Topic
  • 3 replies
  • Latest Post - ‏2012-10-01T10:19:06Z by SystemAdmin
OlgaLind
OlgaLind
34 Posts

Pinned topic primary morphological analysis?

‏2012-09-28T08:23:01Z |
Hi,

I'm searching for the most flexible ways of distiguishing semantic Subject and Object in some sentences, for example:

A cat is bitten by a dog.
A cat is biting a dog.

Currently to create a "Victim" annotation I have to "hardcode" active and passive verb forms in the Dictionary by creating separate entries for "is biting" and "is bitten" forms with Active and Passive categories. But it would be more convinient if I had a separate morfological rule annotation that would be able to extract from text all passive forms - and the similar rule for active forms. What could be done here?

Thanks.
Updated on 2012-10-01T10:19:06Z at 2012-10-01T10:19:06Z by SystemAdmin
  • SystemAdmin
    SystemAdmin
    197 Posts

    Re: primary morphological analysis?

    ‏2012-09-28T14:18:00Z  
    Unfortunately this is not possible. Advanced users can do so by modifying the JFST parsing rules grammar, however doing it is quite "dangerous" and is not encouraged because it may affect rules, and it is hard to debug if something goes wrong.

    On the other hand, in your model, you may need to do a classification as you did, because in reality you will need a specific set of verbs that you will need for your model in terms of victim and perpetrator. For example, if you are creating a model for police reports, you will need a set of verbs specific to your model. So you will still need to create this dictionary with verbs.

    Hope this makes sense
  • OlgaLind
    OlgaLind
    34 Posts

    Re: primary morphological analysis?

    ‏2012-10-01T07:57:25Z  
    Unfortunately this is not possible. Advanced users can do so by modifying the JFST parsing rules grammar, however doing it is quite "dangerous" and is not encouraged because it may affect rules, and it is hard to debug if something goes wrong.

    On the other hand, in your model, you may need to do a classification as you did, because in reality you will need a specific set of verbs that you will need for your model in terms of victim and perpetrator. For example, if you are creating a model for police reports, you will need a set of verbs specific to your model. So you will still need to create this dictionary with verbs.

    Hope this makes sense
    Hi,

    Thank you for the response!

    I think, probably, there are some ready custom annotators (such as http://uima.apache.org/sandbox.html#whitespace.tokenizer)that I could use? I mean the annotators that are able to extract passive/active voice verb forms from natural text? Please tell me, if you have any clue.
  • SystemAdmin
    SystemAdmin
    197 Posts

    Re: primary morphological analysis?

    ‏2012-10-01T10:19:06Z  
    • OlgaLind
    • ‏2012-10-01T07:57:25Z
    Hi,

    Thank you for the response!

    I think, probably, there are some ready custom annotators (such as http://uima.apache.org/sandbox.html#whitespace.tokenizer)that I could use? I mean the annotators that are able to extract passive/active voice verb forms from natural text? Please tell me, if you have any clue.
    I am afraid I am not aware of any ready custom annotators like you describe them.
    Hopefully other users may have some ideas and can share them.