Has anyone used YodaQA for natural language processing? How easy is it to link to a document database other than Wikipedia?

We're thinking we can create a bot to use AI to analyze our developer and user documentation and provide a written or spoken answer in reply. YodaQA comes linked to Wikipedia for starters, but we'd need to link to our own source info. I'm trying to get an idea of the development time required to set up the AI and then to link to the database.

  • $\begingroup$ i am interested in the exact same topic: creating a question answering system for a programming API. in my case, the API is the Revit API. i have heaps of existing documentation and would like to implement a question answering system based on that. i would be very interested in hearing what you find out your end, and also in any answers to this question anyone can provide. thank you! ontology sources for Revit API: thebuildingcoder.typepad.com/blog/2017/01/… $\endgroup$ – Jeremy Tammik Jan 15 '17 at 11:46
  • $\begingroup$ Look at yodaqa issue #17 Domain Adaptations for a pretty extensive discussion of this and closely related topics. $\endgroup$ – Jeremy Tammik Jan 15 '17 at 18:35

Question Answering systems can be grouped into two categories: Simple, textual ones and hybrid QA systems. A textual QA system parses the input text with a POS tagger and tries to understand the grammar. It is influenced by Chomsky grammars and understands natural language like computer source-code. The assumption is, that all information are given by the text itself.

In contrast, hybrid QA systems combining textual data with linked data. Linked data is equal to a RDF database in the tripe storage format. The ontology in the background is used for enhancing textual information. Sometimes this concept is called “grounded Question answering system”. In the “QALD Challenge” many QA-systems were benchmarked against each other, and at least one paper describes the usage of YodaQA in the QALD challenge.

The QALD challenge is an ongoing benchmark contest founded 6 years ago and provides different tasks. Apart from the DBpedia and the biomedical dataset track, there is also a task available called “HOBBIT project” (Holistic Benchmarking of Big Linked Data) which is a question answering task for large scale applications.

  • $\begingroup$ You didn't answer the question. $\endgroup$ – Brian O'Donnell Mar 10 '19 at 4:34

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.