Reuters and ML

A couple of months ago I had a look at the Calais initiative, which provides a free API to semantic information gathered by Reuters. But it was interesting to find out about its new service for sentiment analysis, called Reuters NewsScope Sentiment Engine.

I was, somehow, surprised that it's a paid service. I have no clue on the quality of their classifiers neither about its costs, but the decision to employ machine learning to attend this common need (specially in its beloved financial world) really indicates Reuters' position as an IT company.

I guess it's all still in the information provider realm, but since we trend towards an era where most information will not be digested by human beings - one has to invest in algorithmic information retrieval strategies in order to deliver knowledge, and machine learning definitely remains at the core of it.



* as a side note, Reuters seems to use the .NET framework for all its development. Is it odd that most software companies no not care about multi-platform solutions?

MASSES

A competição Multi-Agent System for Stock Exchange Simulation (MASSES) foi lançada no dia 10/09/2008 e será realizada durante o XXII Simpósio Brasileiro de Engenharia de Software (SBES), que ocorrerá de 13 a 17 de outubro/2008 em Campinas-SP.

Dêem uma olhada no site oficial.

Hug a developer

A day in the life of a developer is indeed quite amazing.

Microsoft's Solver Foundation

The new iniciative seems very promising! It's also interesting how Microsoft is focusing on DSLs and declarative/functional programing (LINQ, F#) to enabled easier modeling of problems.

The exact features of this new managed library are not clear. Will download to check it out.

RoboChamps Winners

I almost joined the competition for one of the disciplines in my masters' degree. I cannot stress how cool is the Concurrency and Coordination Runtime and the folks at Microsoft are doing an amazing jog applying it to the field Robotics.

It was to great surprise that I saw that brazilian was one of the winners of the first challenge. His name is Jackson Matsuura and he's a professor of ITA (Instituto de Tecnologia da Aeronáutica). Cool \m/ !

Go read the oficial news here.

Giant IBM seizing pot of gold

Wow. Thesee are stunning news. IBM to acquire ILOG must be one of the cleverest move IBM has made in the past five years. It will definitely enforce its position as world's top business solution provider.

Open source solutions are still very poor at this point in time. COIN-OR, being the best, has still a lot of ground to cover to even become a serious option.

Alan king, from IBM, has commented on COIN-OR mailing list saying that "No-one from IBM responded to this" and "The message from our management is that IBM Research is committed to open source, and COIN-OR in particular". His view is optimistic in the sense that the impact could "could go both ways: adding ILOG to IBM creates more potential distractions for IBMers who contribute to COIN, on the other hand, incorporating ILOGers into IBM means that the pool of potential contributors to COIN will go up!".


I guess we will be hearing a lot more about operation research and optimization in the tech news throughout the next few months.

ruby is the most beloved language!

Matz did a little experiment using google to calculate love/hate ratio for some programming languages:



The screenshot is from RubyConf 2007's key note from Confreaks, which has a lot of videos that are worth watching. By the way, just watched Ryan Davies's video at GoRuCo 2008 and highly recommend it.

We're looking for the mouse!

I was lucky to find, on twitter!, a reference to
Clay Shirky's essay and I really suggest it to everyone.

I rarely watch TV and am actually slightly proud of that fact. The information flows too slowly and the format isn't efficient at all - the device seems appropriated only to convey uni-directional entertainment content. And remember that even books function better - we can annotate it, read it fast, fast-forward it, etc.

It wouldn't be fair if the TV gets replaced by a computer with a browser and gets still to be called "TV". Perhaps, one shouldn't care for the name too much - but, heck, I am a computer scientist.

I'm currently working for globo.com which is a part of the larger Rede Globo (media corporation -> large revenue from TV), and the article relates to how I feel towards the greater potential of the web company in comparison to that of the TV itself. As it happens across industries, computers/IT/programmers should be core to anything that really matters.


ps: As a side note, it was also amusing that he chose to reference wikicrimes, which is strongly similar to a project I'm currently working :)

WebSlice, Activity XML and IE8

As part of the Microsoft Open Specification Promise, both WebSlice and Activity are really cool ideas.

I could not imagine anything much simpler than what was proposed. And while some people ignorantly keep reject everything Microsoft, I'm pleased to see the old activex days behind and very excited to welcome the new Microsoft to the new web!

Scale up!

These two open source projects seem awesome: Hypertable and Hadoop. Both inspired in most of Google's research papers.

I currently had a go with MogileFS - but not sure how it compares with Haddop's distritued file system. And how does Gearman compare with the Reduce functionality of Hadoop?


Hopefully will find the time to play with these technologies :)

Twitter Badge with links

I just added the twitter badge to this blog, but it annoyed me that the text was not linkified as it is in twitter's page.

hence, had to modify it slightly:
http://www.pedro.not.br/js/blogger.js (link probably won't be permanent)

Hopefully, this patch will be applied in the original too, since it's definetly a feature most users would want.

DSL

Tropecei no link para o esboço do livro do Martin Fowler sobre DSLs nesta entrevista sobre DSL com Phillip Calçado (ex-globo.com).

Achei curioso os exemplos estarem em C#. Torço para o Ayende conseguir lançar um livro antes em boo! :)

Urchin >> Machine Learning!

Google made available Urchin. It's an interesting move to enable enterprises do some data crunching... :)

I'll try get my hands on the beta.

Terracotta

Terracotta seems an interesting option for clustering java aplications. Although very limited in comparison with GigaSpaces, it's open source and the fact it manipulates your bytecode at the JVM level might make it really simple for developers deploy distributed apps.

I should also take a deeper look at the asm project - very cool!

TuneWiki

TuneWiki is powered by GigaSpaces and an overview of the architecture can be seen here.

Stay tuned :)