A very useful site for programmers is the Quick Reference Site, maintained by Tim Sinaeve. This site features:
Cual es mi sorpresa al consultar El Mundo, ver que la portada de la sección de televisión presenta como noticia principal la increible actuación de José Manuel Lúcia en Pasapalabra. Es digno de contemplarse.
Tras esta impresionate hazaña, que le ha valido a Jose Manuel la nada despreciable cifra de 396.000 euros de premio, viene a mi mente mi post de ayer en el que IBM se plantea como próximo reto de la Inteligencia Artificial el construir un sistema capaz de competir a nivel humano en el programa Jeopardy!
El programa Jeopardy! es un popular concurso norteamericano de preguntas de memoria, que exige no sólo una gran capacidad de almacenamiento (en el estándar humano), sino también una enorme rapidez de recuperación de datos. IBM plantea competir con su tecnología de respuesta a preguntas (Question Answering) llamada Watson. Aunque para un profano puede parecer fácil, el campo de QA es uno de los más activos por su interés práctico pero también por su dificultad.
¿Sería Pasapalabra un reto de igual magnitud para un sistema de Procesamiento de Lenguaje Natural? Indudablemente, no en mi opinión. Si en Jeopardý la dificulta consiste en interpretar la pregunta, en Pasapalabra nos remitimos a una interpretación básica (detectar "Contiene la X" vs. "Con la C"), y luego recuperar una palabra dada su definición. Esta segunda parte se puede hacer por técnicas de Recuperación de Información básicas sobre un diccionario exhaustivo previamente indexado. ¿O no?
Para la reflexión, ¿qué ocurre con las categorías gramaticales? Tomemos esta definición improvisada: "Con la V: Ejercicio del derecho de elección de los representantes políticos en los órganismos gubernamentales". ¿La respuesta es "Votar" o "Voto"?
Otra aplicación del Procesamiento del Lenguaje Natural relacionada con Pasapalabra es la de elección de las palabras que conforman el "rosco". El camino trivial es la elección aleatoria, pero la dificultad del rosco sería muy variable y a veces escandalosamente excesiva. La solución es catalogar las palabras por su dificultad. ¿Y como calcularla? Opino que la aproximación más razonable es usar un corpus representativo del español, y asignar a las palabras de un diccionario una dificultad según su frecuencia o popularidad, con atención a la lematización. La variación morfosintáctica debe contemplarse para, por ejemplo, reducir la flexión verbal al infinitivo que aparece en un diccionario. El detalle de seleccionar las palabras de manera que mantengan un nivel de dificultad que equilibre el rosco es un problema de optimización que se puede resolver con algorítmica habitual (entiendo que en este caso, podría valer una aproximación de avance rápido (greedy) o a lo sumo de programación dinámica.
Para finalizar, felicitar a José Manuel por su hazaña, y agraderle su inspiración para este post y el magnífico momento que me ha hecho pasar contemplando el vídeo.
AIML se ha convertido casi en un estándar para el desarrollo de chat bots, es decir, programas capaces de conversar con la gente como personas reales (bueno, ese es el objetivo, pero apenas pasan el Test de Turing, aunque pueden engañar incluso un experto por un tiempo).
Estoy en el proceso de construcción de un robot basado en AIML se asemeja a un niño de habla hispana.He probado dos enfoques principales, a saber:
Pablo Gervásme sugirió, para evitar empezar de cero, usar un robot de habla inglesa.Bueno, si consigo encontrarlo!Así que mi pregunta es la siguiente:
¿Alguna vez has encontrado un bot de habla inglesa parecido a un niño, construido sobre AIML?
Referencias
Abu Shawar, Bayan; Atwell, Eric. A chatbot system as a tool to animate a corpus. ICAME Journal, vol. 29, pp. 5-24. 2005.
Abu Shawar, Bayan; Atwell, Eric. Using corpora in machine-learning chatbot systems. International Journal of Corpus Linguistics, vol. 10, pp. 489-516. 2005.
Question Answering is an increasingly popular Natural Language Processing task, defined in the wikipedia as "the task of automatically answering a question posed in natural language. To find the answer to a question, a QA computer program may use either a pre-structured database or a collection of natural language documents (a text corpus such as the World Wide Web or some local collection)." In other words, an Information Retrieval system / Search Engine does not return a list of references or documents, but it actually is able to cleverly process the question, digg into the documents, and get an appropriate answer.
Question Answering is a very active research topic among NLPers. There are competitions like the TREC and CLEF tracks on Q&A, and some systems have been deployed to real applications (from within Search Engines like Ask.com to domain-specific systems like
EAGLi for biomedicine). But, now, IBM gains popularity for their system Watson with aiming at a nearly Turing Test: beating humans at Jeopardy!
Jeopardy is a game demanding knowledge and quick recall, covering a broad range of topics, such as history, literature, politics, film, pop culture, and science. IBM believes that this is the final challenge for their system Watson, one of the leaders on this area. Do you remember Deep Blue? Ok, this time is not to beat a person in terms of specific skills (only), like reasoning and hypotetizing, but also in dialog! This is very close to a real Turing Test!
If Watson is able to beat the winner of Jeopardy, will we be closer to a real AI? Are HAL or Skynet waiting for us behind that door?
AIML has been turn into nearly an standard for writing chat bot, that is, programs able to chat with people like persons (well, that is the goal, but they hardly pass the Turing Test, although they may fool even an experft for a while).
I am in the process of building an AIML-based bot resembling an Spanish-speaking child. I have tested two main approaches, namely:
Pablo Gervás suggested me to avoid starting from scratch by using an English-speaking bot. Nice, if I can find it! So my question is:
Have you ever found an English-speaking chat bot resembling a child, based on AIML?
References
Abu Shawar, Bayan; Atwell, Eric. A chatbot system as a tool to animate a corpus. ICAME Journal, vol. 29, pp. 5-24. 2005.
Abu Shawar, Bayan; Atwell, Eric. Using corpora in machine-learning chatbot systems. International Journal of Corpus Linguistics, vol. 10, pp. 489-516. 2005.
Sixteenth International Symposium on String Processing and Information Retrieval
25-27 August 2009
Saariselka, Finland
SPIRE 2009 covers research in all aspects of string processing, information retrieval, computational biology, pattern matching, semi-structured data, and related applications. Typical topics of interest include (but are not limited to):
Deadline for papers: May, 1, 2009
Very, very interesting meeting, organized by the Antiphishing Working Group (APWG). The press release:
Electronic crime responders, investigators and counter-electronic crime technologists will join law enforcement and public policy officials from across the globe in Barcelona in May for the APWG's Counter-eCrime Operations Summit, uniting thought-leaders worldwide to plan the next stage in the global confrontation against electronic crime.
The third annual APWG operations conference (CeCOS III), to be held on May 12-14 in Barcelona, Spain will engage questions of operational challenges and the development of common resources for the first responders and forensic professionals who protect consumers and enterprises from electronic crime threats every day.
CeCOS III will present: informative case studies by electronic crime responders and security specialists; examinations of technologies developed and used by electronic crime gangs to exploit Internet infrastructure and user's PCs and client devices; discussions about the technologies and techniques of educating and protecting consumers; and presentations about the development of shared resources like common data formats for ecrime reporting, alerting and coordinating mechanisms.
APWG Chairman David Jevans said, "National governments, international treaty organizations, law enforcement agencies and industry associations the world over are looking at coordinating data exchange for electronic crime. CeCOS III will work to build bridges between these constituencies that engage the threats that electronic crime poses against consumers and enterprises everywhere every day."
CeCOS III is an open conference for members of the electronic-crime fighting community, hosted by the APWG and sponsored by LaCaixa, Telefonica, S21sec, GMV, MarkMonitor, EMC's RSA security division, Deloitte España and Ecija. Although sponsorship is principally from industry, the CeCOS programs are considered the most vital events to investigators and managers of electronic crime from across private and public sectors.
In Tokyo last year at CeCOS II, some 250 delegates attended from law enforcement agencies, technology companies, financial services firms, security services firms, government agencies, consumer advocacy groups and research centers around the globe, bringing together some of the most advanced counter-electronic crime thought leaders from East Asia, Europe, South America and North America.
Parties interested in proposing presentations or participating in panel discussion for CeCOS III can send email to: proposals@antiphishing.org. Parties interested in sponsoring some part of the event can contact Deputy-Secretary General Foy Shiver at fshiver@antiphishing.org.
A preliminary working agenda can be found at the APWG Web page for the Summit.
Final call for participation INEX 2009
INEX 2009 is examining focused (sub-document) retrieval using the new Wikipedia 2009 Collection. The collection size is over 50 Gigabyte with 2.5 million articles in semantically marked-up XML format. For the main ad hoc retrieval track the collection will be made available in Plain Text as well as XML so that non-XML based passage retrieval systems can also be used.
Other tracks are: Book Searching (using a collection of 50,000 full text books in XML), Efficiency, Entity Ranking, Interactive (iTrack), Question Answering (QA@INEX), Link-the-Wiki, and XML-Mining. All INEX Participants are expected to participate in topic creation, and assessment.
The schedule of the main ad hoc track is as follows:
Other tracks will follow variants of this schedule. Relevance assessments will be provided by the participating groups using the INEX assessment system. Each participating organization will judge about 3 topics. Please note that the assessment of each topic may take one-person 1 to 2 days to complete!
VideoCLEF 2009
Video Analysis and Retrieval Benchmark Evaluation
VideoCLEF 2009 is a track of the CLEF benchmark campaign dedicated to developing and evaluating tasks involving access to video content in a multilingual environment. In 2009, organizers offer four video analysis and retrieval tasks, which will be carried out on Dutch television documentaries. Participants can approach these tasks using their own choice of methods and features. The provided video data will include speech recognition transcripts, shot boundaries, shot-level keyframes and archival metadata.
VideoCLEF 2009 takes place on the following schedule:
Third Workshop on Analytics for Noisy Unstructured Text Data
23-24 July 2009, Barcelona, Spain
In conjunction with the Tenth International Conference on Document Analysis and Recognition
Noisy unstructured text data is ubiquitous in real-world communications. Text produced by processing signals intended for human use such as printed/handwritten documents, spontaneous speech, and camera-captured images, are prime examples. ICR/OCR error rates on paper documents can range widely from 2-3% for clean inputs to 50% or higher depending on the quality of the page image, the complexity of the layout, aspects of the typography, etc. Individual variability in handwriting make this a particularly difficult form of input and error rates here are often substantially higher than for machine print text. Telephonic conversations between call center agents and customers often see 30-40% word error rates, even using state-of-the-art ASR techniques. In spite of the tremendous challenges such data presents, it is pervasive in applications of interest to corporations and government organizations.
Recognition errors are not the sole source of noise; natural language and the creative ways that humans use it can create problems for computational techniques. Electronic text from the Internet (emails, message boards, newsgroups, blogs, wikis, chat logs and Web pages), contact centers (customer complaints, emails, call transcriptions, message summaries), and mobile phones (text messages) is often noisy, containing spelling errors, abbreviations, non-standard words, false starts, repetitions, missing punctuation, missing case information, and pause-filling words such as "um" and "uh" in the case of spoken conversations.
The Third Workshop on Analytics for Noisy Unstructured Text Data (AND-09) is devoted to issues arising from the need to contend with noisy inputs, the impact noise can have on downstream applications, and the demands it places on document analysis. Topics of Interest (but not limited to):
Dates
The goal of the Open Health Natural Language Processing Consortium is to establish an open source consortium to promote past and current development efforts and to encourage participation in advancing future efforts. The purpose of this consortium is to facilitate and encourage new annotator and pipeline development, exchange insights and collaborate on novel biomedical natural language processing systems and develop gold-standard corpora for development and testing. The Consortium promotes the open source UIMA framework and SDK as the basis for biomedical NLP systems. Applications created within UIMA consist of software components (referred to as annotators) and their associated configuration files and external resources. Within the framework, one can also create complete pipelines composed of a sequence of annotators and the data flow between them.
Via the BioNLP list.
In the list I am subscribed, there are announces of new data collections from time to time. These resources are extremely valuable for Bio-NLP research, we must disseminate them and strongly appreciate their builders work:
A number of tools are available at the Bio-NLP Resources page compiled by Martin Krallinger and his group.
The following conferences and jornals are needing reviewers. I think it is a very good idea to support these conferences and journals (it does not take so much time, it matters, and it makes CV).
SEMINARIO MAVIR
Miércoles 22 de abril 2009 en la UNED
Con motivo de la presencia de Satoshi Sekine y Andrew Borthwick en el congreso WWW2009 de Madrid, hemos organizado para el próximo miércoles 22 de abril a partir de las 16h dos seminarios que se celebrarán en la ETSI Industriales de la UNED.
TÍTULO: Recent Advances on Minimally Supervised Knowledge Discovery
PONENTE: Satoshi Sekine (NYU)
HORARIO: miércoles 22/04/2009 a las 16h00
TÍTULO: Challenges on Wep People Search: the Spock experience
PONENTE: Andrew Borthwick (Spock Networks)
HORARIO: miércoles 22/04/2009 a las 17h00
LUGAR DE CELEBRACIÓN: Salón de Grados ETSI Industriales, UNED
c/ Juan del Rosal, 12 28040 Madrid
SEMINARIO MAVIR
Jueves 23 de abril 2009 en la UNED
TÍTULO: Semantically Enhanced Information Retrieval: an Ontology-based Approach
PONENTE: Míriam Fernández (UAM)
HORARIO: jueves 23/04/2009 a las 12h00
LUGAR DE CELEBRACIÓN: Sala 2.24 (segunda planta) Facultad de Psicología, UNED
c/ Juan del Rosal, 10 Ciudad Universitaria 28040 Madrid
The papers for the forthcoming World Wide Web Conference 2009 in Madrid have been released. It is possible to access them at the ePrint archive set up for that purpose. I cannot forget this conference since the hit by Brin & Page at WWW98 with their PageRank and data structures that make Google.
Among published papers, there is a selection of those nominated for the best paper award:
My favourite is Visual Diversification of Image Search Results! But I cannot vote :-)
Human Computation Workshop (HCOMP 2009)
KDD-09 Workshop, Paris France
June 28, 2009
The organizers invite you to participate in the first annual Human Computation Workshop (HCOMP 2009), to be held on June 28th in conjunction with the KDD-09 conference in Paris, France.
Human computation is a new research area that studies the process of channeling the vast internet population to perform tasks or provide data towards solving difficult problems that no known efficient computer algorithms can yet solve. The goal of this half-day workshop is to bring together academic and industry researchers in a stimulating discussion of existing human computation applications (e.g. games, CAPTCHAs, Mechanical Turk) and future directions of this new subject area.
The organizers solicit papers related to various aspects of both general human computation techniques and specific applications, e.g. general design principles; implementation; cost-benefit analysis; theoretical approaches; privacy and security concerns; and incorporation of machine learning / artificial intelligence techniques.
An integral part of this workshop will be a demo session where participants can showcase their human computation applications. Detailed information about the workshop and submission procedures can be found at the website.
The workshop proceedings will be included in the ACM Digital Library. Deadline for submission is April 18, 2009 8pm Eastern Time.
Jin-Dong Kim, from the Tsujii Laboratory, University of Tokyo, has announced the start of the Parsed MEDLINE(R) data download service.
This service provides an access to syntactically-parsed MEDLINE abstracts. The abstracts were parsed with a wide-coverage HPSG parser, the Enju parser (version 2.2). The original data is the 2009 baseline release of the MEDLINE database, which includes approximately 18 million records.
For detail, refer to the usage web page.
Themos Kalafatis has an interesting blog about Practical Applications of Data Mining, Text Mining and Information Extraction entitled Life Analytics, in which there a number of examples on using Twitter for Sentiment Analysis, and a recent post about ScoutLabs, which is a company that offer Social Media Monitoring for corporations.
My opinion is that there is a great opportunity right now for Opinion Mining and Sentiment Analysis on Social Media, as collecting and classifying opinions and metions to brands, trade marks, services and products, executives and other staff, etc., can be very valuable for:
In KDNuggets, there is a list of Social Networking groups about Data Mining, Analytics, Web Mining, and Text Mining:
Third International Workshop on Data Mining and Audience Intelligence for Advertising (ADKDD'09)
June 28, 2009, Paris, France
Iconjunction with The 15th International Conference on Knowledge Discovery and Data Mining (SIGKDD'09)
Advertising, especially online advertising, is growing rapidly and brings about large volumes of data along with challenging data mining problems. Following on the success of ADKDD 2007 and 2008, ADKDD 2009 is to be held in Paris France, in conjunction with KDD 2009, to provide a high-level international forum for the academic community and the industry to present the state of the art of algorithms and applications of advertising.
Topics
Papers on all aspects of data mining and audience intelligence for advertising are solicited. Areas of interest include, but are not limited to:
Dates
XXV CONGRESO DE LA SEPLN
8-10 Septiembre 2009
Palacio Miramar, Donostia - San Sebastián
La XXV edición del Congreso Anual de la Sociedad Española para el Procesamiento del Lenguaje Natural (SEPLN) se celebrará en Donostia-San Sebastián los días 8, 9 y 10 de septiembre de 2009 en el Palacio de Miramar.
La ingente cantidad de información disponible en formato digital y en las distintas lenguas que hablamos hacen imprescindible disponer de sistemas que permitan acceder a esa enorme biblioteca que es Internet de manera cada vez más estructurada.
En este mismo escenario, hay un interés renovado por la solución de los problemas de accesibilidad a la información y de mejora de explotación de la misma en entornos multilingües. Muchas de las bases formales para abordar adecuadamente estas necesidades han sido y siguen siendo establecidas en el marco del procesamiento del lenguaje natural y de sus multiples vertientes: Extracción y recuperación de información, Sistemas de búsqueda de respuestas, Traducción automática, Análisis automático del contenido textual, Resumen automático, Generación textual, y Reconocimiento y síntesis de voz.
El objetivo principal del congreso es ofrecer un foro para presentar las últimas investigaciones y desarrollos en el ámbito de trabajo del Procesamiento del Lenguaje Natural (PLN) tanto a la comunidad científica como a las empresas del sector. También se pretende mostrar las posibilidades reales de aplicación y conocer nuevos proyectos I+D en este campo.
Además, como en anteriores ediciones, se desea identificar las futuras directrices de la investigación básica y de las aplicaciones previstas por los profesionales, con el fin de contrastarlas con las necesidades reales del mercado. Finalmente, el congreso pretende ser un marco propicio para introducir a otras personas interesadas en esta área de conocimiento.
Se anima a grupos, investigadores y empresas a enviar comunicaciones, resúmenes de proyectos o demostraciones en el ámbito de las tecnologías de la lengua en alguna de las áreas temáticas siguientes:
Fechas importantes
I am reviewer for this conference:
First International Conference on Networked Digital Technologies
Ostrava, Czech Republic, July 28-31, 2009
The proposed conference aims to enable researchers build connections between different digital applications. Currently a number of institutions across the countries are working to evolve better models to provide collaborative technology services for scholarship by creating shared cyberspace thro expert collaboration, but this is a challenge for the institutions for a number of reasons. In the last few years, the landscape of digital technology applications projects for the various disciplines in humanities, social sciences, and sciences appears induced by many initiatives. For the creation of research clusters, the research community has thousands of databases, websites, local computing clusters, and web-based tools around individual themes, interests and projects. In most cases, these tools and resources are and were created to meet the specific needs of a particular community. In many cases, the funding and support for these critical initiatives is fragile and temporary, and directed in piecemeal fashion. There is a need to provide concerted efforts in building federated digital technologies that will enable the formation of network of digital technologies.
Topics include but not limited to:
Important dates
Wandering on the opinion mining and sentiment analysis topic, I have discovered that Bing Liu, a referent on this topic, has posted his tutorial at the World Wide Web Conference 2008 at his home page. While the survey by Bo Pang and Lillian Lee for Foundations and Trends in Information Retrieval is a must, this stuff is (obviously) easier to read. Mi opinion :-D is first to read the tutorial slides, then the survey.
A good point with Bing Liu's stuff is that he has posted data at his Opinion Mining, Sentiment Analysis, and Opinion Spam Detection project. The data consists of (short) customer reviews of several products like cameras, routers, cell phones and so. This is an extract from the Linksys Router opinions:
router[+2]##This router does everything that it is supposed to do, so i dont really know how to talk that bad about it.
setup[+2], installation[+2] ##It was a very quick setup and installation, in fact the disc that it comes with pretty much makes sure you cant mess it up.
install[+3]##By no means do you have to be a tech junkie to be able to install it, just be able to put a CD in the computer and it tells you what to do.
Besides, Bing has worked on opinion spam!
As soon as I find the time, I will try to write a tutorial on how to train and test a baseline opinion classifier with WEKA.
Un nuevo artículo, en este caso sobre privacidad en Flickr, para Linux+:
Gómez Hidalgo, J.M. Privacidad en Flickr. Linux+ (ISSN: 1732-7121), Número 53, Abril, 2009.