly public domain data sources.
- Used for the text-classification model that is part of the query pipeline to generate more data.
Important Links
To make queries to this pipeline you can visit the HuggingFace space hosted at this link. **Please note that the queries made to the streamlit app are stored anonymously in a relational database.
To see some data analytics on a public jupyter notebook hosted on #DeepNote you can access this link
To see details on how the dataset came to be prepared, please visit the owner's blog at - sankalpsrv.in/blog
File Structure
The training file (filename - "train.csv") contains a set of contract clauses taken from the datasheet for training a text-classification model.
The datasheet (filename - "CombinedDatasheet.ods") contains a classification matrix with contract clauses from judgments (n=60) with annotated discussions for further machine learning training and inference.
The judgment PDFs zip folder (filename - "PDFs.zip") contains a list of judgments which contain contractual clauses.