Trying to learn more about Answer / Inference engines

Question

I'm trying to learn more about how answer / inference engines work, the code behind it.

Are there any famous or well done algorithms, good books, or papers on this topic?

How do systems like Google Now ( The answer not predictive part ), Siri, and Wolfram | Alpha work?

I know they use Natural Language Processing and Machine Leaning, but how do they answer questions based from a collection of knowledge / facts?

You could start from [here](http://en.wikipedia.org/wiki/Inference_engine). — luiso1979, Sep 27 '13 at 07:02
The book "Taming Text" covers this section and even builds a simplistic QA engine using open source frameworks. — Thomas Jungblut, Sep 27 '13 at 09:19
@luiso1979 Thank you, that helped me learn more about them, but not learn how to make one. — Django Johnson, Sep 30 '13 at 01:04
@ThomasJungblut I think I own that book! Who is the book you are talking about written by? What section in the book is this at? — Django Johnson, Sep 30 '13 at 01:05

score 1 · Answer 1 · answered Sep 27 '13 at 14:39

You ask a very broad question. There are many implementations of inference engines, but they would all rely on natural language processing and searching algorithms at their core so I would focus on that.

Try the book Artifical Intelligence : A Modern Approach. It has sections on both NLP and Search and is very good.

score 1 · Accepted Answer · answered Oct 01 '13 at 11:32

1

Question/answering is not done with one algorithm. It can be a combination of NLP algorithms such as part of speech tagging, semantic analysis, semantic and/or lexical parsing etc. Then many approaches can be used like supervised learning, clustering or just storing the information and indexing them.

Can you maybe explain what you are trying to do?

answered Oct 01 '13 at 11:32

Aspasia

1,521
1
13
23

Thank you for breaking it up into parts. That is exactly what I needed. – Django Johnson Oct 05 '13 at 01:08
no problem. I think it really depends a lot on the domain, the nature and the number of data you are dealing with. Good luck! – Aspasia Oct 11 '13 at 12:52

score 0 · Answer 3 · answered Sep 27 '13 at 14:57

0

This task is called Question Answering. A few years ago there was an annual competition for it: the data are still available, and widely used in research papers.

answered Sep 27 '13 at 14:57

Ben Allison

7,244
1
15
24

Trying to learn more about Answer / Inference engines

3 Answers3