Tuesday, November 25, 2014

DannyFest 2008

Today I was reminded that we (Cleo and I) have not, yet, lived up to our promise of editing a special volume for Danny.

We had the Festschrift DannyFest at PARC in 2008
http://www.parc.com/event/634/daniel-g-bobrow-festschrift-celebration-dannyfest.html. The videos of the talks are in the blog there, as are the pictures in flckr. But there is no volume, yet. And we should try to make good on that.

As soon as I manage to get IMLA 2011 done, this will be the new project.

Sunday, November 23, 2014

Wow!!! Google Translate is using our work!

















So cool. I was reading this paper about how to create a logical ontology associated to a lexical ontology
and what did I discover? That our work on a Portuguese WordNet is used by Google Translate!
Check it out Inside Google Translate – Google Translate and you'll see the
OpenWordNet-PT!!!

Yay!!!

Wednesday, November 5, 2014

Publishing Models




The world of scientific publishing is in turmoil. I don't know how the ACL manages to get all their papers online for free for everyone, but
I want to have the same happening in the other areas that I work.

Meanwhile I have to choose better the places where I will send my work.
In the small sample above, Springer got another 200 dollars and I haven't been able, yet, to share the USB memory sticks with my colleagues at Nuance.
(thinking about how to do it, though...)
I never been very strategic with where to send papers, a silly mistake to make all these years, time to put it right, I guess.
(Here's a blog post about it, don't agree with all he says, but some definitely makes sense, http://idibon.com/top-nlp-conferences-journals/ )

hmm, forgot to mention that I only went to two of these conferences:
the one in Nepal my collaborator Gerard de Melo (from Tsinghua University) presented the paper
Sense-Specific Implicative Commitments. and the PROPOR one, my collaborator Alexandre Rademaker
(from FGV and IBM/Research Rio) presented the paper
NomLex-PT: A Lexicon of Portuguese Nominalizations.

Tuesday, October 14, 2014

Happy Ada Lovelace Day 2014!

This year for Ada Lovelace Day an algebraic logician Helena Rasiowa, famous for the Rasiowa-Sikorski Lemma in Set Theory. I have wanted for a long time to buy her book with Sikorski "The Mathematics of Metamathematics, but I also would've liked to have An Algebraic Approach to Non-Classical Logics.

Anyways one can always read Melving Fitting's  tribute to her.

Monday, October 13, 2014

World domination Brain-style

As I was discussing in "A modest proposal for Portuguese"  we have so far two open-source lexical resources for Portuguese (OpenWordNet-PT and NomLex-PT) but I can see many different applications for these and many interesting ways of building from these in to new resources and systems.

Despite the lack of official funding our group  (Alexandre RademakerGerard de Melo, Livy Real, Claudia Freitas, Dario Oliveira, Suemi Higuchi) is growing and we even managed to finally get a paper in one of the most important conferences in Computational Linguistics in Brazil, PROPOR 2014 edition. This was about the use of corpora to improve NomLex-PT, the slides (presented by Alexandre) are here.

This last September, apart from our Workshop on Logics and Ontologies for Natural Languages, we've managed to have at least  three informal working meetings (one in the Livraria Argumento, one in FGV, after my talk and one in PUC with Dario, who's usually in Sao Paulo).

 Besides the FGV talk, I spoke both at PUC-Rio and at COPPE about our generic plan of world domination Brain-style... Jokes apart, I tried to describe how we can work in several fronts (building lexical resources, using these resources for information extraction, ontology building, reasoning with logic from text  and question answering,  amongst others..) in an informal, but tightly integrated collaboration.

I think it's working. This year we had some eight papers in total. Two at the Global WordNet Association meeting proceedings: one a progress report on OpenWordNet-PT, the other a description of how we created NomLex-PT from a translation of the original NomLex enriched with electronic dictionaries and manually verified. Then  one poster at LREC, explaining how we grew NomLex-PT  and then integrated it with OpenWorNet-PT. Then the PROPOR paper above which extends NomLex-PT with corpora information from the AC/DC collection. And two extra posters for the workshop TorPorEsp, one of the Verb Lexicon  of OpenWordNet-PT and one completing our stock of Portuguese nominalizations, this time using the Spanish lexicon AnCora-Nom.

Need to add to this list the paper on the application of the work on Portuguese lexical resources to our historical corpus, the Dictionary of Historic Brazilian Biographies, Fun Information Extraction from a Historical Dictionary and the work in English on the Implicative Lexicon (ImpLex) at CICling, Sense-Specific Implicative Commitments.

Finally there is the paper with Vivek Nigam on using term rewriting in Maude to reason with natural language representations, "Towards a Rewriting Framework for Textual Entailment". I gave a very short talk about it in Brasilia, at LSFA 2014, and a more leisurely on at COPPE, where after lunch Celina, Simone, Petrucio and I went to see this beautiful colonial church.


Thursday, October 9, 2014

MAA Profile

Yep, there are days that you feel that the work you do gets no recognition, no one pays any attention, and when they do, is simply to say it's rubbish.. So you might as well collect some pick-me-up posts, here and there, to remind yourself that each day is a different day.

This is one of these "pick-me-up posts": a while back the Mathematical Association of America (MAA)  asked me for a working profile, while I was still at PARC. Don't remember when this was done, but the picture is from 2004, the year Paola graduated. Given the way things disappear from the web, it seems sensible to take a picture and keep it for myself.


Natural Language and Computer Science (NLCS2) in Vienna

Some complain of too many tourists, too many "mozarts", too much gilding in Vienna. Personally I had a great time and enjoyed it very much.

It is true that it was difficult to see the talks you wanted in the Vienna Summer of Logic, as there were so many. It's true that in July the place is full of tourists and the many guys dressed as Mozart are a bit off putting, but we did have a great Workshop on Natural Language and Computer Science.

First it was fun that the phd student assigned by the central organization, Carol Blasio, to help out with machines, water, air conditioning, etc, happened to be a friend from Brazil. This was a good omen.

Then our workshop (organized by Larry Moss, Christian Retore' and myself) was full,  a few people decided not to come in, because there was no place to sit, or even stand. The discussions were good: I hate workshops where there are no questions and no dispute. The point of academic meetings is disagreeing, politely, so that everyone can improve their arguments. If there are no disagreements, then the organizers stood too much in their comfort zone and this was definitely not the case in our workshop.

We merged our workshop Second Workshop on Natural Language and Computer Science, which had many submissions, with the workshop on Natural Language Services for Reasoners.
This caused some problems with the program, as some people didn't realize that we had a full day of talks, assuming instead that there was a huge break, when the talks part of  NLSR were happening.
Even now, when trying to find a program of the workshop that shows what actually happened I have difficulties. Never mind, all's well that ends well.

I decided to write this post only now, as I was waiting for the technical report with the papers to be ready. Of course the technical report from the University of Coimbra (lovely place!) has been ready for a few weeks now, but other things intervened. Anyways before it gets too late I would like to thank Pedro Quaresma for producing the Technical Report and  my managers at Nuance for allowing me to organize meetings like this. (Shame we could not get all the papers that were presented into the volume ...)

This gives me an excuse to write another blog post with the program and some more comments later on.