Back to top

Methods for Handling Search Engine Exceptions

  • Thesaurus entries – Each thesaurus entry is, in fact, an exception.

- For example, “car” as a synonym for “automobile”, “iWatch” as a synonym for “Apple Watch,” “IBM” as a synonym for “International Business Machines, Inc.”

- Search engines need these to bridge the gap between the searcher’s language and the writer’s language.

  • Irregular variations – Such as irregular nouns ("mouse" to "mice") and verbs ("catch" to "caught") are typically handled with thesaurus entries or language dictionaries.
  • Boosts and blocks – Search results that are manually boosted or removed (blocked) for a query. These are manually specified.
  • Best bets – These are recommended destinations for a query.

- Different than “boosts and blocks” because the recommendations do not necessarily need to have been indexed into the search engine

  • Field snapping – Identifying certain words as keywords for filters (e.g. “snap the word” to the field).

- The typical example is for colors, where a query such as |red shoes| will automatically “snap” the word ‘red’ to the ‘color’ field.

  • Taxonomies and ontologies – Are methods for classifying content, business objects, and key concepts so that the search engine can perform more intelligent searches.

- In this new view, each taxonomy entry or ontology entry is a codified exception case.

  • One box entries – These are query patterns which cause special features to show up in the search engine results.

- Usually these are regular expressions which, when matched, cause special functions to show up in the search results.

- For example, a product code does a lookup into a product database and shows metadata information about the product in a special box at the top of the search results.

  • Question answering (QA) systems – These are systems which provide the answer directly rather than search results which may (or may not) contain the answer.

- Previously we had assumed that question/answer should be handled by an entirely separate system.

  • Flagging sensitive or non-compliant searches – Many organizations need to handle searches on sensitive topics (e.g. searches top-secret information) or suspicious topics (e.g. searches which violate compliance rules such as searches for insider information or in violation of HR policies) using special methods.

- For example, such searches may be subject to special security checking or auditing.

  • Other techniques – Such as personalization based on user path, recommendations engines, query chaining, foul language handling, and semantic searching also all contain elements of exception handling.