Information Retrieval · Cluster
Search UX
Result presentation and interaction.
- 01 Generating Search Result Snippets with Query-Biased Summaries How a result list turns a matched document into a two-line fragment that shows the query terms in context, and what the summarizer has to give up to do it in milliseconds.
- 02 Relevance Feedback Through Query Expansion and Term Reweighting How a search system uses a user marking results as relevant to rewrite the next query, and why the classic Rocchio and BM25 expansion methods diverge in what they trust.
- 03 Faceted Navigation and the Counts on Each Filter How a search interface narrows a result set by attribute facets, and where the per-facet document counts come from without a full re-query.
- 04 Spelling Correction and the Did-You-Mean Suggestion How a search engine decides a query is misspelled, generates candidate corrections, and chooses whether to show them or to run the query as typed.
- 05 Query Autocomplete and the Prefix Trie Behind It How the dropdown under a search box is generated from a prefix match, how the candidates are ranked, and why the index must answer partial-query lookups in single-digit milliseconds.
- 06 Pagination, Ranking Stability, and the Shuffling Result Set How a search result list is split into pages without re-ranking the whole set, why stable pagination and live index updates conflict, and what the page offset actually costs.
- 07 Result Grouping and Near-Duplicate Collapsing How a result list collapses many near-identical documents into a single entry with a cluster, and how the engine detects the duplicates fast enough to do it on every query.
- 08 Personalization Signals and Click-Based Reranking How a search engine folds a user history and click behaviour into a ranking that was previously query-only, and what the feature vectors and the training targets actually look like.
- 09 Zero-Result Queries and Automatic Query Relaxation How a search interface responds when no document matches, the sequence of constraint relaxations it can try in order, and how it decides which relaxation to show.
- 10 Structured Result Cards and the Schema That Feeds Them How a search interface renders a recipe, event, or product as a structured card rather than a text snippet, and how the structured data gets from a page into the index.