A shortcut from
words to pages.
A search can begin with the word you remember, rather than every page you have forgotten.
An educational explanation of a general technology concept.
Reverse the lookup
A collection stores documents and their contents. An inverted index adds another route: start with a term and find the documents containing it. Preparing that route in advance can avoid reading the whole collection again for every search.
In a hypothetical set of field notes, note A says “river stones” and note B says “river plants.” An index can associate “river” with both notes, while “stones” points only to A. Looking up the term produces candidate documents.
| Indexed word | Notes containing it |
|---|---|
| river | A, B |
| stones | A |
| plants | B |
Prepare the words
Text analysis turns a passage into searchable units called tokens. Rules may also normalize them, for example by treating uppercase and lowercase letters alike. Other rules can connect word forms or selected synonyms.
These choices affect what counts as a match. In our hypothetical notes, a search for “River” can find “river” if case is normalized. A search for “rocks” will not automatically find “stones” unless the system has a rule or another method connecting them.
Compatible processing of stored text and incoming queries helps keep the comparison meaningful. There is no single analysis rule that suits every language, identifier or collection.
Finding is only the beginning
Once candidates are found, ranking decides which appear first. A system might consider how often a term appears, how unusual it is across the collection and where it occurs. Finding a word and judging a document's usefulness are different tasks.
An index also needs maintenance when documents change. Search results may reflect the indexed version rather than the latest underlying text. A result is therefore an invitation to inspect the source, its context and its date.