Question answering
Question answering is the computational task of producing, selecting, or extracting an answer to a natural-language question using available information or knowledge.
AI-Powered Search
Question answering is the task of producing an answer to a natural-language question from available information.
- How it is used
- Returns a short answer to a searcher’s question instead of requiring inspection of a results list.
- How deep it goes
- Works through dataset curation, reader fine-tuning, holdout checking, and retrieval-time reranking, concentrating on extractive rather than abstractive answers.
- Examples
- Uses outdoors question-and-answer posts, a minimalist-shoes answer span, and a ten-example holdout check.
- What you can do after
- Build it from scratch using a search retriever and a fine-tuned Transformers reader to extract and rank candidate answers.
Effective Conversational AI
Question-answering is a conversational AI capability that responds directly to a user's question, usually without follow-up interaction.
- How it is used
- Delivers direct answers to frequently asked questions, usually without follow-up interaction.
- How deep it goes
- Explains utterance-to-intent-to-response mapping, example-based training, and response trade-offs without descending into implementation code.
- Examples
- Uses store-hours and cake-price questions, cake ordering, and account-balance requests to distinguish answers from completed tasks.
- What you can do after
- Build it from scratch with trained question examples and controlled responses, then evaluate answer accuracy and unnecessary follow-ups.
Read how Effective Conversational AI explains question answering →
Natural Language Processing in Action, Second Edition
A question-answering system produces answers to questions expressed in natural language.
- How it is used
- Builds FAQ bots and document-grounded applications that answer natural-language questions.
- How deep it goes
- Explains matching, retrieval, extraction, and generation with runnable pipelines, but defers deeper model internals.
- Examples
- Uses data-science FAQs, book sentences, Haystack, TF–IDF, FAISS, BERT-based retrievers, and transformer readers.
- What you can do after
- Builds a basic FAQ matcher or configures a Haystack retrieval-and-reading pipeline.
Read how Natural Language Processing in Action, Second Edition explains question answering →
Introduction to Generative AI, Second Edition
Question answering is the task of producing answers to questions posed in natural language.
- How it is used
- Frames question answering as a core large-language-model application for responding to natural-language questions.
- How deep it goes
- Distinguishes task formats and answer-generation modes, then explains representation, parsing, pretraining, and uncertainty without implementation details.
- Examples
- Uses Nobel Prize, Beethoven, and first-president questions alongside text passages, tables, multiple-choice items, and conversational datasets.
- What you can do after
- Evaluate question-answering systems by distinguishing answer formats, context use, reading-comprehension settings, and risks of unsupported responses.
Read how Introduction to Generative AI, Second Edition explains question answering →
Related concepts
- conversational aikind of · Conversational AI is artificial intelligence that enables software systems to interact with users through natural-language dialogue.
- extractive question answeringincludes · Extractive question answering is a technique that answers a question by selecting relevant text directly from supplied source material.
- retrieve-and-readimplemented by · Retrieve-and-read is a question-answering architecture in which a retriever selects relevant information and a reader or language model uses it to produce an answer.
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.