Mailing-List: contact dev-help@opennlp.apache.org; run by ezmlm
Precedence: bulk
Reply-To: dev@opennlp.apache.org
Received-SPF: pass (athena.apache.org: domain of nando.nlp@gmail.com
 designates 74.125.82.176 as permitted sender)
MIME-Version: 1.0
In-Reply-To: 
 <dbdc781773094e77a4e672cbdb28df2f@DB3PR05MB057.eurprd05.prod.outlook.com>
References: 
 <CACQuOSVu51-8vS=tCyZXNVcAZLiQuWhuDpksW1uZOCGQg-3v_g@mail.gmail.com>
	<dbdc781773094e77a4e672cbdb28df2f@DB3PR05MB057.eurprd05.prod.outlook.com>
Date: Thu, 26 Sep 2013 14:14:37 -0600
Message-ID: 
 <CACQuOSWcRNGuzrLnbFoFF+x9og1BeNwg-5Zms7xxaCUwjKCY+Q@mail.gmail.com>
Subject: Re: Triplet Extraction with OpenNLP
From: Carlos Scheidecker <nando.nlp@gmail.com>
To: dev@opennlp.apache.org
Content-Type: multipart/alternative; boundary=f46d04428cc2e27c4504e74f0492

--f46d04428cc2e27c4504e74f0492
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: quoted-printable

Thanks Svetoslav,

Would you have an example on that?

cheers,

Carlos.


On Thu, Sep 26, 2013 at 5:09 AM, Svetoslav Marinov <
svetoslav.marinov@findwise.com> wrote:

> Hi Carlos,
>
> This is not exactly answer to your question but I am not really convinced
> that a Phrase structure tree is the best way to extract triplets. A
> dependency graph is a much better option.
>
> There would be a number of NPs and PPs that are neither the subject nor
> the object, and not sure at all whether an adjective can be an object.
>
> However, if you want to use OpenNLP and the parse tree, maybe you can
> consider mapping the tree to FrameNet, thus you will see what kind of
> arguments a verb will have and which of these can potentially be the
> subject and the object.
>
> Best,
>
> Svetoslav
> ________________________________________
> Fr=E5n: Carlos Scheidecker <nando.nlp@gmail.com>
> Skickat: den 26 september 2013 11:37
> Till: dev@opennlp.apache.org
> =C4mne: Triplet Extraction with OpenNLP
>
> Hello all,
>
> I am interested in performing Triplet Extraction.
>
> For that, I need to traverse the parse tree.
>
> I know how to use the ChunkMe, however I am not sure how to use the Parse=
r
> so that I can create a tree to traverse it.
>
> Ideally, I want to obtain the subject, predicate and object.
>
> To find the subject I need to search in the NP subtree selecting the firs=
t
> descendent of NP that is a Noun via breadth first search.
>
> To find the predicate I will search the VP subtree, the deepest verb
> descendent on that tree will give the predicate.
>
> Now for the object(s) they can be in 3 different subtrees. PP, NP and ADJ=
.
> In NP and PP they will be the first noun while on the ADJ we need to loca=
te
> the first adjective.
>
> Therefore, what I need to learn is how to create the parser and the main
> tree so that I can navigate the subtrees.
>
> Thanks for the help,
>
> Carlos.
>

--f46d04428cc2e27c4504e74f0492--