nlp - Identifying and comparing syntactic structure of questio-sentence -

I am getting a user query and trying to understand the syntax.

My goal is to identify the exact quote sentence from the question entered by the user. Like

  Obama is the President of USA, who is his wife?  

So I am able to implement the Anaphora resolution and point to its to Obama I can convince and punish the above.

  Obama is the President of America, who is Obama's wife?  

But how do I get the exact question from the sentence i.e. Obama's wife

I try with the code I can identify> Pylinkgrammar which gives 54 links to the above sentence, like

  linkparser> Linking 54, Cost Vector = (Absent = 0 DIS = 8.05 LAN = 24) + -------------------------- XP- ---- ------------------------- + ------------------- --- & gt; ; WV ---------------------- & gt; + | + ------------------- Xx ------------------- + - - & gt; WV- & gt; + --- SIS --- + | + ---- wd --- + - ss - + - om - + --- mp - + - js + + + + - q- + ds ** c + + - | | | | | | | | | | | | Left Wail Obama [!] Is.v president.tv United States, who is his wife.  

I want to define patterns for various question types such as W5H1, on the basis of combination etc.

But how do I write rules for these systems, any suggestions and references will be very commendable

You can try to remove various possible sub-questions (hypotheses) to test the text to be written between your original text and the text and the references