Why do results include answers that do not include the word or phrase I searched for?

Environment:

Phrase Searching

Issue:

When I search our end-user pages, I frequently get results that include answers that do not include the word or phrase I searched for. For example, when I search our site using the word "indices" in the search box, I get several answers listed, even though none of them include that word.

Also, when I search for a phrase, I get answers that do not include the whole phrase. For example, when I search on "market share", I get more answers than if I just search on "market".

Resolution:

When a search is performed in Oracle B2C Service, the application first excludes words listed in the exclude text files in the File Manager. Typically, the exclude files includes common and frequently-used words such as pronouns and prepositions, including "your", "at", and "the". For more information on excluding words and phrases from searches, refer to Answer ID 1309: Excluding Words from Searches as well as Answer ID 1985: Stopindex and Startindex Tags.

The exclude_answers.txt file specifies the words to exclude when searching answers on the end-user pages or the administration pages. The exclude_incidents.txt file specifies words to exclude when searching for incidents at the Account - Questions pages or the administration pages.

Common word stems: After excluding certain words from the search, the searching algorithm used by Oracle B2C Service next stems each word that remains in the search. This word-stemming method takes each remaining word of your query and stems the word back to its root component. The search is performed on the root words, which means the search results include words that have the same root word. Typically, word-stemming allows matches for words with different suffixes, such as -s, -ed, and -ing. As a result, your search results include variant forms of the words included in the search.

For example, if you search on the word "indices", the Oracle B2C Service search feature returns "index" and "indexes" as a match for "indices", since the application searches on the stemmed word.

Notes: The words listed in the Product and Category menus are included when searching on a word or phrase. That is, if you search on a word that is in a specific product, the answers associated to that product are returned with the search results, even if the answer does not include the word specifically.  You may see answers returned that do not contain your search item if one or more of the following is true:

1. The answers contain words that are aliases to query words.
2. The answers contain synonyms of query words (in the case of Similar Phrases search).
3. The answers contain keywords matching the query words.

Multiple word phrases: When multiple words are included in the Phrases search method, by default, words are joined as an OR statement, not an AND statement. This means that when you search on a phrase such as "market share", the search results include answers that include both "market" and "share" as well as answers that only include "market" and answers that include only "share". Typically, answers that include both words are listed towards the top of the list. Answers that include only one of the words generally appear lower in the list since they are not usually as strong a match.

If you want to search for answers that include both "market" and "share", use a + before each word, namely "+market +share". The plus sign indicates that the word must be in the answer in order for it to be returned in the search results.

You can specify whether multiple words should be joined with an OR connector or an AND connector. For more information on this, refer to Answer ID 2049: Searching with an AND Join Between Multiple Word Phrases.

For more information on the different search types used in Oracle B2C Service, refer to Answer ID 566: Search Options on the End-user Pages.