Role of white-box logic/models in cTAKES

[Peter Klügl <pe...@averbis.com>]

Hi,

my (uninformed) view on cTAKES was that it is mainly based on black-box
machine learning models.

There were some mentions of rule-based approaches on the mailing list
and a quick look in the source code revealed to me some functionality
that is based on FSMs and regular expressions (and the grey area of rule
logic implemented in plain java).

I'm just curious. Is this code actively used in cTAKES and is there a
general position of the cTAKES community on rules-based/white-box
approaches?

Best,

Peter

Re: Role of white-box logic/models in cTAKES

[Peter Klügl <pe...@averbis.com>]

Hi,

Am 05.08.2015 um 22:16 schrieb Pei Chen:
> Peter,
> Good to hear from you again!
> Yes, I believe there are some regex and rules based annotators that
> are in used (and probably the future for as long as it out performs
> other methods for certain tasks.)
> I don't think there is specific position form the community on this
> approach.  (ASF's 'Do-acracy')
> Were you thinking of writing some Annotators in Ruta?

I was just curious.

Actually, I write quite a lot rule-based annotators these days, but
always in a commercial context (and mostly for other sectors). So, I
cannot contribute them to UIMA or cTAKES unfortunately. I am still
planning to provide ready to use ASL rule-based components for Ruta, but
all the time I can invest is spent on the system itself (so that it
performs better in an industrial setting).

In case someone wants to create rule-based Ruta annotators for cTAKES,
be sure that he/she can count on my help and support.

Best,

Peter


> --Pei
>
>
> On Wed, Aug 5, 2015 at 3:47 AM, Peter Klügl <pe...@averbis.com> wrote:
>> Hi,
>>
>> my (uninformed) view on cTAKES was that it is mainly based on black-box
>> machine learning models.
>>
>> There were some mentions of rule-based approaches on the mailing list
>> and a quick look in the source code revealed to me some functionality
>> that is based on FSMs and regular expressions (and the grey area of rule
>> logic implemented in plain java).
>>
>> I'm just curious. Is this code actively used in cTAKES and is there a
>> general position of the cTAKES community on rules-based/white-box
>> approaches?
>>
>> Best,
>>
>> Peter

Re: Role of white-box logic/models in cTAKES

[Pei Chen <ch...@apache.org>]

Peter,
Good to hear from you again!
Yes, I believe there are some regex and rules based annotators that
are in used (and probably the future for as long as it out performs
other methods for certain tasks.)
I don't think there is specific position form the community on this
approach.  (ASF's 'Do-acracy')
Were you thinking of writing some Annotators in Ruta?

--Pei


On Wed, Aug 5, 2015 at 3:47 AM, Peter Klügl <pe...@averbis.com> wrote:
> Hi,
>
> my (uninformed) view on cTAKES was that it is mainly based on black-box
> machine learning models.
>
> There were some mentions of rule-based approaches on the mailing list
> and a quick look in the source code revealed to me some functionality
> that is based on FSMs and regular expressions (and the grey area of rule
> logic implemented in plain java).
>
> I'm just curious. Is this code actively used in cTAKES and is there a
> general position of the cTAKES community on rules-based/white-box
> approaches?
>
> Best,
>
> Peter