Semantic search allows retrieval of documents based on the meaning of a query, even if the exact keywords are not present. Topic extraction from documents helps improve semantic search by identifying relevant topics. Neo4j with GDS provides a rich toolset for topic-based semantic search, allowing representation of documents and topics as a knowledge graph and enabling efficient searches.
Table of contents
Topic Extraction with Neo4j GDS for Better Semantic Search in RAG ApplicationsA Dataset of Recent MoviesUsing an LLM to Extract ThemesClean up Themes and Generate Text EmbeddingsCluster Themes Using Neo4j Graph Data ScienceUse Traditional NLP Techniques to Find Themes That Share StemsExplore similar themes that don’t share stemsCreate IS_SIMILAR Relationships With the K Nearest Neighbors AlgorithmTest Weakly Connected Components CommunitiesLeiden Communities Can Be Tuned for Larger or Smaller CommunitiesSummarize Theme GroupsCompare Retrievers Based on Different Vector IndexesConclusion13 Impressions