ctakes-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Tomasz Oliwa <ol...@uchicago.edu>
Subject RE: cTAKES dictionary lookup behavior question
Date Mon, 16 Nov 2015 16:36:05 GMT
Sean,

I created a JIRA entry for this bug at: https://issues.apache.org/jira/browse/CTAKES-389

It would be great you could check in a fix for it.

Regards,
Tomasz

________________________________________
From: Finan, Sean [Sean.Finan@childrens.harvard.edu]
Sent: Monday, November 16, 2015 10:20 AM
To: dev@ctakes.apache.org
Subject: RE: cTAKES dictionary lookup behavior question

Hi all,

This is not intended behavior, it is a bug.  I will check in a fix soon ...

-----Original Message-----
From: Tomasz Oliwa [mailto:oliwa@uchicago.edu]
Sent: Thursday, November 12, 2015 6:53 PM
To: britt fitch; dev@ctakes.apache.org
Subject: RE: cTAKES dictionary lookup behavior question

Britt,

I observed it also depends on what the "missed" word is.

"baby to" , "baby too" match C1305907 of "baby tooth", however "baby token" does not match
it.
"electrolyte le", "electrolyte lev" match C0428284 "electrolyte level", but "electrolyte dev"
does not match.

It seems if the "missed" word contains the same characters that the word found in the fast
dictionary starts with, a match is made?

Is there any way to tweak or customize this behavior?

Thanks,
Tomasz


________________________________
From: britt fitch [britt.fitch@wiredinformatics.com]
Sent: Thursday, November 12, 2015 5:36 PM
To: dev@ctakes.apache.org
Subject: Re: cTAKES dictionary lookup behavior question

The rare words, given the example terms below are "primary", "milk", and "baby".
The lookup allows for a certain number of "misses".
The "baby to" hits on "baby" as the rare word.
"baby to" compared to "baby tooth" is 1 "miss" and qualifies as a match. (in practice, if
I recall correctly, "to" is actually discarded entirely, so the comparison is actually "baby"
: "baby tooth").

Others can correct my napkin logic though.

This is a pretty common scenario when a single term ends up matching to a larger term because
of the allowance of misses.

For example:

"oxygen" > "oxygen therapy"
"pathology" > "pathology department" , "pathology procedure"
"exercise" > "exercise pain management"

Those are just some quick examples. It depends heavily on what the ontology contains though.

Cheers,

Britt


Britt Fitch
Wired Informatics
265 Franklin St Ste 1702
Boston, MA 02110
https://urldefense.proofpoint.com/v2/url?u=http-3A__wiredinformatics.com&d=BQIF-g&c=qS4goWBT7poplM69zy_3xhKwEW14JZMSdioCoppxeFU&r=fs67GvlGZstTpyIisCYNYmQCP6r0bcpKGd4f7d4gTao&m=nrheHTAYzgKYX9njwAR5G_NJXfSe_sbYbOMaifjWZwQ&s=6LcknYupSIqPd8Uml-tNRhwLudfDpVLBcC5JjZFhFQo&e=
Britt.Fitch@wiredinformatics.com

On Nov 12, 2015, at 6:27 PM, Tomasz Oliwa <oliwa@uchicago.edu<mailto:oliwa@uchicago.edu>>
wrote:

Hi,

cTAKES has a dictionary lookup behavior that I cannot explain, you can verify the queries
via the cTAKES demo that has been posted here at: https://urldefense.proofpoint.com/v2/url?u=http-3A__52.27.22.206-3A8080_index.jsp&d=BQIF-g&c=qS4goWBT7poplM69zy_3xhKwEW14JZMSdioCoppxeFU&r=fs67GvlGZstTpyIisCYNYmQCP6r0bcpKGd4f7d4gTao&m=nrheHTAYzgKYX9njwAR5G_NJXfSe_sbYbOMaifjWZwQ&s=UmyBQ5X4UBJggOqmIQkANeD0eUz0nrLqGN8Z6__iB8o&e=
 but it also happens with the current 3.2.2 version and the fast dictionary UMLS lookup

SENTENCE:  Took  the baby to  the hospital.
           VB   DT   NN  IN  DT     NN
          |===|     |======|
          Event     Anatomy
                    C1305907

It finds the "baby tooth" annotation. The only CUI texts in the default fast dictionary for
C1305907 are

C1305907|primary tooth
C1305907|milk tooth
C1305907|baby tooth

How can "baby to" trigger the "baby tooth" annotation?

Regards,
Tomasz


Mime
View raw message