Improving Search Accuracy and Performance for a Major Book Publisher

Adam Forsythe-Cheasley

Nesight greatly improved search functionality on a customer’s business portal by integrating Solr search into the Plone CMS. Weighting of metadata and spell-checking suggestions were key to better search results.

Netsight recently completed a project to improve the search on a global publisher’s large, document-rich website. Previously, the site had used the standard Plone search. Although this standard search functionality is adequate for smaller sites, when the amount of content in a site starts to grow the search can become difficult to customise and start to return unexpected results.

The main problem usually stems from a user’s expectation of how search should function, with regards to the title of an item of content vs the body content. If users enter the search string ‘hello world’, for example, they might expect a page with the title ‘Hello World’ to appear higher in the search results than a page with the title ‘Another Page’ and containing the phrase ‘hello world’ many times in the body.

To solve this problem, we need to weight certain attributes of a page higher than others. In this way, a page’s title could carry more weight than its body. In this case, Netsight used the Solr search platform. Solr offers many advantages over the standard Plone search: better speed, scalability and many advanced monitoring and configuration tools. Solr also integrates well with Plone, acting as a drop-in replacement for the standard search functionality. In this case, it allowed us to add weighting to the various attributes of their content, which greatly improved the quality of their search results.

Netsight was also able to improve the UI for the search results, adding a simple content type filter if users only wanted to search for a specific type of content. We also added to the results spell-checking suggestions, which help users quickly perform searches again if there is a typo in a search string.

After working closely with the client on fine-tuning of attribute weighting and other configuration of Solr search, overall the feedback has been very positive.  We will look to implement this search technology again for future projects.

Tweet about this on TwitterShare on FacebookShare on Google+