Semantic gap
A semantic gap is a difference between intended or represented meaning and the forms or representations through which that meaning is expressed or processed.
Retrieval Augmented Generation, The Foundational Ideas
A semantic gap is a discrepancy between the meaning or representation of an input and the meaning or representation of a relevant target.
- How it is used
- Explains why direct dense retrieval can fail when queries and relevant documents use different levels of detail or vocabulary.
- How deep it goes
- Explains sparse-query and dense-document mismatch, vector separation, and zero-shot exacerbation, then motivates document-to-document matching.
- Examples
- Contrasts a search for “COVID-19 anxiety rates” with a detailed paper title about pandemic-related generalized anxiety disorder.
- What you can do after
- Diagnose why direct vector search can miss relevant documents when query and corpus language differ.
Read how Retrieval Augmented Generation, The Foundational Ideas explains semantic gap →
Inside AI
A semantic gap is the mismatch between data representations that a computational system can process and the deeper real-world meanings they signify.
- How it is used
- Frames why capable AI systems can process data yet fail to infer context, meaning, constraints, or intended significance.
- How deep it goes
- Defines the mismatch and illustrates it through handwriting and puzzle solving, but stops short of explaining implementation details.
- Examples
- Uses a handwritten 9, natural-language and image tasks, game playing, and the Farmer-Wolf-Goat-Cabbage riddle.
- What you can do after
- Evaluate AI claims by distinguishing pattern-based success from tasks requiring context, deduction, planning, or explicit constraints.
Also covered in
- LLM Evaluation and Alignment, The Foundational IdeasIn early access (MEAP)
Related concepts
- HyDEaddressed by · HyDE is a retrieval technique that generates hypothetical document or answer content from a query and uses that content or its embedding to retrieve relevant documents.
- semantic similarityaddressed by · Semantic similarity is the degree to which two pieces of content or representations correspond in meaning, regardless of differences in wording or form.
Questions about semantic gap
What is semantic gap?
A semantic gap is a difference between intended or represented meaning and the forms or representations through which that meaning is expressed or processed.
Which Manning books explain semantic gap?
Retrieval Augmented Generation, The Foundational Ideas and Inside AI.
What is the difference between semantic gap and semantic error?
A semantic gap is a difference between intended or represented meaning and the forms or representations through which that meaning is expressed or processed. A semantic error is a contradiction between the meaning or context depicted in media and its observable details.
These guides are generated from the text of Manning books by Manning's knowledge graph and link to the sections they draw on. The full books are available on liveBook and at manning.com.

