Skip to main content

Advanced search

At a glance

Advanced search combines multiple strategies to find the most relevant information in your knowledge base. It leverages both meaning comprehension (semantic search) and exact term matching (lexical search).

Semantic searchLexical search
PrincipleUnderstands the meaning of the queryMatches exact words
SynonymsWorks ("car" finds "vehicle")Does not work (only the exact word is found)
TyposTolerantSensitive
Proper nounsLess preciseVery precise
Use caseOpen questions, explorationPrecise references, codes, identifiers

Example:

  • Semantic search: "How to handle customer returns?" finds a document titled "Refund and exchange policy".
  • Lexical search: "REF-2024-0042" finds exactly the document containing that reference.

Ontologie automatically combines both approaches for every search. The system:

  1. Runs both types of search in parallel.
  2. Merges the results by weighting each source according to relevance.
  3. Ranks the results by overall relevance score.

There is nothing to configure: hybrid search is active by default.

Available filters

Refine your results using the following filters:

FilterDescriptionExample
By sourceLimit to documents from a specific sourceOnly GitHub documents
By document typeFilter by file formatPDF, Markdown, web pages
Minimum scoreExclude results below a relevance thresholdScore above 0.5
Graph contextInclude related entities in resultsAdd ontology relations
TagsFilter by tags assigned to documentsTag "procedure" or "policy"

Hybrid weighting

The system automatically adjusts the weight of each search strategy based on the type of query:

  • Conceptual questions (e.g. "What is..."): semantic search is favored.
  • Factual questions (e.g. "What is the number of..."): lexical search is favored.
  • Mixed questions: both approaches contribute equally.

Practical examples

Finding a procedure

Query: "How to onboard a new supplier?"

The system finds relevant documents even if they are titled "Supplier onboarding process" or "Partner integration guide" thanks to semantic search.

Finding a specific document

Query: "Service note SN-2024-003"

Lexical search directly identifies the document with that exact reference, even if the document body does not mention those terms in the text.

Exploring a topic

Query: "What are the risks associated with outsourcing?"

Hybrid search combines documents explicitly discussing "outsourcing risks" with those covering related topics such as "contractor management" or "supplier compliance".

See also

Need help?

Contact us: Support and contact.