You’ve probably used a Question Answering (QA) system. Most of them are just a FAQ turned into a horrible search interface. If you don’t answer the exact question they answered, don’t bother. Other QA systems are basically just keyword search that let you put in questions.
So what is a proper question answering system? The answer seems obvious, “it is a system that answers your questions.” But to do it properly it needs to recognize synonyms, close enough answers and other aspects of the meanings of questions specifically and language generally.
In their talk, “Enriching Solr With Deep Learning for a Question Answering System” at this year’s Activate conference, Lucidworks data scientists Savva Kolbachev and Sanket Shahane showed a powerful question answering system that they constructed by adding deep learning using Fusion. They both showed how to produce more accurate answers as well as how to scale the approach given the weights of deep learning models.
Their talk covers techniques as well as the more technical, mathematical and statistical details and include a demo of how Fusion enriches Solr’s functionality. Additionally they detail highlighting using sentiment analysis.
If you’re trying to create an Information Retrieval system such as a QA system, or even if you’re just really interested in deep learning, you’re definitely going to watch this talk.