Path: csiph.com!usenet.pasdenom.info!goblin2!goblin.stu.neva.ru!newsfeed.xs4all.nl!newsfeed1.news.xs4all.nl!xs4all!post.news.xs4all.nl!not-for-mail
MIME-Version: 1.0
In-Reply-To: <CAHzaPEMnb0iAbp_V9EK7BkR_mk1Kj1RCnYrj0ztZ-uR5XA-bsQ@mail.gmail.com>
References: <90b8ca83-fb81-40d6-a864-f1c0e07bca76@googlegroups.com> <CAHzaPEMnb0iAbp_V9EK7BkR_mk1Kj1RCnYrj0ztZ-uR5XA-bsQ@mail.gmail.com>
Date: Tue, 1 Oct 2013 17:04:00 +0200
Subject: Re: extraction tool using CRF++
From: Joost Molenaar <j.j.molenaar@gmail.com>
To: python-list@python.org
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: quoted-printable
Precedence: list
Newsgroups: comp.lang.python
Message-ID: <mailman.562.1380639848.18130.python-list@python.org>
Lines: 79
NNTP-Posting-Host: 2001:888:2000:d::a6
Xref: csiph.com comp.lang.python:55219

Hi Ron,
In the python/ subdirectory of the CRF++ source package there's a
README with instructions on how to use the CRFPP python module.

HTH,

Joost

On Tue, Oct 1, 2013 at 4:24 PM, Vlastimil Brom <vlastimil.brom@gmail.com> w=
rote:
> 2013/10/1 cerr <ron.eggler@gmail.com>:
>> Hi,
>>
>> I want to write an extraction tool using CRF++ (http://crfpp.googlecode.=
com/svn/trunk/doc/index.html).
>> I have written a trainings file and a template:
>> training:
>> banana  FOOD    B-NP
>> bread   FOOD    I-NP
>> template:
>> U01:%x[0,1]
>> U02:%x[1,1]
>>
>> and now I want to go ahead and extract the foods from a sentence like "h=
ow do I make a banana bread". Also, I'm unsure how I interface to crf++ wit=
h python, I compiled and installed it from source as described on the above=
 website but I don't have a crf module available in python...
>> --
>> https://mail.python.org/mailman/listinfo/python-list
>
>
> Hi,
> I have unfortunately no experience with CRF++; if there is no python
> wrapper for it available, the usage might not be (easily) possible -
> depending on the character of this library, you may try accessing it
> e.g. via ctypes.
>
> Alternatively, you may try another packages already available, e.g.
> NLTK:  http://nltk.org/
>
>>>> import nltk
>>>> any(synset.lexname =3D=3D "noun.food" for synset in nltk.corpus.wordne=
t.synsets("apple"))
> True
>>>> any(synset.lexname =3D=3D "noun.food" for synset in nltk.corpus.wordne=
t.synsets("bread"))
> True
>>>> any(synset.lexname =3D=3D "noun.food" for synset in nltk.corpus.wordne=
t.synsets("wine"))
> True
>>>> any(synset.lexname =3D=3D "noun.food" for synset in nltk.corpus.wordne=
t.synsets("book"))
> False
>>>> any(synset.lexname =3D=3D "noun.food" for synset in nltk.corpus.wordne=
t.synsets("pencil"))
> False
>
> # of course there might be some surprise, probably due to polysemy ore
> some specifics of the semantic description...
>
>>>> any(synset.lexname =3D=3D "noun.food" for synset in nltk.corpus.wordne=
t.synsets("dog"))
> True
>>>> any(synset.lexname =3D=3D "noun.food" for synset in nltk.corpus.wordne=
t.synsets("white"))
> True
>>>>
>
> cf.
> http://nltk.org/
> http://nltk.googlecode.com/svn/trunk/doc/howto/wordnet.html
> http://www.velvetcache.org/2010/03/01/looking-up-words-in-a-dictionary-us=
ing-python
> http://wordnet.princeton.edu/man/lexnames.5WN.html
>
> hth,
>    vbr
> --
> https://mail.python.org/mailman/listinfo/python-list