Meet us at the AI Summit, June 28th from 10 AM to 3 PM ET at the HLS Booth in the Industry Lounge. Book now
was successfully added to your cart.

Biomedical Question Answering

Combine information extracted from medical research, clinical trials, ontologies, and other sources into a knowledge graph – and answer natural language questions about the content.

Entities from unstructured text with NLP


Information to form a graph

Visualize & Explore & Discover

Facts from the knowledge graph

Roche Biomedical
Knowledge Graph
  • 10k+ relation types / 500k clinical trials / 85M entities / 260M facts / 5B+ additional facts
  • Build at scale using named entity extraction
  • Supports adding private facts into the graph
  • Tested on BioASQ challenge

Biomedical Q&A at Roche

  • 10k+ Predicates (Relation types)
  • 500k Clinical Trials
  • 85M Entities
  • 260M Extracted Facts from Text
  • 5B+ Total Facts
  • The end-to-end system includes a user interface for graph visualization, exploration, search, and question answering
  • Knowledge graph built at scale, combining multiple free-text data sources and ontologies
  • Supports adding private documents and facts to the graph
  • Tested on BioASQ challenge

The Clinical Knowledge Graph

Combining Spark NLP & Neo4j
  • Extracted named entities & relations between from
  • Typical questions:
  • What are the drug interactions and adverse drug events of “Lipitor”?
  • What are the routes by which the drug containing API “Rofecoxib” is administered?
Automated question answering about clinical guidelines
  • Joint work with Kaiser Permanente
  • Answering both specific and general questions (‘how to treat diabetes?’)
  • Automatic identification documents or section of documents with corresponding clinical guidelines – from curated set of clinical guidelines
  • Indexing free text documents including PDF and DOCX
  • Scalable to a large number of documents.