Inspiration The project was inspired by the difficulties MSME innovators face in identifying existing patents and checking the similarity of their invention ideas before filing a patent application. What it does The system uses AI-based semantic similarity to retrieve relevant prior-art patents, display similarity scores, generate summaries, compare technical features, and estimate novelty. How we built it We designed a system using Python, FastAPI, React, Sentence Transformers, and FAISS to process patent documents, generate embeddings, and retrieve similar patents efficiently. Challenges we ran into Key challenges include processing lengthy patent documents, capturing technical meaning, retrieving relevant patents accurately, and generating reliable similarity explanations. Accomplishments that we're proud of We developed a clear architecture integrating semantic search, patent summarization, similarity analysis, and novelty estimation into a single proposed system. What we learned We gained knowledge of natural language processing, semantic embeddings, vector search, patent retrieval, and the application of AI to real-world innovation challenges. What's next for AI-based patent prior-art searching Future improvements include integrating live patent databases, improving retrieval accuracy, supporting multiple languages, and introducing claim-level and patent diagram comparisons.

Built With

Share this project:

Updates

Submission history