Semantic similarity
Semantic similarity is the degree to which two pieces of content or representations correspond in meaning, regardless of differences in wording or form.
LLM Evaluation and Alignment, The Foundational Ideas
Semantic similarity is the degree to which texts express related meanings despite possible differences in wording.
- How it is used
- Uses semantic similarity to evaluate generated text against references when valid paraphrases make word overlap misleading.
- How deep it goes
- Explains contextual embeddings, cosine similarity, word matching, and metric selection, including factual and interpretability limitations.
- Examples
- Compares foreign-car paraphrases and contrasting sentences, and illustrates contextual meaning with running a company versus running a marathon.
- What you can do after
- Evaluate or understand generated text by selecting semantic and lexical metrics for the task and inspecting consequential factual errors separately.
Read how LLM Evaluation and Alignment, The Foundational Ideas explains semantic similarity →
Build an Advanced RAG Application (From Scratch)
Semantic similarity is the degree to which two pieces of text express related meaning, typically estimated from their representations.
- How it is used
- Supports comparing sentence meaning, matching paraphrased cache queries, and assessing retrieval candidates within RAG systems.
- How deep it goes
- Explains vector comparison and SBERT training architectures, then stops short of prescribing a universal similarity threshold or retrieval policy.
- Examples
- Uses Uber revenue paraphrases and mixed API-documentation and financial-filing searches to expose matching and false-retrieval behavior.
- What you can do after
- Evaluate semantic matches while separating meaning similarity from domain-appropriate retrieval.
Read how Build an Advanced RAG Application (From Scratch) explains semantic similarity →
AI-Powered Search
Semantic textual similarity is the task or measure of determining how closely two text segments correspond in meaning.
- How it is used
- Uses document co-occurrence to discover related terms, expand search queries, rank graph nodes, and generate content-based recommendations.
- How deep it goes
- Explains distributional reasoning, stop-word effects, document-set comparison, scoring, traversal, and query construction without deriving the full equation.
- Examples
- Demonstrates health-term discovery from “advil” and topic-based recommendations from a “star wars” document.
- What you can do after
- Apply it via a knowledge-graph traversal and search API to expand queries and recommend contextually similar documents.
Natural Language Processing in Action, Second Edition
Semantic similarity is the degree to which two language items have similar meanings or contextual associations.
- How it is used
- Supports semantic search and identifies related or redundant passages when wording differs.
- How deep it goes
- Explains embeddings, normalization, cosine matrices, heatmaps, and an approximate redundancy threshold, without detailing transformer internals.
- Examples
- Uses movie-plot retrieval, MiniLM sentence embeddings, pairwise similarity matrices, and a heatmap of manuscript passages.
- What you can do after
- Apply it via sentence-transformers and cosine similarity to retrieve related text or inspect possible redundancy.
Read how Natural Language Processing in Action, Second Edition explains semantic similarity →
Effective Conversational AI
Semantic similarity is the degree of shared meaning between two pieces of text, typically estimated from their representations.
- How it is used
- Finds relevant document passages for chatbot questions by comparing meaning rather than exact words.
- How deep it goes
- Explains vector-space matching and runtime retrieval, but does not specify a similarity formula or threshold.
- Examples
- Uses indexed document chunks and user questions, then compares generated answers with expected responses.
- What you can do after
- Evaluate response relevance by scoring semantic similarity and informativeness, comparing generated and expected answers, and reviewing precision, recall, and F1.
Read how Effective Conversational AI explains semantic similarity →
Math and Architectures of Deep Learning
Similarity of meaning between terms or documents, as opposed to exact string overlap.
- How it is used
- Measures document relatedness through shared topics when explicit term overlap misses meaningful connections.
- How deep it goes
- Explains semantic similarity through topics, subspaces, variance, and principal components.
- Examples
- Uses documents d5 and d6, the terms gun and violence, and a toy document-vector dataset.
- What you can do after
- Evaluate document relatedness through topic-based representations rather than relying only on explicit word overlap.
Read how Math and Architectures of Deep Learning explains semantic similarity →
Building Reliable AI Systems
A measure of how closely the meanings of two pieces of content correspond.
- How it is used
- Uses similarity to retrieve relevant passages and compare generated claims with grounding data.
- How deep it goes
- Explains spatial proximity and threshold filtering, but gives no similarity formula or scoring implementation.
- Examples
- Uses clustered policy documents, an international-roaming FAQ query, and claim comparisons against knowledge-base data.
- What you can do after
- Evaluate retrieved passages for relevance and compare generated claims with grounding data.
Read how Building Reliable AI Systems explains semantic similarity →
Retrieval Augmented Generation, The Foundational Ideas
Semantic similarity is the degree to which two linguistic items have similar meanings or conceptual content.
- How it is used
- Supports similarity search and forms a learned foundation for retrieval-augmented generation systems.
- How deep it goes
- Explains training signals, domain adaptation, and production limitations, but stops short of specifying model architecture or loss functions.
- Examples
- Contrasts million-document Sentence-BERT indexing with Word2Vec lookup and notes sentence-level context limits.
- What you can do after
- Apply via Sentence-BERT to perform similarity search while weighing domain, context-length, and indexing constraints.
Read how Retrieval Augmented Generation, The Foundational Ideas explains semantic similarity →
Also covered in
- Build AI-Enhanced Web Apps
- Introduction to Generative AI, Second Edition
- MongoDB 8.0 in Action, Third Edition
- The Art of AI Product Development
- Build an AI Agent (From Scratch)
- Building LLM Applications with DSPy
- AI Recommender SystemsIn early access (MEAP)
What to know first
- language modelA language model is a computational model that represents patterns in linguistic sequences to assign probabilities, predict, generate, or evaluate language.
- embeddingAn embedding is a numerical vector representation of an item in a feature space, encoding selected properties or relationships for computational tasks such as comparison or retrieval.
- knowledge baseA knowledge base is an organized body or representation of information that people or computational systems can consult to support retrieval, reference, reasoning, or tasks.
- BERTBERT (Bidirectional Encoder Representations from Transformers) is a pretrained Transformer encoder language model that produces contextual text representations using information from both directions.
Related concepts
- BERTScoremeasured by · BERTScore is an automatic text-generation evaluation metric that compares candidate and reference texts using contextual representations of their tokens.
- word-matching metriccontrasts with · A word-matching metric evaluates generated text by measuring lexical overlap or other relationships among surface-level tokens, words, or n-grams, often against reference text.
- lexical similaritycontrasts with · Lexical similarity is the degree to which texts resemble one another in their surface-level words, tokens, or other forms.
- Retrieverused by · A retriever is a software component that searches or queries an information source and returns items relevant to a query for use by downstream processes.
Questions about semantic similarity
What is semantic similarity?
Semantic similarity is the degree to which two pieces of content or representations correspond in meaning, regardless of differences in wording or form.
Which Manning books explain semantic similarity?
LLM Evaluation and Alignment, The Foundational Ideas, Build an Advanced RAG Application (From Scratch), AI-Powered Search, Natural Language Processing in Action, Second Edition, Effective Conversational AI, Math and Architectures of Deep Learning, Building Reliable AI Systems and Retrieval Augmented Generation, The Foundational Ideas.
What should I know before learning semantic similarity?
Manning books that teach semantic similarity usually introduce language model, embedding, knowledge base and BERT first.
What is the difference between semantic similarity and word-matching metric?
Semantic similarity is the degree to which two pieces of content or representations correspond in meaning, regardless of differences in wording or form. A word-matching metric evaluates generated text by measuring lexical overlap or other relationships among surface-level tokens, words, or n-grams, often against reference text.
What is semantic similarity also called?
Semantic similarity is also called semantic textual similarity.
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.







