Research Humanities

M. van de Camp, PhD student

http://www.tilburguniversity.edu/nl/webwijs/show/?uid=m.m.v.d.camp

This project aims at the development of a generic, reusable toolkit for historical text mining. Operating on any large heterogeneous set of documents from some arbitrary period but covering a particular historical domain, a “historical web of domain knowledge” is generated: an associative network of concepts (historically relevant entities) and their relations, linked to a timeline. This network can be browsed or searched with textually and/or visually oriented interfaces for timelines and maps, allowing for both system-internal component evaluation and system-external task-based evaluation.

multipleResearchers

Dr. K. Zervanou

http://www.tilburguniversity.edu/webwijs/show/?uid=k.zervanou

Dr. M. van Zaanen

http://www.tilburguniversity.edu/webwijs/show/&uid=m.m.vanzaanen?uid=m.m.vanzaanen

Many sources of data, such as natural language or music, have an inherent internal structure. This structure describes the regularities and restrictions of the data. This research aims at making the structure explicit, which leads to a description of the underlying (formal) language of the data. This information can then be used, for instance, to group similar elements in the data together. The research is divided into two parts. The first part aims at learning language models that describe the data and can be used to introduce explicit structure on top of the unstructured data. The second part aims at using the identified structure in specific applications. For instance, in the context of mood classification of (music) lyrics, language models for each of the different moods can be used to identify the underlying mood of a particular lyric. Similarly, language models can be applied within spelling correction systems that require sentential context to identify and correct errors.

Prof. dr. A. Plaat

http://www.tilburguniversity.edu/webwijs/show/&uid=584103?uid=584103

The central question this chair is concerned with is how we can model the behavior of people in organizations in order to better understand it, in order to increase the effectiveness of organizations. It is about people’s behavior, about organizations and about modeling. A model is an abstraction of reality. It is a scientific thought construct. The researcher leaves out unimportant elements to be able to concentrate on the hidden essence. With the advent of the computer the modeling possibilities have increased. Computer models make it fairly easy to imitate transitions from one state to another. We can incorporate certain actions into the model which the computer will subsequently perform. That way, interactions can be simulated. Well-known applications are weather simulations and simulations of the landing on the moon (Hartmann 1996). There are tremendous challenges in the field of organization simulation. There is as yet no realistic computer simulation of cooperative behavior in organizations. It requires pushing the boundaries of what is currently possible. This chair conducts research on modeling forms of cooperation and negotiation in order to reconstruct the interests behind them, to arrive at a better understanding of behavior, to fathom the consequences, to be able to imitate behavior in management training sessions, and to find out how we can get people to work together more effectively.

S.Wubben, PhD student

http://www.tilburguniversity.edu/webwijs/show/?uid=s.wubben

The aim of the MEMPHIX (memory based paraphrasing using implicit and explicit semantics) project is to develop a way to automatically generate paraphrases. The ability to paraphrase can serve to explain something or to provide feedback in dialogue. Generating shorter paraphrases is useful for subtitles or news feeds. Paraphrasing can also change the register of a text. It might also help to increase performance of question answering, dialogue systems and machine translation. In the MEMPHIX project, a system is built that learns to generate paraphrases on the basis of examples. This is done by treating paraphrasing as a monolingual machine translation task. While the generation of paraphrases can be driven in the first place by surface similarities (leaving semantics completely implicit), explicit semantic information may also play a role. Such information may be computed through automatic means (parsing, semantic role labeling, co-reference resolution). The project compares the direct implicit route with the use of explicitly computed semantics. The project has access to a Dutch corpus of over a million words developed in the DAESO project, consisting of pairs of texts that express paraphrased or at least comparable information from various domains. However, the data from DAESO alone might not be enough. The first step in this project has been to automatically collect more aligned paraphrases by mining the web: headline clusters are acquired from Google News. For each cluster, the available paraphrase candidates are selected using surface similarities. These aligned paraphrases can then be used to train an MT system.

A.P.C.I. Hong, PhD student

http://www.tilburguniversity.edu/webwijs/show/&uid=a.p.c.i.hong?uid=a.p.c.i.hong

Dealing adequately with conflicts of interest is pivotal to the management of organizations. Managers should be trained to recognize and to deal with strategic behavior in order to improve their effectiveness in managing an organization. The goal of the research is to improve training tools for managers, for instance, by developing serious games or computer simulations that provide insight into the strategic behavior and conflict resolution. The research focuses on the computational modeling of four characteristics of strategic behavior in organizations. The modeling of: (1) social relations between agents in an organization, (2) hidden agenda’s and strategic behavior in collectives of agents, (3) communication and social interactions, and (4) strategic behavior of agents in an organization.

PhD

Dr. Ir. P.H.M. Spronk

http://www.tilburguniversity.edu/webwijs/show/?uid=p.spronck

Prof.dr. A. Plaat (PhD supervisor)

http://www.tilburguniversity.edu/nl/webwijs/show/?uid=per.vanderwijst

Dr. P. van der Wijst (PhD co-supervisor)

http://www.tilburguniversity.edu/webwijs/show/&uid=a.plaat

R. Mattheij, PhD student

http://www.tilburguniversity.edu/webwijs/show/?uid=r.j.h.mattheij

A home networking system, sometimes called a domotic system, connects all household appliances and other electrical devices in a home to a common network, giving a global view and command. Currently home networking systems include computer home networking, networking of consumer electronics, networking of communication devices, the classical home automation, and recently smart energy meters. This project addresses the interaction of users with domotic systems. At the core of this projectis the belief that intelligent systems should stimulate users to adopt energy saving behaviour by means of persuasion, rather than by taking over control. Our aim is to persuade users to change their energy-consumption behaviour by developing personalized agents in the form of displays or robotic interfaces that provide feedback and suggestions in an attractive and non-disturbing way. The agents collect information on consumptionpatternsthrough,e.g., power-consumption meters, and use that information for generating feedback and suggestions. The main challenge of the research is to combine psychological and technological knowledge as to identify and exploit successful human-agent interactions. This project is a collaboration with the Human-Technology Interaction group of Eindhoven University of Technology and Smart Homes.

PhD

Prof.dr. E.O. Postma (PhD supervisor)

http://www.tilburguniversity.edu/webwijs/show/&uid=e.o.postma

Prof.dr. J. van den Herik (PhD supervisor)

http://www.tilburguniversity.edu/webwijs/show/&uid=h.j.vdnherik

Dr. Ir. P.H.M. Spronk (PhD co-supervisor)

http://www.tilburguniversity.edu/webwijs/show/?uid=p.spronck

Dr. M. Reynaert

http://www.tilburguniversity.edu/webwijs/show/?uid=reynaert

multipleResearchers

A mashup is a webapplication that combines data from more than one source into an integrated experience. A “political mashup” brings together political information produced by political parties (their promises and actual deeds) and information on the reception of these by professional users and the general public. Such a political mashup will enable novel historical research, creates challenging opportunities for computational linguists, and offers valuable and innovative testing grounds for information scientists. We aim to collect, standardize in digital format, and integrate these four types of political data: election manifestos and party websites, parliamentary proceedings, political news, and political user generated content found on the web. We restrict the corpus to data ranging from 1946 to the present. Integration of these data sources consists of creating connections between them along three dimensions: temporal happening in the same period, being about the same event; political Researchers involving the same parties and/or members of parties and/or other politically influential Researchers; political issue being about the same political issue. The ICT infrastructure created by the project aims to make large-scale comparative research (both diachronic and synchronic) on thisdata possible and effective. We propose a true mashup in which users do not even experience different sources.Access to the data is bysemantically meaningful items, like events, political issues, persons, and parties, not only by search terms. Users can group related pieces of data along meaningful dimensions and display data in several ways.

Dr. M. Reynaert

http://www.tilburguniversity.edu/webwijs/show/?uid=reynaert

The SoNaR project, a trans-national collaboration funded by the STEVIN programme between the universities of Nijmegen, Leuven, Ghent, Utrecht, Twente, and Tilburg, aims at the creation of a large corpus of minimally 500 million words of contemporary written Dutch, to serve as a generic reference for future research on linguistics and language use, for the development of dictionaries and grammars, and for research in language and speech technology. For all these purposes it is vital that large amounts of text are available, in a digital format that facilitates exploitation by different types of users. The corpus is based on a design made in the predecessor STEVIN D-CoI project. In the corpus, standard Dutch texts are included written after 1954 by writers from the Netherlands and Flanders. Texts can be written by native speakers, but may also be Dutch texts translated from another language by professional translators. Texts are gathered from various domains and genres, and a wide coverage of topics is aimed at. As far as possible, texts are included in their complete form. Special attention is given to texts from new media, such as websites, SMS messages, e-mail, and chat. Research into these new types of text is relatively new, and is in need of reference corpora. The role of Tilburg University in this trans-national research project is to be the node where all incoming texts are normalized (cleaned, corrected, and tokenized), and converted to a standard XML format, using automatic means wherever possible.

J. Janssens, PhD student

http://www.tilburguniversity.edu/webwijs/show/&uid=303996?uid=303996

The POSEIDON project is coordinated by the Embedded Systems Institute in Eindhoven and is a collaboration of various universities and Thales in Hengelo. The focus of the project is on anomaly detection in the domain of Maritime Safety and Security. The approach followed is based on probabilistic machine learning. Observations in the maritime domain are used to create a statistical model of what is considered to be normal or usual. Deviations from the model are considered to be outliers depending on the local context of observations. Dedicated algorithms are developed and evaluated on many datasets taken from a large variety of domains. The best algorithms will be applied to the maritime domain by tailoring the data representations as to reflect the relevant domain knowledge in an appropriate manner.

G. van Lankveld, PhD student

http://www.tilburguniversity.edu/webwijs/show/&uid=411527?uid=411527

Intelligent beings have the ability to adapt to changes in their environment. In Life Science research often simulations are used in which computer-controlled agents fulfil the Function of intelligent beings. Usually these agents are not adaptive, since adaptive behaviour in such environments can only be relied upon when it meets certain requirements of effectiveness and efficiency that are in conflict with each other. In this research we investigate existing and newly-designed machine-learning techniques to implement reliable, rapid adaptive behaviour in complex environments. We will use modern computer games as experimental environments, since such games can be considered advanced simulations of the real world. The problem statement of our proposed research is as follows: “To what extent is it possible to design reliable machine-learning techniques that allow situated agents in complex environments, in particular in modern computer games, to rapidly adapt to environmental changes?” In the research, as far as the “environmental changes” are concerned, the current focus is on adapting towards human players. In particular, we intend to adapt to the psychological profile of the human player. To that end, that psychological profile must be established in-game. We intend to do so based on solid psychological grounds.

multipleResearchers

Dr. Ir. P.H.M. Spronk

http://www.tilburguniversity.edu/webwijs/show/?uid=p.spronck

Prof.dr. J. van den Herik (PhD supervisor)

http://www.tilburguniversity.edu/webwijs/show/&uid=h.j.vdnherik

Dr. P. Vogt

http://www.tilburguniversity.edu/webwijs/show/&uid=343858?uid=343858

Dr. Paul Vogt and J. Douglas Mastin are currently analyzing the first data sets containing videos of interactions between infants and their caregivers, which they acquired during months of fieldwork in Mozambique in 2010. The research is part of Paul Vogt's NWO VIDI project called CASA MILA, which focuses on the relation between the use of multimodal gestures and vocabulary development among infants. Douglas Mastin will soon return to Mozambique to collect additional data. Later this year the researchers will start to collect similar data in the Netherlands for cross-cultural comparisons.

multipleResearchers

J. D. Mastin, PhD student

http://www.tilburguniversity.edu/webwijs/show/?uid=j.d.mastin

H. Buisman, PhD student

http://www.tilburguniversity.edu/nl/webwijs/show/?uid=h.j.buisman

Social signal processingis about processing signals that are produced during social interactions between agents (human-human or human-machine) or signals that provide information about these agents. In this project the focus lies on the development of speech analysis algorithms that extract and interpret the elements of speech that are not about what people say, but rather about how people say what they say, also referred to as prosody. This work is expected to improve experimental behavioral research and eventually contribute to providing machines with empathic capabilities.

PhD

Prof. dr. J. van den Herik (supervisor)

http://www.tilburguniversity.edu/webwijs/show/&uid=286943?uid=286943

Prof. dr. E. Postma (supervisor)

http://www.tilburguniversity.edu/webwijs/show/&uid=370030?uid=370030

Dr. M. Goudbeek (co-supervisor)

http://www.tilburguniversity.edu/webwijs/show/&uid=671693?uid=671693

Dr. M.Postma (co-supervisor)

http://www.tilburguniversity.edu/webwijs/show/&uid=720333?uid=720333

Dr. A. Alishahi

http://www.tilburguniversity.edu/webwijs/show/?uid=a.alishahi

My primary research interest involves the development of computational models of human language acquisition. Computational modeling is an effective tool for studying human cognition: whereas linguistic and psychological theories often give a high level explanation for the experimental data, computational models provide a detailed account of the underlying mechanisms for the cognitive task at hand. Moreover, the behaviour of a model can be directly compared to that of humans through computational simulation.