An Intelligent Semantic Search System For Context-Sensitive Query Interpretation
DOI:
https://doi.org/10.64252/vpjr4b23Keywords:
Semantic Search, Vector Database, Qdrant, Sentence Transformers, Vision Transformers, Multimodal Retrieval, Image Embeddings, Text Embeddings, Similarity Search, Streamlit InterfaceAbstract
This paper presents a Streamlit-based semantic search application that leverages transformer embeddings and a Qdrant vector database to enable context-aware information retrieval. Users can input or upload text, which is encoded into dense vector embeddings using Sentence Transformer models (such as mixedbread-ai/mxbai-embed-large-v1 and all-MiniLM-L6-v2). These embeddings are stored in Qdrant, an open-source vector similarity search engine, which organizes the data into collections with efficient indexing. At query time, the user’s natural language query is converted to a vector and compared against stored vectors using cosine similarity to retrieve the top K most semantically similar documents. The Streamlit frontend displays results and visualizes similarity scores (e.g. via bar charts) in real time. Custom UI elements (background image, styled containers) are used to create an engaging interface. The architecture, implementation, and user workflow are described in detail.
