Chapter 8. Building an example question answering system

 

In this chapter

  • Applying techniques for automatically tagging documents
  • Enabling document and subdocument tags to be leveraged in search
  • Reranking documents returned by Solr based on additional criteria
  • Generating possible answers to users’ questions

In the previous chapters, we’ve looked at different technologies and approaches independently. Though we’ve still managed to build useful applications focusing on one or two technologies, often you need to combine several of the tools we’ve described so far to get the job done. For instance, search and tagging (classification) with faceting are a natural fit, as are clustering and search, when it comes to helping users find and discover new and relevant content for their information needs. For the purposes of this chapter, you’ll build a question answering (QA) system capable of answering fact-based questions from users (written in English) using search, named-entity recognition, and string matching, among other techniques. Though most of the other chapters stand on their own, in this chapter, we assume you’ve read the previous chapters and so we don’t explain the basics of Solr and other systems again here.

8.1. Basics of a question answering system

8.2. Installing and running the QA code

8.3. A sample question answering architecture

8.4. Understanding questions and producing answers

8.5. Steps to improve the system

8.6. Summary

8.7. Resources