1 reply to this topic

[vbehzadan]

Hi,

I'm currently working on a "smart" question answering system,which is supposed to get the question from user in plain english,break the question string in groups of words,then rate each Q/A pair in the knowledge base to find the most related answer.

First of all, I'm planning to make this app web based, so i've chosen Python CGI as the primary language and php/mysql/ajax for the user interface and the knowledge base DB. Is there any other suggestions anyone would like to add?

And the main question: I'm not sure about the algorithm I have to use to rate the Q/A pairs. one is to compare frequency of each group word in each pair and rate accordingly, but I dont find this solution accurate enough. Please share any ideas you've got for this issue.

Cheers,
VB

[marwex89]

Something keyword-based maybe? That's probably what I would try for the main search. What you suggested could be used as a secondary rating system (after you have found the entries that matched any of the selected keywords).