The idea
Semantic search can recognise related wording, while lexical search preserves exact names, identifiers and uncommon terms. Many collections contain both kinds of query. Combining the two can be useful, but the combination needs its own evaluation.
Design choice
Retrieve candidate passages from both methods and define how rankings are combined. Do not add raw scores unless their scales are comparable. Examine cases with product codes, abbreviations and phrases that have several meanings.
Worked example
A query for an exact error code should retain its lexical match, while a question describing the same symptom in ordinary language can benefit from semantic candidates.
Before you ship
- Keep separate baseline results
- Evaluate exact identifiers explicitly
- Inspect combined ranking failures
Examples are illustrative. Adapt the checks to your own environment.