Enhancing Equity Research with AI: A LangChain-Based News Analysis Framework

Abstract

The advancement in chatbot technology and large-scale data processing has significantly transformed financial analysis and equity research and many other things. This research presents a LangChain-based framework to process financial documents, including company reports and market trends, to generate actionable investment insights. By integrating state-of-the-art technologies such as Large Language Models (LLMs) and advanced Natural Language Processing (NLP) techniques, the framework will support efficient data extraction, semantic analysis, and response generation. The proposed system begins by extracting key information from PDF documents, such as company financials and stock performance reports, which are then processed and segmented into manageable text chunks. These chunks are embedded into high dimensional vectors using techniques like Word2Vec or Doc2Vec, allowing the system to capture semantic relationships and store them in a semantic knowledge base. The knowledge base is further enhanced with tools like FAISS for efficient similarity search and information retrieval. In the second phase, the system responds to user queries by analyzing the context and questions posed. It retrieves relevant information from the knowledge base and generates responses using a generative AI model like GPT, ensuring high relevance and accuracy. The research also compares the system’s efficiency in answering various types of investment-related questions, showcasing the chatbot\'s capability to assist users in making informed decisions. The framework’s versatility and scalability, supported by cutting edge AI models and semantic search, demonstrate its potential to revolutionize the way equity research is conducted, providing financial analysts and investors with a more efficient and accessible tool for market analysis in the digital age.

Introduction

The project introduces a LangChain-powered, LLM-based news research and equity analysis chatbot. It processes news articles and financial documents to generate summaries and context-specific answers. The system is inspired by equity research chatbots and uses modular tools like text loaders, chunking, embeddings, and FAISS-based vector search to retrieve and generate insights from large datasets.

Literature Review

Recent studies show that NLP and LLMs are effective in analyzing financial texts and generating personalized investment insights. Previous models lacked real-time adaptability and user-specific responses. This project addresses that gap by combining LLMs with generative AI, focusing on intuitive interfaces and accurate, context-aware responses.

Methodology

The approach consists of two main phases:

Phase I: Knowledge Base Creation

1. Data Extraction: Financial PDFs are parsed using PyPDF.

2. Chunking: Large texts are split into manageable sizes due to LLM token limits.

3. Embedding: Chunks are converted into semantic vectors (e.g., Word2Vec).

4. Indexing: FAISS organizes vectors for fast semantic search.

5. Knowledge Base: All processed data is stored for query matching.

Phase II: User Query Response

1. Prompt Segmentation: User input is divided into context and query.

2. Query Embedding: Input is vectorized and chunked.

3. Similarity Matching: FAISS retrieves semantically relevant data.

4. Response Generation: A generative model composes domain-specific answers.

5. Ranking: Answers are ranked, and the best one is shown to the user.

Results & Discussion

The chatbot's performance was tested using Infosys' 2023–24 quarterly transcripts. Questions were categorized into:

1. Fact/Number-Based: Highly accurate and fast responses.

2. Behavioral Brief: Answered if discussed in transcripts; otherwise, model indicated lack of data.

3. Investment Advice: Declined due to regulatory constraints.

The analysis showed that:

• Fact-based questions had the highest accuracy and fastest responses.

• Investment advice queries had the slowest response time and lowest accuracy.

• There is an inverse relationship between response time and accuracy.

Conclusion

This research presents a comprehensive LangChain based framework for enhancing equity research through AI-powered news analysis. The system demonstrates significant improvements in information retrieval accuracy, response quality, and user satisfaction compared to traditional approaches. The modular architecture ensures scalability and maintainability, while the semantic search capabilities enable nuanced understanding of financial documents. The experimental results validate the effectiveness of combining large language models with vector databases for financial analysis tasks. The system\'s ability to process vast amounts of financial data in real-time while maintaining contextual understanding represents a significant advancement in automated equity research. Key contributions include the development of a domain-specific embedding model for financial texts, an efficient retrieval system using FAISS, and a comprehensive evaluation framework for financial AI systems. The case studies demonstrate practical applicability across various financial use cases, from earnings analysis to regulatory compliance. While challenges remain, particularly in computational requirements and potential hallucination issues, the framework establishes a solid foundation for future developments in AI-powered financial analysis. The integration of explainable AI features and multimodal capabilities in future versions will further enhance the system\'s utility for financial professionals. The proposed framework has the potential to democratize access to sophisticated financial analysis tools, enabling smaller firms and individual investors to leverage advanced AI capabilities previously available only to large institutions. As the financial industry continues its digital transformation, such AI powered tools will become increasingly essential for maintaining competitive advantage.

References

[1] Y. Zhang, J. Zhao and Y. Wang, \"Stock Movement Prediction Using News Sentiment Analysis,\" IEEE Access, vol. 7, pp. 37649–37658, 2019. [2] R. Hu, Y. Liu, and M. Bian, \"Financial Sentiment Analysis for the Chinese Market Based on LSTM,\" Journal of Computational Science, vol. 44, pp. 101– 122, 2020. [3] S. Ding, H. Zhang, Z. Liu, and Y. Zhang, \"Deep Learning for Event-Driven Stock Prediction,\" in Proceedings of IJCAI, 2015, pp. 2327–2333. [4] J. Chen, Y. Sun, and W. Li, \"A Personalized News Recommendation System Using LSTM-Based Sentiment Analysis,\" in IEEE Int’l Conf. on Big Data (Big Data), 2018. [5] A. Vaswani et al., \"Attention Is All You Need,\" in Advances in Neural Information Processing Systems (NeurIPS), 2017. [6] OpenAI, GPT-3.5 Technical Report, [Online]. Available: https://openai.com/research/gpt-4, 2023. [7] A. Jain and B. Singh, \"Conversational AI in Finance: A Review,\" ACM Transactions on Management Information Systems, vol. 12, no. 4, pp. 1–22, 2021. [8] S. R. Varghese and A. Gupta, \"Trust and Engagement in Financial Chatbots: A UX Perspective,\" in Proc. of the ACM Conference on Human Factors in Computing Systems (CHI), 2020.

Copyright

Copyright © 2025 Thejaswini K S, Yashwanth N, Vivek D, Chandana H S, Mrs. Sindhu K S. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.