You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@ctakes.apache.org by Oleg Bogatiryov <ol...@ctco.lv> on 2017/02/20 07:31:40 UTC
Ctakes relation extraction
Hello to everyone.
I am pleased to join the group.
I am trying to extract relation from the document.
Ideally I'd like to get the graph or tree of dependencies/relations from
the clinical documents.
Could you please let me know how can I achieve it ?
I am able to run CVD and RelationExtractorAggregate analysis engine but
there is no useful information
in results that can be used in order to build a relation graph.
Thanks in advance,
Oleg.
Re: Ctakes relation extraction
Posted by "Dligach, Dmitriy" <dd...@luc.edu>.
Hi Oleg,
You may want to look into learning about UIMA and UIMAFIT.
Once you are somewhat comfortable with these frameworks, take a look at the package:
ctakes/ctakes-relation-extractor/src/main/java/org/apache/ctakes/relationextractor/pipelines/
There are several pipelines that may work for you (e.g. take a look at RelationExtractorPipeline.java). Many of them should generate XMI files which contain the relation annotations you are looking for. Once you have generated the XMI files, you can parse them using UIMAFIT. E.g. take a look at RelationAnnotationViewer.java which you should use as an example.
Hope this helps!
Dima
> On Feb 24, 2017, at 03:07, Oleg Bogatiryov <ol...@ctco.lv> wrote:
>
> Hi Dmitriy.
>
> Thank you for your reply.
>
> I am not sure how can I extract location_of and degree_of.
> AggregateTemplateFiller doesn't return anything in CVD as well as
> TemporalAggregatePipeline.
> There is BinaryTextRelation class that as I understand should fill in
> relations but no data is displayed.
>
> I was trying to execute RelationExtractorPipelineSingleCas but it doesn't
> print anything related to relations.
>
> Could you please help me with extraction of relations from the med text ?
>
> As I am a new to ctakes and uima could you please provide step to step
> instructions and exact classes to use.
>
>
> Thanks in advance,
> Oleg.
>
> -----Original Message-----
> From: Dligach, Dmitriy [mailto:ddligach@luc.edu]
> Sent: 20 февраля 2017 г. 17:48
> To: dev@ctakes.apache.org
> Subject: Re: Ctakes relation extraction
>
> Hi Oleg,
>
> The relation extraction AE currently only handles location_of and degree_of
> relations as described here:
>
> https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3994852/
>
> The dependencies are handled by the dependency parser (i.e. a separate
> module).
>
> So you’ll need to run both and collect all annotated relations from the the
> CAS.
>
> Hope this helps.
>
> Dima
>
>
>
>> On Feb 20, 2017, at 01:31, Oleg Bogatiryov <ol...@ctco.lv>
>> wrote:
>>
>> Hello to everyone.
>>
>>
>>
>> I am pleased to join the group.
>>
>>
>>
>> I am trying to extract relation from the document.
>>
>> Ideally I'd like to get the graph or tree of dependencies/relations
>> from the clinical documents.
>>
>>
>>
>> Could you please let me know how can I achieve it ?
>>
>>
>>
>> I am able to run CVD and RelationExtractorAggregate analysis engine
>> but there is no useful information
>>
>> in results that can be used in order to build a relation graph.
>>
>>
>>
>> Thanks in advance,
>>
>> Oleg.
>>
>
RE: Ctakes relation extraction
Posted by Oleg Bogatiryov <ol...@ctco.lv>.
Hi Dmitriy.
Thank you for your reply.
I am not sure how can I extract location_of and degree_of.
AggregateTemplateFiller doesn't return anything in CVD as well as
TemporalAggregatePipeline.
There is BinaryTextRelation class that as I understand should fill in
relations but no data is displayed.
I was trying to execute RelationExtractorPipelineSingleCas but it doesn't
print anything related to relations.
Could you please help me with extraction of relations from the med text ?
As I am a new to ctakes and uima could you please provide step to step
instructions and exact classes to use.
Thanks in advance,
Oleg.
-----Original Message-----
From: Dligach, Dmitriy [mailto:ddligach@luc.edu]
Sent: 20 \u0444\u0435\u0432\u0440\u0430\u043b\u044f 2017 \u0433. 17:48
To: dev@ctakes.apache.org
Subject: Re: Ctakes relation extraction
Hi Oleg,
The relation extraction AE currently only handles location_of and degree_of
relations as described here:
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3994852/
The dependencies are handled by the dependency parser (i.e. a separate
module).
So you\u2019ll need to run both and collect all annotated relations from the the
CAS.
Hope this helps.
Dima
> On Feb 20, 2017, at 01:31, Oleg Bogatiryov <ol...@ctco.lv>
> wrote:
>
> Hello to everyone.
>
>
>
> I am pleased to join the group.
>
>
>
> I am trying to extract relation from the document.
>
> Ideally I'd like to get the graph or tree of dependencies/relations
> from the clinical documents.
>
>
>
> Could you please let me know how can I achieve it ?
>
>
>
> I am able to run CVD and RelationExtractorAggregate analysis engine
> but there is no useful information
>
> in results that can be used in order to build a relation graph.
>
>
>
> Thanks in advance,
>
> Oleg.
>
Re: Ctakes relation extraction
Posted by "Dligach, Dmitriy" <dd...@luc.edu>.
Hi Oleg,
The relation extraction AE currently only handles location_of and degree_of relations as described here:
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3994852/
The dependencies are handled by the dependency parser (i.e. a separate module).
So you’ll need to run both and collect all annotated relations from the the CAS.
Hope this helps.
Dima
> On Feb 20, 2017, at 01:31, Oleg Bogatiryov <ol...@ctco.lv> wrote:
>
> Hello to everyone.
>
>
>
> I am pleased to join the group.
>
>
>
> I am trying to extract relation from the document.
>
> Ideally I'd like to get the graph or tree of dependencies/relations from
> the clinical documents.
>
>
>
> Could you please let me know how can I achieve it ?
>
>
>
> I am able to run CVD and RelationExtractorAggregate analysis engine but
> there is no useful information
>
> in results that can be used in order to build a relation graph.
>
>
>
> Thanks in advance,
>
> Oleg.
>
Re: Ctakes relation extraction
Posted by "Abramowitsch, Peter" <pa...@hearst.com>.
Agree completely with Guergana. The causality annotations reflect only
semantics. In no way do they reflect "truth" for any given patient.
However, these relations can be used to acquire data for statistical
purposes - for instance to reveal that % of lipid panels that were
conducted to rule out xyz.
Peter
On 2/20/17, 10:24 AM, "Savova, Guergana"
<Gu...@childrens.harvard.edu> wrote:
>A word of caution -- the definition of a causal relation in the clinical
>narrative is much vaguer than in the general domain. Even if the clinical
>narrative asserts a relation between a medication and a sign/symptom
>(a.k.a. adverse event), it might not be necessarily the case. Even more,
>the lack of an asserted explicit causal relation does not mean that such
>is lacking.
>Regards,
>--Guergana
>
>Guergana Savova, PhD, FACMI
>Associate Professor
>PI Natural Language Processing Lab
>Boston Children's Hospital and Harvard Medical School
>300 Longwood Avenue
>Mailstop: BCH3092
>Enders 144.1
>Boston, MA 02115
>Tel: (617) 919-2972
>Fax: (617) 730-0817
>Guergana.Savova@childrens.harvard.edu
>Harvard Scholar:
>https://urldefense.proofpoint.com/v2/url?u=http-3A__scholar.harvard.edu_gu
>ergana-5Fk-5Fsavova_biocv&d=DwIFAw&c=B73tqXN8Ec0ocRmZHMCntw&r=5LM1YwNyMUq7
>CWiSepCCsjTjwuVF4uswNF8BK5Orm10&m=uOutBBTdNsYiDP3HAuF-_UFIGcTT6qqeuYEBj7dQ
>N2I&s=lj2ltefhatVPII-9YxnA0QHcsvaeL22-sVSCFLZ_QQM&e=
>
>-----Original Message-----
>From: Abramowitsch, Peter [mailto:pabramowitsch@hearst.com]
>Sent: Monday, February 20, 2017 1:19 PM
>To: dev@ctakes.apache.org
>Subject: Re: Ctakes relation extraction
>
>There is another, more generic NLP engine, StanfordCoreNLP which does
>have more advanced CORel annotation capabilities, but it is not
>specifically tuned to clinical concepts and relationships. So for
>instance, causal relationships might be detected if expressed in standard
>english, but certainly not by clinical acronyms. But you might play
>with it just to get a sense of what is possible in the open source space.
>
>Regards, Peter
>
>
>
>On 2/19/17, 11:31 PM, "Oleg Bogatiryov" <ol...@ctco.lv> wrote:
>
>>Hello to everyone.
>>
>>
>>
>>I am pleased to join the group.
>>
>>
>>
>>I am trying to extract relation from the document.
>>
>>Ideally I'd like to get the graph or tree of dependencies/relations
>>from the clinical documents.
>>
>>
>>
>>Could you please let me know how can I achieve it ?
>>
>>
>>
>>I am able to run CVD and RelationExtractorAggregate analysis engine but
>>there is no useful information
>>
>>in results that can be used in order to build a relation graph.
>>
>>
>>
>>Thanks in advance,
>>
>>Oleg.
>>
>
RE: Ctakes relation extraction
Posted by "Savova, Guergana" <Gu...@childrens.harvard.edu>.
A word of caution -- the definition of a causal relation in the clinical narrative is much vaguer than in the general domain. Even if the clinical narrative asserts a relation between a medication and a sign/symptom (a.k.a. adverse event), it might not be necessarily the case. Even more, the lack of an asserted explicit causal relation does not mean that such is lacking.
Regards,
--Guergana
Guergana Savova, PhD, FACMI
Associate Professor
PI Natural Language Processing Lab
Boston Children's Hospital and Harvard Medical School
300 Longwood Avenue
Mailstop: BCH3092
Enders 144.1
Boston, MA 02115
Tel: (617) 919-2972
Fax: (617) 730-0817
Guergana.Savova@childrens.harvard.edu
Harvard Scholar: http://scholar.harvard.edu/guergana_k_savova/biocv
-----Original Message-----
From: Abramowitsch, Peter [mailto:pabramowitsch@hearst.com]
Sent: Monday, February 20, 2017 1:19 PM
To: dev@ctakes.apache.org
Subject: Re: Ctakes relation extraction
There is another, more generic NLP engine, StanfordCoreNLP which does have more advanced CORel annotation capabilities, but it is not specifically tuned to clinical concepts and relationships. So for instance, causal relationships might be detected if expressed in standard
english, but certainly not by clinical acronyms. But you might play
with it just to get a sense of what is possible in the open source space.
Regards, Peter
On 2/19/17, 11:31 PM, "Oleg Bogatiryov" <ol...@ctco.lv> wrote:
>Hello to everyone.
>
>
>
>I am pleased to join the group.
>
>
>
>I am trying to extract relation from the document.
>
>Ideally I'd like to get the graph or tree of dependencies/relations
>from the clinical documents.
>
>
>
>Could you please let me know how can I achieve it ?
>
>
>
>I am able to run CVD and RelationExtractorAggregate analysis engine but
>there is no useful information
>
>in results that can be used in order to build a relation graph.
>
>
>
>Thanks in advance,
>
>Oleg.
>
Re: Ctakes relation extraction
Posted by "Abramowitsch, Peter" <pa...@hearst.com>.
There is another, more generic NLP engine, StanfordCoreNLP which does
have more advanced CORel annotation capabilities, but it is not
specifically tuned to clinical concepts and relationships. So for
instance, causal relationships might be detected if expressed in standard
english, but certainly not by clinical acronyms. But you might play
with it just to get a sense of what is possible in the open source space.
Regards, Peter
On 2/19/17, 11:31 PM, "Oleg Bogatiryov" <ol...@ctco.lv> wrote:
>Hello to everyone.
>
>
>
>I am pleased to join the group.
>
>
>
>I am trying to extract relation from the document.
>
>Ideally I'd like to get the graph or tree of dependencies/relations from
>the clinical documents.
>
>
>
>Could you please let me know how can I achieve it ?
>
>
>
>I am able to run CVD and RelationExtractorAggregate analysis engine but
>there is no useful information
>
>in results that can be used in order to build a relation graph.
>
>
>
>Thanks in advance,
>
>Oleg.
>