Skip to content
Roshil Shrestha
All projects

Proof of concept

Document Verification & Classification

A Streamlit app that classifies uploaded documents, predicts dispute codes from comments and extracts entities.

NLP

Overview

A proof-of-concept Streamlit application for verifying and classifying documents. Users upload PDFs and enter queries; NLP models predict dispute codes, classify the document type and extract the relevant entities.

The problem

Checking a dispute means reading documents of many different types and matching them to the right dispute code by hand. The app needed to handle varied inputs without a dedicated model for every document type.

What I did

  • Developed the Streamlit application and its NLP pipeline.
  • Fine-tuned RoBERTa for dispute code prediction.

How it works

  • Fine-tuned RoBERTa sequence classification that predicts dispute codes from user comments.
  • Zero-shot BART classification of document types, without training a model per type.
  • GLiNER for named entity recognition.
  • PDF text extraction with PyMuPDF.

Outcome

  • One interface to upload a PDF, ask a question, and get the dispute code, document type and entities back.