Re: AI KR CG interests

Thanks a lot for sharing RR
If and when you have a moment, it may be useful to extract some quick
lessons from what you are doing

- problem
- how and why KR can help solve the problem,
- what specific KR technique/solution is applied
- how the solution works
- what lesson can be generalized

The problem I am working on now is the fragmentation/lack of coherence of
the body of knowledge of
AI, in particular in relation to ethical challenges, accountability,
explainability etc
(improved AI KR as a path to  explainability, reliability etc?)
both within the standardization domain (across standards such as IEEE ISO
etc)
as well as within W3C .

I d be interested to hear from the others as well, maybe we can compile some
useful cases from CG members

Cheers

PDM

On Thu, Jun 13, 2019 at 9:00 PM Ronald Reck <rreck@rrecktek.com> wrote:

>
>
> On Thu, 13 Jun 2019 07:43:29 +0800, Paola Di Maio <paoladimaio10@gmail.com>
> wrote:
>
> > Tell us about your interest in AI KR, what can the CG do for you
>
>
> Right now I am working on representing results from NLP semantic (term)
> analysis in the legal domain.
> I render the results in RDF and build DTM/TDM models in R using the
> Predictive Analytics Framework
> product on AWS.
> https://aws.amazon.com/marketplace/pp/B00SK3VN1E
>
> I am working with my partner, Skye Suh, on two specific efforts.
>
> 1. https://sites.google.com/view/fire-2019-aila/track-description
> Our hopes are that our domain representation models provide us a
> competitive advantage.
>
>
> 2. We have a paper at Graphorum 2019 – Chicago, IL
> https://www.dataversity.net/graphorum-2019-chicago-il/
>
> Our presentation focuses on text analytics to assist in making an informed
> decision to decrease the probability of application denial.  More
> specifically, as relates to the non-immigrant visa (L-1A and L-1B) used by
> foreign companies to bring key employees to the United States. The process
> of applying, obtaining, and entering in on L-1 status costs upwards of USD
> 5,000.00, and routinely take 3 to 4 months for a decision.  Further, USCIS
> routinely issues Requests for Evidence (RFE), which require the petitioner
> to resubmit or prepare evidentiary documentation, which results in
> escalated costs and time delays. The use of text analytics can be used to
> increase the probability of obtaining an approved petition.  Our process,
> using The Predictive Analytics Framework, an open source based operating
> environment allows the user to review, extract, and isolate the specific
> use of word choices that promote an approved application. We describe our
> process to construct a document-term matrix (DTM) from petitions and then
> fit a model to the corpus thereby showing how classification can allow us
> to improve the likelihood of petition acceptance. Text analytics can be a
> powerful tool in decision making to decrease time and effort preparing
> applications and deciding whether to continue with the application process.
>
> We also have several AI KR representation interests for object detection
> in computer vision CV.
> http://vpics.sphere188.com/
>
> More specifically, we have been using Haar feature-based cascade
> classifiers for use in security, sports and property management domains.
>