ctakes-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Erick Velazquez <erick.lero...@gmail.com>
Subject Re: Building a dictionary from ontologies [EXTERNAL]
Date Wed, 24 Jan 2018 16:23:29 GMT
Hi Sean, 
Thanks for your help. So now I got my BSV file, but I don’t find the documentation that
explains how to include it into the cTAKES analysis. Is there any document that can help me?
Kind regards, 

Erick 

> On Jan 23, 2018, at 5:21 PM, Finan, Sean <Sean.Finan@childrens.harvard.edu> wrote:
> 
> Hi Erick,
> 
> Each synonym gets a single line in the bsv file.  So:
> 
> OWL00770 | TOOO | Right parietal lobe
> OWL00770 | TOOO | parietal lobe, right
> OWL00770 | TOOO | parietal lobe on the right
> 
> If you don't have a tui then you can simplify the lines by leaving the second column
empty:
> 
> OWL00770 | | Right parietal lobe
> OWL00770 | | parietal lobe, right
> OWL00770 | | parietal lobe on the right
> 
> 
> Sean
> 
> -----Original Message-----
> From: Erick Velazquez [mailto:erick.lerouge@gmail.com] 
> Sent: Tuesday, January 23, 2018 5:15 PM
> To: dev@ctakes.apache.org
> Subject: Re: Building a dictionary from ontologies [EXTERNAL]
> 
> Hi Sean,
> 
> Thank you for your answer!
> 
> I would like to show you my results. As an example, I got this:
> 
> 
> 
> OWL00770 | TOOO | Right parietal lobe
> 
> 
> 
> The third column is the text.
> 
> You suggested to me to use the uri but what I get from the ontology is only a web link.
Then, I don’t use the preferred text option.
> 
> When you say that text should also contain synonyms what do you mean? That means that
every token in the text column is considered as a synonym? Then in my example, right would
be interpreted as a synonym of parietal and lob?
> 
> 
> 
> Kind regards,
> 
> 
> 
> Erick Velazquez 
> 
> 
>> On Jan 22, 2018, at 2:47 PM, Finan, Sean <Sean.Finan@childrens.harvard.edu>
wrote:
>> 
>> Hi Erick,
>> 
>> There is a fourth option that should work
>> 
>> cui | tui | text | preferredText
>> 
>> I would create an importer that creates a -fake- cui.  The cui need not (in this
case should not) start with 'C'.  So, I would import per-owl uri using something like OWL00001.
 
>> 
>> tui can be empty, in which case "T000" will be used, =forcing ctakes to create annotations
of unknown semantic type.  
>> 
>> text(s) should contain your synonym(s).
>> 
>> preferredText can be your owl uri.
>> 
>> This should allow you to fake it with an imported owl.  Upon deconstruction of the
cas you will want to look at the preferredTerm for each annotation and ignore the cui and
tui.
>> 
>> Sean 
>> 
>> ________________________________________
>> From: Erick Velazquez <erick.lerouge@gmail.com>
>> Sent: Monday, January 22, 2018 11:14 AM
>> To: dev@ctakes.apache.org
>> Subject: Building a dictionary from ontologies  [EXTERNAL]
>> 
>> Hello,
>> 
>> I’m building a dictionary from an ontology (OWL), but there is no CUI, neither
TUI in the information. Since the format of a dictionary in cTAKES is CUI | TUI | TEXT, or
CUI | TEXT, is there any specification to create CUIs for terms?
>> Thanks,
>> 
>> Erick
> 


Mime
View raw message