Configuring search results

Before you begin

Learn the basic logic of search in the application and explore which items can be searched; see Search-in-BMC-Helix-Digital-Workplace. 

Enhance search results 

On the Search page of the Configuration tab, you can amplify the accuracy of search results with the following search options:

• Configure the maximum number of results per source
• Select whether requests and approvals appear in the Top Hits tab

The following image shows the Search configuration page:

To configure the maximum number of results per source

By configuring Maximum results per source, you limit the number of search results per source in a global search. Source is an integrated system from which items (services or knowledge articles) are retrieved. You can select any number from 1 to 100. The default value is 25. 

To select whether requests and approvals appear in the Top Hits tab

By using Include in Top Hits results, you include or exclude requests and approvals from search results. If you disable this feature, the end user must click the Requests and Approvals tabs to search these categories. This feature is enabled by default.

To enable dynamical re-ranking of search results

(Available in version 22.1.05 and later) Search results for different sources can be returned by using an improved ranking technology. When the feature is enabled, the content of the primary search results from the different sources is re-ranked by using the Okapi BM25 ranking function. During this second round of ranking, the search result data is parsed into a single document corpus. The parsing process uses a text processing pipeline that includes:

• Upper- to lowercase transformation.
• Punctuation removal.
• Tokenization (breaking sentences into words).
• (English only) Lemmatization (determining basic word forms; for example, the basic form of "better" is "good").
• Stop word filtering (overly common words of the language are removed). 

When all of the search result documents are parsed, the corpus represents a record of the total number of documents, the average size of each document, and the number of documents that would match a particular search term. Each document within the corpus has a record of the number of unique terms within that document, and the number of occurrences of each unique term. These values are then used to determine the relevance score of each term within the user’s search query. The user’s query is also parsed by the same pipeline to facilitate matching.

This feature is disabled by default.

To enable natural language searching

By enabling Natural language searching, you can leverage conversational language (such as "I need to order a new laptop") instead of a keyword-based search. The main purpose of this feature is to determine whether a user is looking for a Catalog item or a knowledge article. The search engine picks up the meaning of a sentence or word group, recognizes the end user's intent, and defines the appropriate category, thus reducing the time spent searching. This feature is enabled by default.

Natural language searching also enables end users to experience the following additional search capabilities:

• Additional resources in the Related items section. 
  If there are no additional results, the Related items section is not displayed. The following image shows the Related items section:

• Top three search results for each category.  
  On the Top Hits page, you can see summaries with top three search results within individual categories. Click See more to expand all the results for specific categories. The following image shows the top three results for the catalog items: