ATG Search results contain a ranked list of matching statements, sorted by relevancy. These raw results could contain duplicate or similar statements or come from a limited set of documents, offering little variety of information for the end-user. To prevent these problems, ATG Search offers a variety of grouping levels, described in the sections that follow.
For information on the parameters that control these modes, see Using Grouping Modes in the Query Component Reference chapter.
Grouping by Statement
In grouping by statement, ATG Search reviews the list of matching statements and collapses those that are similar. Parameters control the similarity metric, the maximum size of result groups, and the number of result groups to return. Individual results that belong to a group that exceeds the size limit are eliminated from the response.
For example, if the query requests the first 20 result groups by statement, but restricts the number of groups to one, then ATG Search returns the 20 most relevant distinct statements, eliminating less relevant similar statements. If the query restricts the number of groups to two, then ATG Search returns the 40 most relevant results in groups of two, one group for each distinct matching statement (according to the similarity metric). The result objects of the response denote which group they belong to.
Because ATG Search operates at the sentence level, grouping by statement allows the user to view the most relevant results in the order of their relevance. A drawback of this option is that it is not always clear whether a particular document or subset of documents have a heavy concentration of relevant responses.
Grouping by Document
In grouping by document, ATG Search reviews the list of matching statements and collapses those that come from the same index item, forming groups of document results. Parameters control the size of result groups and the number of the result groups to return. Individual results that belong to a group that exceeds the size limit are eliminated from the response.
For example, if the query requests the first 20 result groups by document, but restricts the number of the groups to 1, then ATG Search returns the 20 most relevant statements from 20 unique documents, eliminating less relevant statements from the same document. If the query restricts the number of groups to 3, then ATG Search returns the 60 most relevant results in groups of 3, one group for each unique document. The result objects of the response denote which group they belong to.
Grouping by document allows the user to make a quick survey of the kinds of answers available in the returned documents, but less relevant statements can end up higher on the result list than more relevant statements.
Grouping by Property
In grouping by property, ATG Search groups matching statements according to a metadata property of their documents. Thus, the grouped results may not share the same document or the same text. The typical scenario for this is in a commerce application that has indexed SKUs as searchable items, but wants to return results at the product-level. The results can be grouped by a “product id” metadata property, so that all SKUs from the same property are in the same group.
Grouping by Field
In this grouping method, the matching statements are grouped by the name of the text field that it came from. Typically, only structured content has text fields, although unstructured content has URL and title fields. This feature allows specialized displays of the results based on where it was found in the content. For example, a display might want to show results from the name field of an item separately from the results from the description field.
