Built with datascienceportfol.io

Finance Insights App

Project photo

Project links

Skills

About this project

Financial Document Q&A System

Project Overview

This project is a Streamlit-based web application designed to analyze financial statements in Excel (.xlsx) and PDF (.pdf) formats. The system extracts key metrics — Revenue, Expenses, and Profit — from uploaded financial documents and allows users to query them interactively in natural language.

The app demonstrates skills in data extraction, preprocessing, regex-based parsing, and interactive dashboard development. It simplifies financial analysis by automating data lookup and providing instant answers.

Key Responsibilities

Data Extraction & Processing

  • Implemented PDF table extraction using pdfplumber to parse yearly financial data.

  • Handled Excel data ingestion with pandas and openpyxl.

  • Converted raw text into structured data (Year, Revenue, Expenses, Profit).

  • Performed typecasting and cleaning to ensure numeric consistency.

Natural Language Query Parsing

  • Built a simple regex-based Q&A system to interpret user questions.

  • Enabled detection of metrics (Revenue, Profit, Expenses) and time references (years).

  • Returned contextual answers or full trend tables if no specific year was mentioned.

Interactive App Development

  • Designed a Streamlit UI with:

    • File uploader for PDF/Excel documents

    • Text input box for user queries

    • Answer output section

  • Implemented feedback messages (upload success, missing data).

Exploratory Testing & Validation

  • Verified extraction accuracy with a 2000–2025 financial dataset.

  • Cross-checked consistency between Excel and PDF inputs.

  • Tested multiple queries, including year-specific (e.g., “Revenue in 2012”) and general (e.g., “Show me all profits”).

Business Insights & Use Cases

  • Faster Analysis: Automates manual lookup of financial data from long documents.

  • Accessibility: Enables non-technical users to interact with financial statements.

  • Scalability: Can be extended with data visualization (charts for trends) or NLP models for richer queries.

Key Results

  • Successfully extracted and structured 26 years of financial data (2000–2025).

  • App answered year-specific questions with 100% accuracy when data existed.

  • Provided tabular outputs for broader queries (all Revenues, Profits, or Expenses).

  • Demonstrated integration of file processing, regex parsing, and interactive UI in one workflow.


Challenges Addressed

  • Parsing tabular data consistently from unstructured PDFs.

  • Handling queries with different phrasings (e.g., "expenses" vs "expense").

  • Ensuring numeric fields were correctly formatted as integers.

  • Balancing simplicity with functionality for non-technical users.

Future Scope

  • Add data visualizations (trend charts for Revenue, Expenses, Profit).

  • Expand Q&A with NLP models for more natural language flexibility.

  • Integrate databases for persistent storage of multiple financial documents.

  • Deploy online with Streamlit Cloud / Docker for live demonstrations.