Page tree
Skip to end of metadata
Go to start of metadata

The Settings screen in the Querying section of the Admin UI provides options to control how user queries are interpreted by LucidWorks Search and how results are displayed to users. Use this screen to make changes to those settings. Note that some of the settings in the Admin UI will only apply if the "lucid" query parser (see below) is used.

The following options are also available. All of these settings can also be configured with the Settings API. After editing any setting, click Update to save your changes.

 

 

 

 

 

 

 

 

Fields Must Be Configured To Support Query Settings

Icon

Many of the query settings require corresponding index configurations for each field that should be used for that functionality. For example, if spell-check is enabled for user queries, fields should be selected to populate the spell-check index. This applies to:

Default Sort

Defines how results will be sorted automatically for users. Users can change the sort option while viewing the results of their search. The default is relevance, other options are date and random.

Settings API attribute name: default_sort

Query Parser

Selects the query parser to be used for parsing queries. By default, LucidWorks Search uses the Lucid Query Parser (lucid). Other options are dismax, edismax, and lucene. For information about other available query parsers, see Query Parser Options.

Settings API attribute name: query_parser

Spell-check

Will use an index of terms to suggest possible alternate spellings to users above their search results. By default, this parameter is turned on (and the index will be created when new content is added). Fields must be selected to be used for spell checking; by default, the author, body, description, and title fields from crawled documents will be used.

Settings API attribute name: spellcheck

For each query result, displays a link to a query for similar results. By default, this parameter is turned on.

Settings API attribute name: show_similar

Display Facets

Displays facets (groupings of documents by common elements such as category or author) for results. By default, this parameter is turned on.

Settings API attribute name: display_facets

Unsupervised feedback

Enables automatic unsupervised feedback of all user queries. Unsupervised feedback is a relevancy tuning technique that analyzes documents returned by a user's query and uses the top 5-10 terms from those documents and re-executes the query with those terms added. By default, this feature is disabled. For more information, see #Unsupervised feedback emphasis.

Settings API attribute name: unsupervised_feedback

Unsupervised feedback emphasis

When Unsupervised Feedback is enabled, unsupervised feedback emphasis has two choices:

  • Emphasize Relevancy (the default)
  • Emphasize Recall

Emphasize Relevancy assures that no additional documents are included or excluded as a result of unsupervised feedback, with the only effect being to rank more relevant documents higher based on the additional query terms found in the top-ranked documents of the original results. This option uses the list of feedback terms to perform an AND operation with the original user query.

Emphasize Recall performs traditional unsupervised feedback which may return more documents than the original query. You can use this option to discover alternative keywords beyond those used in the original query which are likely to be relevant to the user. One potential disadvantage is that it may return many more documents of dubious relevance even as more documents of relevance are added to the results, but that may be advantageous for the discovery process. This option takes the list of feedback terms and performs an OR operation with the original user query.

For more information on unsupervised feedback, see Unsupervised Feedback in Understanding and Improving Relevance.

Settings API attribute name: unsupervised_feedback_emphasis

Include stop words in searches

Enables specialized stopword handling. By default, this parameter is turned on. To modify the list of stopwords, see Stop Words.

Settings API attribute name: query_time_stopwords

Auto-complete

Will suggest possible query terms to users via a drop-down menu in the query box. By default, this parameter is turned on, however the index must be created with the "Generate autocomplete index" activity scheduled via the Indexing Settings screen. See Auto-Complete of User Queries for more information about configuring Auto-Complete.

Settings API attribute name: auto_complete

Boost recent

Boosts the relevance scoring of more recent documents. By default, this parameter is turned on.

Settings API attribute name: boost_recent

Use synonyms

Allows the use of synonyms when returning query results. By default, this parameter is turned on. For more information about modifying the list of synonyms, see Synonyms. This parameter will only work if the 'lucid' query parser is selected as the default or specifically used in the search request.

Settings API attribute name: query_time_synonyms

Enable click tracking and boosting

Click Scoring allows administrators to incorporate user clicks into the LucidWorks Search relevance ranking algorithm. There is only one step to enable Click Scoring; however, a boost data file must be created by scheduling and running the Click Activity. After running that activity, if Click Scoring is enabled, boost data will be available for use in relevance calculations. Once enabled, user queries and associated document clicks will be logged to the $LWS_HOME/data/logs/click_<collection>.log file.

Icon

There is currently a known issue where Click Scoring will not properly process calculated boost information until LucidWorks Search is restarted. So, when enabling Click Scoring, please also schedule a full LucidWorks Search restart. For details on how to restart, see the section Starting and Stopping LucidWorks Search.

For more information about Click Scoring, see Click Scoring Relevance Framework.

Using Click Scoring with NearRealTime Search

Enabling Solr's NearRealTime (NRT) search by configuring the update_handler_autosoftcommit_* parameters with the Settings API or the Auto-soft-commit* settings in the Admin UI has some impacts on how user clicks are processed by LucidWorks.

In order to avoid a performance hit on NRT search when Click Scoring is enabled, documents added between the last "hard" commit and the current "soft" commit are not augmented with click-through data.

Deletions since the last hard commit are processed as usual (i.e., documents deleted are not visible), but their term statistics are still included in the global term statistics (which includes the fields added by Click). Added documents since the last hard commit will not get any click-related fields until the next hard commit, even if a document with the same unique key was deleted and replaced by a new, updated, version of the document.

Settings API attribute name: click_enabled

Related Topics

  • No labels