DiceTechJobs/ConceptualSearch

Trains Word2Vec or LSA models on domain-specific documents using gensim, then generates Solr synonym files (with optional payload weighting) or vector clusters to enable semantic matching beyond keyword queries. The pipeline includes preprocessing, optional keyphrase extraction, model training, and multiple output formats—synonyms for query expansion or clustered vectors for field-based conceptual search without requiring custom plugins. Integrates directly with Apache Solr via synonym filters and optional custom plugins (PayloadEdismax, PayloadAwareDefaultSimilarity) for weighted semantic retrieval, or works with any search engine supporting synonym files.

259 stars. No commits in the last 6 months.