The open-source pipeline that runs NER, relationship extraction, and GNN sentiment propagation over raw documents.
Parses emails, CSVs, and PDFs into coherent document streams.
Finds entities and SVO relationships via zero-shot models.
Merges mentions via cosine similarity over SBERT embeddings.
Propagates localized sentiment across the network topology.
Persists to Neo4j, triggers webhooks, serves D3 data.
Parse raw bank emails to build a merchant relationship graph. Detect spend clusters, hidden subscriptions, and macro expense trends automatically.
Ingest reviews and support tickets. Run GNN sentiment propagation to map how negative feedback on one feature impacts the perception of others.
Consume unstructured incident logs to assemble service dependency graphs. Detect cascade failures and route P0 alerts dynamically.
Process scientific papers or earnings calls with SONAR multilingual embeddings to map entity co-citation clusters across languages.