Search Evaluation and Explorer

• Paul Libbrecht

We need a tool that supports site administrators to raise the quality of the (Solr-based) search results by performing experiments.

A facet is for users which are invited to use the search tool and provide feedback on the relevance of individual results (relevant/not-relevant/should-be-elsewhere). The tool needs to collect feedback in an effective so it can be used by site administrators.

The other facet is for administrators to explore:

• what is stored in the index for each document (index explorer)
• why a given query yielded a given document and with what score (based on the Solr explain)
• adjustments to the weighting of individual fields (e.g. using the eDisMax query parser)
• how the indexing process (e.g. the language-specific features) transforms and stores pages and documents (combining what Solr and XWiki does)

Search quality evaluation is a traditional field of information retrieval. More can be read about it in the IR Book or in the summary of K Manesh.

• experience with HTML, JavaScript and CSS
• experience setting up a search tool
• knowledge of Java
• desire to learn velocity, Groovy, XWiki objects, and Solr functions

If you are curious about applying CS theories to real world problems, about supporting administrators to use data-scientific approaches to enhance their installations, or about understanding how to make it possible for others to make a great website search. You are at the right place!

2019

Proposed