Alexandre Trilla's Blog

Lean Startup hackers were already there back in the early eighties

Thu, 2 May 2013 23:50:00 +0000

The fancy "lean" adjective that accompanies every rocking tech business issue nowadays is an already old story. I found it out the other day while skimming through Steven Levy's groundbreaking book "Hackers: Heroes of the Computer Revolution". Read more

NLP-Tools broadens its capabilities with a RESTful API service

Thu, 18 Apr 2013 23:00:00 +0000

In the software tool development business, the API is the new language of the developers, i.e., the customers. In this regard, nlpTools keeps pace with the evolution of the industry market and introduces its RESTful API service to facilitate its integration. And in that quest for added-value and kaizen it partners with Mashape to handle the commercialisation issues. The original website still maintains the evaluation service, but further performance features now need to be routed through the Mashape nlpTools endpoint. Read more

Foraging ants as living particle filters

Sun, 24 Feb 2013 20:00:00 +0000

Ant colonies are admirable examples of cooperative societies. Some of its members are prepared to build their complex lairs, some others constitute an army to protect their population, some others explore the outer world and gather food, etc. With respect to the latter function, which to me is the most representative of ant colony behaviour, I coded a simple simulation in JavaScript inspired by the js1k competition. Read more

The abridged build-measure-learn loop: innovate and seek excellence

Tue, 12 Feb 2013 22:30:00 +0000

The principal objective of a tech startup (Research and Development also fit the shoes of tech entrepreneurship without loss of generality) is to learn how to build and run a sustainable business where value is created when a new technology invention is matched to customer need. Therefore, validated learning is a fundamental issue in this uncertain quest for success. Read more

A New Year's resolution: get over specialisation and embrace generalisation to face real world industry problems

Tue, 1 Jan 2013 17:00:00 +0000

Regularisation is a recurrent issue in Machine Learning (and so it is in this blog). Prof. Hinton also borrowed the concept in his neural networked view of the world, and used a shocking term like "unlearning" to refer to it. Interesting as it sounds, to achieve a greater effectiveness, one must not learn the idiosyncrasies of the data, one must remain a little ignorant in order to discover the true behaviour of the data. In this post, I revisit typical weight penalties like Tikhonov (L-2 norm), Lasso (L-1 norm) and Student-t (sum of logs of squared weights), which function as model regularisers Read more

Sure, you can do that... and still get an IEEE published article

Mon, 24 Dec 2012 18:00:00 +0000

This year has been rather prolific with respect to the attained number of research publications. The most noteworthy is the one on the IEEE Transactions on Audio, Speech and Language Processing (TASLP), which is entitled "Sentence-based Sentiment Analysis for Expressive Text-to-Speech". Its abstract is posted as follows: Read more

Perceptron learning with the overused Least Squares method

Fri, 2 Nov 2012 00:00:00 +0000

Following Geoffrey Hinton's lectures on Neural Networks for Machine Learning, this post overviews the Perceptron, a single-layer artificial neural network that provides a lot of learning power, especially by tuning the strategy that is used for training the weights (note that Support Vector Machines are Perceptrons in the end). To keep things simple, 1) no regularisation issues will be covered here, and 2) the weight optimisation criterion will be the minimisation of the squared error cost function, which can be happily overused. In another post, the similarity between using the least squares method and the cross-entropy cost through the negative log-likelihood function (as it is reviewed in class) assuming a Gaussian error was already discussed. So using one or the other won't yield much effectiveness improvement for a classic toy dataset sampled from two Gaussian distributions. Read more

Least Squares regression with outliers is tricky

Mon, 23 Jul 2012 23:00:00 +0000

If reams of disorganised data is all you can see around you, a Least Squares regression may be a sensible tool to make some sense out of them (or at least to approximate them within a reasonable interval, making the analysis problem more tractable). Fitting functions to data is a pervasive issue in many aspects of data engineering. But since the devil is in the details, different objective criteria may cause the optimisation results to diverge considerably (especially if outliers are present), misleading the interpretation of the study, so this aspect cannot be taken carelessly. Read more

On using Hacker News to validate a product idea involving NLP and PHP

Thu, 12 Jul 2012 9:00:00 +0000

The first step to creating a valuable product is to discover what it is exactly wanted or needed by the target customers. The Lean Startup process states it straight, and the Pragmatic Programmer even provides a means to find it out by asking Hacker News (HN). HN is a vibrant community of tech people, hackers is its broadest sense... and entrepreneurs (these concepts need not be disjoint), which can provide a lot of insight into the value of a product idea. Read more

The Passionate Programmer in the late-2000s recession

Sun, 3 Jun 2012 23:20:00 +0000

The present receding economy displays a scenario that is wildly unknown, and this inevitably affects the attitude that we take with respect to our careers, reminding us all of the crucial importance to always be heading to where the magic happens. In addition to the renown advice to not settle, what's utterly of value is to stay hungry in this continuously changing world. Read more

Numerical computation platform for the technical university: values to decide on a proprietary or open-source software model

Fri, 30 Mar 2012 19:20:00 +0000

A discussion on the adequacy of a proprietary numerical computation platform like Matlab or a free open source alternative like Octave is an old story already. But I feel it would be inadequate to stick with rigid values only because of one's preference for a single particular software development model. To me, one is just as good as the other. And I opine with certain authority being a TA at the university who led the migration from Matlab to Scilab for the practise sessions of Discrete-Time Signal Processing (a graphical interface for simulating dynamic systems was required, therefore xcos was needed), which is part of the Master's degree programme in Telecommunications Engineering. Read more

Spelling correction and the death of words

Fri, 23 Mar 2012 23:00:00 +0000

One of the topics treated in this second week of the Natural Language Processing class at Coursera is spelling correction (also treated in the Artificial Intelligence class). It's wonderful to have tools that help proofreading manuscripts, but this comes at the expense of impoverishing our own expression ability. This newspaper article, which links to the original research work conducted by Alexander Petersen, Joel Tenenbaum, Shlomo Havlin and Eugene Stanley, states that spelling correction (not only computerised but also human-made in the editorial industry) causes language to be homogenised, and this eventually reduces the lexicon (old words die at a faster rate than new words are created). So, is this NLP fancy topic actually hurting NLP? What a headache... Read more

Hacking with Multinomial Naive Bayes

Wed, 29 Feb 2012 23:20:00 +0000

Today it's the most significant day of a leap year, and I won't miss the chance to blog a little. I think I can put Udacity aside for a moment to note the importance of Naive Bayes in the hacker world. Regardless of its naive assumption of feature independence, which does not hold for text data due to the grammatical structure of language, the classification decisions (based on Bayes decision rule) of this oversimplified model are surprisingly good. Read more

Natural Language Processing: forthcoming online classes at Stanford, and a unified approach with Multilayer Neural Networks

Sun, 22 Jan 2012 19:30:00 +0000

Needless to say, I was very eager to begin the nlp-class tomorrow. It's a pity that it has just been delayed for administrative reasons. But good things take time, so we'd better be patient 'cos this looks very good (45000 people expecting its launch, this is incredibly exciting). Having used Manning's and Jurafsky's masterpieces for years, it is now enthralling to have the chance to attend their classes, getting my hands dirty with the classical problems, techniques and the whole lot of details of statistical NLP. Read more

L-systems, fractal plants and grammars

Sat, 24 Dec 2011 19:30:00 +0000

One beautiful example of life is vegetation. In fact, since I was little I had always believed that ferns were supposed to be the first form of life on Earth, but now I see that the Wikipedia debunks this fact here. Anyhow, plants do deserve a lot of admiration and respect, they are fascinating to me. Read more

Principal Component Analysis for Blind Source Separation? Shouldn't you use Independent Component Analysis instead?

Fri, 2 Dec 2011 21:00:00 +0000

This week's work in the Machine Learning class has treated the Principal Component Analysis (PCA) procedure to reduce the dimensionality of the feature space. The use of the Singular Value Decomposition (SVD) of the covariance matrix of the features has reminded me of the introductory video to unsupervised learning where Prof. Ng applies the SVD to blindly separate speech sound sources (e.g., the cocktail party problem). And I have felt puzzled because as far as I know, the problem of Blind Source Separation (BSS) boils down to finding a linear representation in which the components are statistically independent, uncorrelatedness is not enough (Hyvarinen, et al., 2001). Read more

Support Vector Machines and the importance of regularisation to approach the optimal boundary

Sun, 27 Nov 2011 23:00:00 +0000

In this post I would like to show some empirical evidence about the importance of model regularisation in order to control the complexity so as to not overfit the data and generalise better. To this end, I make use of this week's work on Support Vector Machines (SVM) in the Machine Learning class. Read more

On figuring out the similarity between the Mean Squared Error and the Negative Corpus LogLikelihood in the cost function for optimising Multilayer Neural Networks

Wed, 23 Nov 2011 23:30:00 +0000

This long post title links with my previous Post 59, where I discussed using different criteria for the cost function and for the derivation of the gradient when training Multilayer Neural Networks (MNN) with Backpropagation. After a fruitful chat with Prof. Xavier Vilasis and some careful reading (Bishop, 2006) and (Hastie, et al., 2003), I have eventually come to realise there is no mystery, as long as the right assumptions are made. Read more

Different cost criteria in Multilayer Neural Network training

Sat, 19 Nov 2011 23:30:00 +0000

It is customary to train Multilayer Neural Networks (MNN) with the Mean Squared Error (MSE) criterion for the cost function (Duda, et al., 2001), especially when the Backpropagation algorithm is used, which is presented as a natural extension of the Least Mean Square algorithm for linear systems, a deal of lexical coincidences with "mean square" altogether. Nevertheless, Prof. Ng in the ml-class presented a somewhat different flavour of cost function for training MNN, recurring to the "un-log-likelihood" error, i.e., the negative of the corpus loglikelihood, that typically characterises the Logistic Regression error wrt the data, and for a good reason: Read more

Inter-language evaluation of sentiment analysis with granular semantic term expansion

Thu, 17 Nov 2011 23:00:00 +0000

Yesterday, my Master Student Isaac Lozano defended his thesis with honours. The topic of his work entitles this post: inter-language evaluation of sentiment analysis with granular semantic term expansion. His work is mainly framed by our former publication (Trilla et al., 2010), but he has extended it by porting EmoLib to Spanish in its entirety and evaluating its performance wrt English, which is the default working language, and also the most supported one by the research community. Read more

Spanish Government presidential candidates face-off analytics

Tue, 15 Nov 2011 8:00:00 +0000

This post is in line with a recent article that presents a novel application to signal the polarity of opinions in Twitter posts, see this piece of news published by Universidad de Alicante. The original work is centred on the face-off debate between Rajoy and Rubalcaba, the two main candidates for the Spanish Government presidential election that will take place next Sunday. I here reproduce a similar experiment with the default configuration of EmoLib, which is grounded on a centroid-based classifier representing the words' emotional dimensions in the circumplex, but with the transcription of the whole debate, which is available for the two candidates: Rajoy, on behalf of the PP, and Rubalcaba, on behalf of the PSOE. What follows is a table that shows some linguistic statistics regarding their speeches (bearing in mind that they spoke the same amount of time) Read more

Multi-class Logistic Regression in Machine Learning (revisited)

Thu, 10 Nov 2011 20:00:00 +0000

In my former post about Multi-class Logistic Regression in ML (see Post 55), I questioned the ml-class about using a multi-class generalisation strategy, i.e., the One-Versus-All (OVA), with a plain dichotomic Logistic Regression classifier when there's a more concise multi-class model, i.e., the Multinomial Logistic Regression (MLR), that is already of extensive use in ML. Read more

Multi-class Logistic Regression in Machine Learning

Sun, 6 Nov 2011 22:00:00 +0000

This week's assignment of the ml-class deals with multi-class classification with Logistic Regression (LR) and Neural Networks. In this post I would like to focus on the former method, though, in line with Post 52. There I missed the use of the Multinomial LR (MLR) to tackle multi-class problems, putting in question the need of a multi-category generalisation strategy, i.e., One-Versus-All (OVA), when there is already a model that inherently integrates this multi-class issue, i.e., MLR. Now, after conducting some experimentation (see the table below), I shall conclude that it must be due to its higher effectiveness, which is measured wrt the accuracy rate, at least for the proposed digit-recognition problem. Read more

VIm, 20 years old

Fri, 4 Nov 2011 8:00:00 +0000

I would like to cite the 22nd pragmatic programming tip (Hunt and Thomas, 1999), which says: "Use a single editor well. The editor should be an extension of your hand; make sure your editor is configurable, extensible and programmable" because VIm just turned 20! It's wonderful news! VIm has been my choice for extending my hand. Over time, the HJKL keys have become my day-to-day user interface, and I'm really satisfied, despite I sometimes mess up emails when editing on a web-based client :-) Regardless that some say that VI-VI-VI is the Editor of the Beast, I believe it's a magnificent piece of art what Bram Moolenaar released 20 years ago, and I definitely stick with the beast. I wonder if there's also a myth about it being coded in one weekend, like Bill Joy's best-known deed... Read more

Natural Language Generator (NLG) released

Tue, 1 Nov 2011 23:30:00 +0000

I recall the tutorial that Noah Smith gave at LxMLS about Sequence Models with much interest. Motivated and inspired by his explanation, today I release a Natural Language Generator (NLG) based on n-gram Language Models (see the CODE section). Read more

Logistic Regression in Machine Learning

Sun, 30 Oct 2011 23:30:00 +0000

This week's work in the Stanford ml-class has covered Logistic Regression (LR) and Regularisation in regression models (both linear and logistic). I do believe Prof. Ng when he states that LR is the most widely used classification method in the world. A well-designed LR strategy is determined to pass any test with flying colours, hence its renown effectiveness. LR is full of wonders, most of which are grasped in the lecture, but I would still like to review a few aspects that I think deserve a deeper analysis. Read more

Linear regression in Machine Learning

Sun, 16 Oct 2011 23:30:00 +0000

Today is the deadline for submitting the first Review Questions of Stanford's Machine Learning course, which regard Linear Regression (LR) with one variable and Linear Algebra. I enrolled late this week and my experience with it has been very exciting. Andrew Ng has an outstanding ability to teach, and the lectures are most understandable with the digital board. But I'm a bit confused wrt the intended audience, because the gradient descent algorithm is presented along with the basic matrix operations (in a sense, the former is taught at 4th year of engineering while the latter at 1st year). Read more

Hyper{space,plane}

Sat, 8 Oct 2011 23:00:00 +0000

Star Wars, Machine Learning and Linux: an Eternal Golden Braid (and yes, I'm also fond of Douglas R. Hofstadter's book). This is a comic strip. Read more

LxMLS retrospective

Thu, 28 Jul 2011 16:00:00 +0000

The 1st Lisbon Machine Learning School LxMLS was warmly received and ended this week in complete success. The summer school has had many engaging qualities, not the least of which has been its magnificent set-up, just check out its schedule. Over 200+ applications, 130 candidates, most of whom were PhD students, enjoyed such a comprehensive program oriented to statistical natural language processing. All the materials, including the slides, lab docs and code (in Python), are available at L2F. Although the banner of the school shows a skew toward the web, I guess it's only a matter of adding a little NLTK clean_html to be just right :-)

Textual media is missed in multimedia

Wed, 15 Jul 2011 23:30:00 +0000

This week, La Salle has hosted the IEEE International Conference on Multimedia and Expo (ICME 2011). As a volunteer I can state that it has been a great experience being involved in its development. Especially if one is somewhat interested in image and audio processing. Nevertheless, one thing I generally miss in multimedia is its dismissal of the textual media. Despite the insistent reference to text in other fields when regarding multimedia (such as Dan Schonfeld and Li Deng, guest editors of the IEEE Signal Processing Magazine, vol. 28, n. 4, July 2011), there have been very few accepted works in the conference that have considered some textual issue. Although text is well available from images (OCR) as well as from speech (transcription via ASR), it is often ignored as a value-added feature. I even seldom hear that text is becoming sort of old-fashioned among the trendy media. But IMO, some of the most difficult problems in human interaction still stem in the machine comprehension of natural language (see The Loebner Prize in Artificial Intelligence), and until an acceptable rate may be achieved in this regard, text and hence Natural Language Processing are determined to be present in original research. This is my silly rant about the conference, which I deeply regret because overall it has been tremendous.

X JPL

Wed, 6 Jul 2011 19:30:00 +0000

Past Friday, July the 1st, I could enjoy the last day of the 10th Free Software Symposium (Jornades de Programari Lliure, JPL, in Catalan), which was held at UPC, Barcelona. In the morning, Prof. Arcadi Oliveres and Prof. Joan Tugores gave an overview of the economic and social impact of new technologies, trying to establish links with the free and proprietary software models. Prof. Oliveres overtly discussed that the lack of information that arrives from the non-independent mass media makes people live a lie. In the same line, Prof. Tugores generalised that the more uneven is a society, the fewer are the business/economic groups that share the power to make (big scale) decisions, and the more easily influenced they are to private (and sometimes dishonest) interests. That reminded me of the Builderberg Group altogether, and Uncle Ben with the often-quoted Spider-Man theme of "with great power comes great responsibility". Read more

Robotics at La Salle with Prof. Chris Rogers

Tue, 28 Jun 2011 19:30:00 +0000

Professor Chris Rogers, from Tufts University, gave a speech promoting creativity and innovation with pre-college engineering education, a kind of mirror talk wrt his probably more renown Talk at Google entitled "LEGO Engineering: From Kindergarten to College". Although I've traditionally been more keen on using Meccano (it's more heavy metal), I must admit that LEGO is astonishing (it's lighter and it doesn't oxidise), especially regarding the more "technical track" available with the use of the NXT microcontroller. Truly enticing. Read more

Free software needs free speech

Thu, 7 Apr 2011 16:00:00 +0000

I just wanted to echo this recent encouraging post to contribute to speech-related free software apps. It discusses the benefits of contributing to such FLOSS applications, which not only need good coders, but also big amounts of recorded speech data to train the acoustic models. The recordings requested a while ago (see Post 36) for Spanish are still being processed and no results are yet available.

Pipeline Skeleton released

Wed, 2 Feb 2011 17:00:00 +0000

Regarding Bob Carpenter's ready-to-distribute pyhi package, a bare bones Python package with all the trimmings (modular structure, configurability, building automation, etc.), and his apparent latter skew to C/C++, I thought it would be interesting and useful to have a similar package in C++. Such framework could be based on a sequential processing structure, which modules could be defined (and redefined) in an external XML config file, and its core implementation could be abstract with regard to concrete application needs (declaring pure virtual functions), thus defining a neat interface ready to be extended for any particular purpose. So, I've just released the Pipeline Skeleton (see the CODE section of my homepage). Read more

Full-text search ability in the blog posts

Tue, 21 Dec 2010 23:30:00 +0000

As the number of posts in the blog increases from time to time, I have thought it would be a good idea to enable a full-text search option there. Hence, thematic posts regarding a search query may be retrieved within a few moments. Read more

Information Retrieval techniques in ASR

Sun, 28 Nov 2010 13:00:00 +0000

I also wanted to blog about Dr. Alex Acero's speech in the FALA 2010 conference. His talk was entitled "New Machine Learning approaches to Speech Recognition", and in brief (quoting his own description), he described some new approaches to Automatic Speech Recognition (ASR) that leverage large amounts of data using techniques from Information Retrieval (IR) and Machine Learning. Read more

TTS in the future

Sun, 21 Nov 2010 00:30:00 +0000

In the FALA 2010 conference, Dr. Heiga Zen gave a speech entitled "Fundamentals and recent advances in HMM-based speech synthesis". He reviewed the growth of Hidden Markov Models (HMM) over the last years in the TTS research community. Indeed, this direction was also evident in the Speech Synthesis Albayzin 2010 Evaluation, where out of the 10 systems participating, 3 were purely concatenative, 6 were based on HMM, and one was as a hybrid approach (HMM-based + concatenative). And it was the latter who won the competition. Read more

FALA2010 contributions released

Wed, 10 Nov 2010 23:30:00 +0000

Today the FALA2010 conference has begun, and our pending papers are now available in the publications section. NLP has been regarded to be one of the most attractive fields in TTS research nowadays, at least according to Heiga Zen, who has given a tutorial session on HMM-TTS synthesis this morning. I hope to report it asap. Read more

Reasons for students to contribute to Open Source

Mon, 4 Oct 2010 15:30:00 +0000

I just wanted to echo Shalin's arguments to contribute to Open Source. In brief, Shalin supports having the chance to work on what one really likes, how beneficial this is for learning tools to face "real world" software problems (non-existent in academic problems), the experience gained from working with some of the best coders, and the attractiveness to companies that this profile yields. I do encourage students to enrol in open source projects for their Bachelor and Master's Thesis Projects. In the end, it is most probable that they get to use some open source tools to supply particular parts, and concentrate on their academic interests. So, why not contribute?

Discrete-Time Signal Processing with Scilab

Thu, 16 Sep 2010 14:00:00 +0000

For years, Matlab has been has been the de facto choice for many tasks at the university, including teaching. While it is reputed to be a technically fabulous tool, IMO it still lacks the free software flavour that open knowledge should have. We as engineers should strive to be able to analyse a system in its entirety. In this sense, Scilab is a great alternative (I already used it with great success for my Master's Thesis). By integrating Scilab into the academic life, students are enabled to gain a deeper knowledge of the system they are to work with, they are freed from the cost of any proprietary tool and thus they are dissuaded from the infringement of the law (let's face reality). Moreover, Scilab enables them working from home as it can be freely installed on any computer. So, there is no longer the need to come to the laboratories of the university, a fact that is most practical for online students. Read more

FALA 2010 conference

Wed, 15 Sep 2010 16:30:00 +0000

Good news. Our work on text classification of domain-styled text and sentiment-styled text for expressive speech synthesis has been selected for presentation at the FALA 2010 conference. The conference will be held on November at Vigo, Spain. Read more

Donate your voice for the wealth of free speech recognition apps

Tue, 7 Sep 2010 10:30:00 +0000

A week ago, the Fernando de los Rios Consortium, who maintains the Guadalinfo portal, launched the Donate your voice competition with the aim of compiling a voice corpus to produce a free acoustic model. This free acoustic model would then be used to develop a speech recognition application to control the desktop of a computer, similar to Magnus, but more professional :) Read more

Sentiment analysis with NLTK

Fri, 13 Jul 2010 22:00:00 +0000

By the beginning of the month, Streamhacker set a demo of sentiment analysis using the Natural Language Toolkit NLTK, a powerful Python set of open source modules for research and development in natural language processing and text analytics. Read more

EmoLib demo - Now available

Thu, 8 Jul 2010 12:00:00 +0000

EmoLib is shown to the world in the form of a web service with servlet technology, available here. Its performance features will (hopefully) be enhanced as my research advances. Read more

Dissertation on sentiment analysis - Now available

Thu, 1 Jul 2010 10:00:00 +0000

I've just made available the supervised work of my dissertation on sentiment analysis under a Creative Commons license. I've left out the doctorate courses I attended and some padding "fashion" stuff. Nevertheless, the whole dissertation is available at the library of the university or on (polite) demand :) Read more

Dissertation on sentiment analysis

Wed, 9 Jun 2010 14:00:00 +0000

Yesterday I defended my dissertation on sentiment analysis, actually entitled "Natural Language Processing techniques applied to speech technologies". The gist of this research is the automatic extraction of affective information from text in order to feed a speech enabled application with this particular human feature. Read more

Robots taking control of La Salle

Mon, 23 Apr 2010 8:00:00 +0000

First of all, let me wish all the Georges have a nice day. Then, I just wanted to mention the robotic platform that La Salle presented yesterday as a complementary part of the Engineering (and Business) courses, the LS Maker (that is not this Light Saber Maker), see this video and this article. My devotion to robots was already shown in an older post. As one of its marketing features, students are enticed to implement a voice recognition system on it! That's awesome! I wish I had had one of these during my undergraduate years!

Swift GNU/Linux tools stay alive

Mon, 19 Apr 2010 17:00:00 +0000

Have a butcher's hook (Cockney English for "look") at this Linux Journal article about my favourite GNU/Linux distro setup: Debian with a most lightweight configuration. I have used it for the past four years, after trying several fashionable distros, and I definitely keep it for the same reasons stated in the article. Nevertheless, to my taste, XFCE has become too complex (and heavy) for me and my PC, and I still rather prefer using Blackbox with bbkeys for keyboard shortcuts. That is really fast on a modest computer!

Moving from Alpine to Roundcube webmail

Thu, 15 Apr 2010 23:00:00 +0000

If you are (like me) still stubborn using text-based interfaces and you feel pushed to using "newer" technologies, this post may be of your interest. Read more

Automating game development with text-based technologies

Thu, 15 Apr 2010 8:00:00 +0000

Since my last post was dedicated to relating speech technologies with game development, and yesterday I came across a similar issue but with text technologies in a Linked-in discussion, the so-called Text-to-Scene conversion, I thought it also deserved a blog entry. Read more

Speech technologies growing among open source game development

Wed, 4 Nov 2009 15:00:00 +0000

Speech technologies are on the rise among open source game development communities. For example, check the announcement of the BennuGD WIZ/PC contest. This suite recently enables the incorporation of a speech synthesis engine in order to permit the interaction of speaking avatars. Game technologies grow mature in time, they need to provide more demanding features and speech technologies seem to be one the most challenging engineering sciences to supply this requirements. Will emotion analysis ever fit in this field? Very likely.

Emotion detection using both acoustic and linguistic information in children's speech

Mon, 27 Oct 2009 20:00:00 +0000

Dr. Michael Wagner, director of the National Center for Biometric Studies at the University of Canberra, Australia, has given a speech on emotion analysis at UPC. He has presented their contribution to the Emotion Challenge that was held last September at Interspeech. They proposed the fusion of acoustic and linguistic features in order to detect emotion states in children's speech.

Text Analysis in Writing Research

Mon, 19 Oct 2009 20:00:00 +0000

Today Dr. Nancy Nelson, from the University of North Texas, has given a speech on Text Analysis in Writing Research at Blanquerna. She has tackled the problem very philosophically, aside from the engineer's eye view of text mining, with an insight into the cultural and historical situations and how these factors influence orientation towards writing. Very interesting.

Spanish science doesn't need scissors

Wed, 7 Oct 2009 8:00:00 +0000

Still with the generalized indignation feeling because of the R&D budget cut proposed by the Spanish Government, today lots of Spanish researchers have agreed to post about this misfortune following the initiative "La ciencia espanola no necesita tijeras".

Talking RSS Reader for Android

Mon, 5 Oct 2009 11:00:00 +0000

First, Google is the gold sponsor of Interspeech'09. Then, it releases a speech synthesizer for Google's mobile platform, Android, the Talking RSS Reader. What will be next? It seems that as long as Google is interested in speech technologies, speech processing will be on the rise. Is it good news or not?

Ridiculous Spanish Government anti-crisis measures threaten R&D budgets

Wed, 30 Sep 2009 15:00:00 +0000

A few days ago this opinion column appeared in a Spanish newspaper. Its author, Dr. Rodriguez, criticizes the budget cut on R&D that the Spanish Government has proposed to fight the present crisis. Since the column is written in Spanish, I am just going to translate it in this post, please excuse my skewed Spanish constructions.

InterSpeech 2009 Opening Ceremony and Sentiment Classification from Text

Mon, 7 Sep 2009 19:30:00 +0000

The InterSpeech 2009 conference Opening Ceremony has been held today. Some of the most remarkable scientists that have passed away last year have been rememebered at the beginning of the event. Among them, Dr. Gunnar Fant, who devoted his life to the study of the vocal tract and the measurement of formant values.

Tutorials Day at InterSpeech 2009

Sun, 6 Sep 2009 19:30:00 +0000

The tutorials before the InterSpeech 2009 conference officially begins (tomorrow) have been held today. In the morning I have attended the "Language and Dialect Recognition" tutorial. This topic wraps the first step on a speech processing pipeline, before a sheer speech recognition engine. It first identifies the language of the speaker and then it spots its dialect. This step permits the use of more refined models for recognition.

Long live the robots

Wed, 29 Jul 2009 19:30:00 +0000

One more year, the CampusBot gathers a huge amount of robotics enthusiasts in Valencia, as part of the CampusParty. I had also once made my first approach to this field, it was my final High School project: WaiterBot. My robot was remotely controlled with a home-made transmitter/receiver pair with a joystick and could carry a glass of e.g. water on a tray without spilling its content. Since it had a couple of orthogonal encoders parallel to its plane of movement it could detect the inclination of the surface it roved and correct this angle with a set of motors and gears. Take a look at the pictures to see what I mean.

R, Octave and Scilab

Fri, 19 Jun 2009 23:30:00 +0000

Today the program for the next Jornades de Programari Lliure has been made available on the web portal of this meeting. Although the program is still provisional, the appointed activities schedule a speech on R and a tutorial on Octave and Scilab. Read more

InterSpeech 2009 conference

Tue, 16 Jun 2009 16:30:00 +0000

Good news. Our work on sentiment classification from sentence-level annotations of emotions regarding models of affect has been selected for poster presentation at the InterSpeech 2009 conference in a session "Prosody, Text Analysis, and Multilingual Models". The conference will be held on September at Brighton, U.K.

HMMs considering multiple observations

Sat, 6 Jun 2009 22:30:00 +0000

The original definition of the (discrete) Hidden Markov Models (HMM) in the speech recognition field [Rabiner, 1989], deals with a single symbol sequence, which implies that the models are shown one single aspect of the speech signal (the spectral behavior features). Thus the models are limited to making predictions based on the features of a speech frame, which is previously framed in time in order to take it for a stationary signal. But speech is by definition a non-stationary time-varying signal (phonemes...), therefore the stationarity assumption is not actually true, but we assume it for simplification. Read more

Thank you, Guadalinex

Tue, 19 May 2009 18:30:00 +0000

The Guadalinex GNU/Linux distribution is one of the most ambitious free software projects in Spain, promoted by the Government of Andalusia (Junta de Andalucia), aimed at spreading the FLOSS culture in the educational sector. This exemplar project deserves my admiration for the technical quality of the distro, the organization of the project and the nice community that supports it. Read more

Ideas, values and attitudes

Sat, 16 May 2009 11:30:00 +0000

With this title, the Most Honorable President of the Government of Catalonia (Generalitat de Catalunya) Dr. Jordi Pujol presented his speech at the VIII Student Conference at Blanquerna with the main topic of "Citizenship and Communication". Communication is a feature that we bloggers, and especially researchers, should develop proficiently, because regardless of how good a work can be it can unfortunately be ignored or even invalidated if the ideas that conveys don't get through. Read more

Working on the sentiment of text

Fri, 1 May 2009 11:30:00 +0000

It's been quite a long time (almost 3 months) since I last blogged. I have been working with Dr. Alias on an article for this year's SEPLN conference San Sebastian 2009. Our work deals with positive/negative text classification from sentence-level annotations of emotions regarding models of affect. Read more

Advances in parallel frameworks

Thu, 5 Feb 2009 22:30:00 +0000

Some days ago Xavier Llora published a technical report about a pilot study of a data-intensive framework running genetic algorithms using Meandre. What's most impressive is the capacity of Meandre to parallelize the computation bottlenecks dividing the load among the cores of the processor that runs the framework. Read more

Back to the genuine Debian flavor

Wed, 28 Jan 2009 22:00:00 +0000

After some time "playing" with Puppeee, some application-level features (no SSL support for Subversion, no LaTeX) didn't satisfy my needs. I suspect it is my business to mind for such specific software. Thus, back to the genuine Debian flavor, I took eeeXubuntu, a 7.10 live CD release of Xubuntu hacked to run on the EeePC. Read more

Don't be evil, Google

Thu, 8 Jan 2009 8:00:00 +0000

Reading Gottiplati's opinion on ONJava, available here, I realize of the critical situation that Java developers are facing presently since the birth of Dalvik, Google's (Java) Virtual Machine (VM) for the Android mobile phone platform. Read more

Puppy Linux flavored SDHC for the EeePC

Sat, 3 Jan 2009 12:00:00 +0000

Not being much content with the little hacking that the EeePC Xandros distro offers I've decided to bring a new software set to my new device, the ASUS EeePC 900. Since the DebianEeePC wouldn't read the 8GB SDHC card I bought for lodging the new OS, I thought I'd give a try with a different GNU/Linux flavor. Read more

ASR in the Asus EeePC

Fri, 26 Dec 2008 23:00:00 +0000

Having fun with the present that Santa Claus brought for me, the UMPC Asus EeePC 900, I am now enoying the device bringing it to one of my fields of study, Automatic Speech Recognition (ASR). The computer has an embedded speech recognition engine that is used to recognize the user's voice commands to launch some applications. It has a modus operandi similar to Magnus, my Master's Thesis: the input speech is analysed continuously in order to determine the spoken commands.

Prof. David E. Goldberg at La Salle #2

Thu, 11 Dec 2008 23:00:00 +0000

Today, Prof. Goldberg has pronounced his second speech, this time, about the matter he gave birth to 28 years ago: Genetic Algorithms (GAs). And there's not much to say about it, it has been simply brilliant. The man that yesterday argued in favor of the approach to monopolies for the sake of technology, a mind I could never share due to my compromise with the free software world, has today proven to be a master in the field of GAs, whatever the tools of use. I expect to be getting my hands dirty with GAs shortly. I'll be feeding this blog with the attempt as soon as I do.

Prof. David E. Goldberg at La Salle #1

Wed, 10 Dec 2008 23:00:00 +0000

Today, professor David E. Goldberg, a worldwide renowned American computer scientist, father of genetic algorithms, has given the first of the two speeches arranged for this week at La Salle, titled "The Creativity Imperative and the Technology Professional of the Future", a business-centered perspective of engineering. Read more

Getting my VPN to work, at length

Fri, 5 Dec 2008 23:00:00 +0000

Getting a Virtual Private Network (VPN) to work under GNU/Linux can be a pain. I have had, like many other people, a lot of trouble to set it right. Some time ago I did have my attempts, I eventually got to establish the connection but the traffic wouldn't route with success. Read more

FreeLing exposed

Wed, 19 Nov 2008 12:00:00 +0000

Last week, on Friday the 14th. the presentation of FreeLing at Universitat Pompeu Fabra was a total success. I think that Dr. Padro, the speaker and one of the developers of FreeLing, showed a tool that is a must to be known by anybody related to the NLP field. Read more

Speech technologies wiki

Thu, 23 Oct 2008 12:20:00 +0000

Following the ideal of free information and responding to the need of having a good means of documentation, a wiki specialised on speech technologies has been created. My Ph.D. Thesis advisor, Dr. Francesc Alias, recommended me to jot down in a safe place anything important I learned and that I would probably need sometime. Read more

NLP disambiguation

Thu, 16 Oct 2008 12:20:00 +0000

This topic aroused my curiosity in my yesterday's attendance at the scientific research methodologies class, conducted by Dr. Mundet, when he suggested that the clear advantage of Barack Obama over John McCain for the Presidential Election 2008 was due, in part, to the special speech techniques that Obama uses. It's none of my business to discuss the political issues of the two candidates, but to analyse the scientific aspect of such fact, if there is some. Read more

Hello World!

Wen, 15 Oct 2008 13:00:00 +0000

This is my first post, and with it, this blog is born. My name is Alexandre Trilla and I am a researcher at the Department of Media Technologies (DTM) at Enginyeria i Arquitectura La Salle, Universitat Ramon Llull, Barcelona. Read more