Friday, May 12, 2017

Universal Dependencies for Textual Inference

In February 2017 I gave a talk at Nuance's AI Lab about a small experiment that I did with Alexandre Rademaker and Fabricio Chalub, from IBM Research, in Rio de Janeiro.

We used the SICK corpus created by the COMPOSES project in Italy, (devised to downplay the difficulties of language understanding), together with Parsey McParseFace,  Google's self-declared most accurate model in the world, to create logical representations of the sentences in SICK.

I reasoned that with all the fanfare about the advances of neural nets in NLP, considering that the corpus is simplified on purpose, these representations should be accurate enough to allow us to do the inferences required. Unfortunately between small errors here and there and big errors in the disambiguation, this experiment did not work the  way I expected it to.

You can see the slides in slideshare. Since the disambiguation, using Freeling's version of personalized PageRank, didn't work at all, we have all the possible word senses from WordNet at the moment in the GitHub repository. Now I am thinking about disambiguation, but also thinking about the kinds of inferences that we want and don't want to make. This project was suggested by Danny Bobrow several years ago.

Tuesday, May 2, 2017

Hurwitz-Radon Transformations

Looking for something else I discovered that my masters' thesis had been put online by the Maths department of PUC. Yay!!!

Had to ask Noemi to download it for me, as it required a PUC login. (a few months back I had asked fedex to scan it for me, but they wanted 600 dollars, phew!)

My Master's supervisor Duane Randall is on the left, in the back of this very old picture.

Now I am planning to latex and translate it, as quartenions and octonions are still something I enjoy thinking about.

Friday, April 21, 2017

Among the Philosophers

 Last week I was in Seattle for the APA/ASL meeting. The APA (American Philosophy Association) is kind of new to me, I had never been to one of their meetings before.

The program for the sessions I was is in this blog post Association for Symbolic Logic at the Pacific APA | Richard Zach.   The picture was taken by Richard Zach, the organizer of the session.

Mine was a very short talk about the Lambek Calculus  and categorical models for it, using Dialectica categories. This is joint work with Harley Eades III, updating the old paper I talked about in Lambek Calculus and Dialectica Categories.

I had one very good question, on why do I not believe in the Lambek calculus as a way of modelling language. The answer is complicated and long, so it will be left for a later blog post. I have many good friends who love the Lambek calculus exactly for this application, so I have to be careful how I describe my issues. Anyways, I do love the Lambek calculus as a logical system and the slides for the short talk are in slideshare.

Friday, March 3, 2017

Legal Start-ups

I have not written for far too long. Lots of work, lots of deadlines and especially lots of reviewing. But working with other people always helps.

Sabrina Praduroux works with Luigi di Caro in the University of Torino, Italy, as part of the MIREL (MIning and REasoning with legal text) project. Last year she had a visiting fellowship  to Stanford,  and we did some work together mapping the legal start-ups in the Valley. A report was written for the MIREL workshop and it is posted at my webpage.

Reasoning with legal texts is something I would like to do more. Happily Luigi is visiting again this Summer. I don't know yet if Sabrina and Livio will be visiting too. When they come next I have to ask what the logo of MIREL is all about. I find it a bit disturbing, I must say.