Create a Suggest Dictionary

This section gathers all the procedures you need to add and configure a Search Suggest dictionary.

This task shows you how to:

See Also
Enable the Suggest in the Mashup UI
Use the Suggest Via the Search API
Export Suggest Dictionary Content to an XML File
Dispatch a Query to Several Suggest Dictionaries

Add a New Suggest Dictionary

  1. In the Administration Console, go to Search > Suggest.
  2. Click Add suggest and select one of the suggest types. For more information, see Available Suggest Types.

Add Allow Lists or Block Lists to a Suggest Dictionary

  1. Expand Block and allow list.
  2. Next to Allow list or Block list, specify your resource file.
    • If you have already created a resource file, click Browse. Select the resource file, which contains all allow list and block list resources created in the Suggest group of the Resource Manager. Then click Accept. If you have created a resource file using cvadmin, type the path to the resource file using the format resourcemanager://group_name/resource_name.
    • OR, create a new resource: click Create new, specify a name for the allow list or block list, and click Accept. This adds the resource to the Suggest group in the Resource Manager, which ensures correct deployment of interdependent resource files in multihost environments.
  3. Click Apply.
  4. (Optional) To define the contents of the resource file, click Edit. This takes you to the Business Console. For more information, see Add a Suggest Block List and Add a Suggest Allow List in the Business Console.

Configure Query-Time Options

Expand Query-time options to specify how the suggest handles queries.

Option

Description

Distance

Allows approximate matching. The higher the distance the more approximate the match.

  • 0: exact match.
  • 1: distance tolerance of 1 between the result and the query
  • 2: distance tolerance of 2 between the result and the query

For more information about approximate matching, see Approximation.

Autocomplete

Appends suggest results to the last query word entered in the search field to autocomplete it.

It only applies to suggests built with the Subexpr matching or Substring matching build options.

Recursive

Discards the leftmost word of the query progressively.

It sends each new subquery to the suggests until you reach the max number of suggestions, or until there is no more word to use.

For example, for a query "A B C", the suggest is called 3 times, with "A B C", "B C", and "C".

Configure Build Options

Important: These options can have a tremendous performance impact, read carefully Performance Considerations and Options for Search Suggest.

For all suggests (except those based on custom dictionaries), you can configure build options.

Build option

Description

Subexpr and Substring matching

Normally, suggest matching is prefix-based: "first" returns entries "first test" and "first image".

Sometimes, you want to be able to do a wider matching, not always prefix-based.

  • Subexpr matching allows you to find matches on every start of word. For example, "first test" returns both for "fir" and for "tes".
  • Substring matching allows you to find matches on every letter. For example, "first test" returns for "fir", for "rs", for "es", ...

Sentence split and Ngram split

For performance reasons, use these options to avoid long entries. By “long”, we mean entries longer than 100 characters (100 bytes).

Sentence and ngram split options allow you to break up a suggest entry into several entries, and to perform matches independently on the chunks.

  • For sentence split, if the entry is multisentence, an entry is created for each sentence.
  • For ngram split, a sliding window of ngrams of a given size is created and an entry created for each step of the window. For example, "a b c d e f" with a split on 4-grams gives entries "a b c d", "b c d e", "c d e f".

Note: 0 means no splitting.

Compute permutations

Computes all permutations for an entry and adds them as separate entries. For example, if you start entering "Angeles", Exalead CloudView automatically suggests "Los Angeles".

Note: Entries longer than 8 words are not permuted for performance reasons.

This action is performed after the sentence split if the Sentence split option is selected.

To apply permutation to Static XML suggest and Static resource suggest types, you need to add permutation=”true” to the SuggestDictionary tag in your XML file or suggest resource in the Business Console.

Max. entry length

The maximum number of characters in a suggest entry.

This is a security measure to prevent overly long entries. They are automatically truncated after the specified length.

0 means no limit.

Max. suggestions

The maximum number of suggestions that can be shown to the user for a given input string. You cannot change this dynamically.

Tokenization config

Specifies the Tokenization configuration to use.

Sanitize entry

This option strips the entry of punctuation, and encloses any UQL operators in quotes.

It is useful when you want to suggest among a list of product references containing "-" (hyphens) or other delimiters, and you do not want any tokenization on these characters.

Build after import

Triggers a build automatically after the index refreshes.

Enable security

Makes use of documents and users’ security tokens to restrict suggestions.

Compile the Suggest Dictionary

Once created, suggests must be compiled in the Administration Console.

Important: Building suggest fails if there is not enough disk space to calculate it. It is best to allocate substantial disk space for the suggest build to copy/compute raw files from temporary files (in build/resources/tmp). If Build options are enabled, for example subexpr matching and substring matching, the required disk space is even bigger. Read carefully Performance Considerations and Options for Search Suggest.

Go to Search > Suggest and click Build now.

For each suggest, you can also schedule suggest builds using the Build scheduling options.