jina-financial-qa-search
Jina Financial QA Search is a production-ready BERT-based Financial Question Answering System built with Jina, PyTorch, and Hugging Face transformers. It uses a two-stage passage reranking approach: first retrieving the top 50 candidate answers, then reranking them using FinBERT-QA, a BERT model fine-tuned on the FiQA dataset that achieves state-of-the-art results. The system indexes answer passages from the FiQA dataset and searches for relevant passages given a financial question, addressing the financial industry's need for automatic analysis of unstructured data at scale. It provides competitive advantages for financial advisers by facilitating decision-making through efficient question answering. Key features include semantic search powered by neural information retrieval, BERT-based reranking with FinBERT-QA, integration with the Jina framework for scalable deployment, and support for indexing and querying answer passage collections. Use cases include financial question answering systems, semantic searc