Projects

The project objective is to bring to the software community an enabling technology to support affective interactions, in particular (i) to create a framework to enrich interactions and applications with an affective dimension, (ii) to implement a toolkit for affective computing combining a set of components addressing affective knowledge acquisition, representation, reasoning, planning, communication and expression, and (iii) to verify under which conditions the hypothesis that emotion, as well as other affective phenomena, contributes to improve rationality and general intelligent behaviour of the synthetic characters, thus leading to more believable interactions between humans and computers.

The general theme of COST 258 was the improvement of quality and naturalness of synthetic speech. The COST action comprised leading research institutions from 17 European countries and was led by Prof. Eric Keller from the University of Lausanne.

Natural sounding speech is a key factor for the acceptability of practical voice output systems whereby the main factors contributing to naturalness are segmental quality and prosody. Current improvements in the segmental quality of synthesized speech have made it clear that truly high-quality speech synthesis now depends crucially on adequate and natural sounding prosody as well. Currently, most research in prosody is directed at intonation and its realization through fundamental frequency (f0), while duration (and amplitude) have been mostly regarded as secondary factors. The project aimed at a better model of segmental duration. Moreover, we wanted to get a clearer understanding of the relation between discourse structure and prosodic parameters.

In the last years, a rapid progress in the fields of computer animation and simulation could be observed, thus a growing number of realistic, believable computer generated synthetic actors in a variety of roles and tasks can be expected in the near future. While research and technical development in this field have concentrated on the development of (individual) personalities for synthetic actors, this project raised the question of the synthetic being as product and participant of different cultures. Its goal was to raise awareness of the social and cultural implications in the development and research community on the one hand and in the user community on the other. This goal was pursued through literature study and analysis of international research work and ended with an international workshop. The results have been published in the book Agent Culture: Human-Agent Interaction in a Multicultural World, edited by R. Trappl and S. Payr.

Demands for more rapid and cost-effective models of foreign language acquisition in business means increasing learner autonomy and customisation with greater reliance on technology. Learning packages need to be multi-standard and portable to different delivery systems. In SELECT, a Language and Culture Learning Workbench was developed, and an electronic delivery learning model for access to language learning resources in English, French and Portuguese was defined and implemented. OFAI was a subcontractor to the University of Wolverhampton. It supplied part of the so-called Language Learner Workbench (LLW). Namely, it developed a tool that enables the user to highlight any word or phrase in a text and have it correctly pronounced in a natural context. This was done by providing large speech corpora aligned with text. These corpora are indexed to allow for the easy identification of all occurences of a word or phrase.

This nationally sponsored project (Austrian Federal Ministry of Science and Transport) is concerned with the realization of synthetic speech with paramaters corresponding to the emotional state of the speaker. In this attempt we went beyond the means of usually utilized copy-synthesis and use a parameter space covering not only f0, duration and amplitude, but also voice quality parameters, spectral energy distribution, harmonics-to-noise ratio, and articulatory precision. Results from previous studies on these parameters in natural emotional speech were used to control the phoneme-to-speech synthesizer also used in the VieCtoS project.

The FACILE project aimed at the development of a system for the categorisation of texts from the area of finance and business news in an exact and specific way. The intended users of the system were institutions from finance and commerce that have a vital interest in up-to-date business information stemming from online news agencies and periodicals. An important consideration in FACILE has been its use across country and language borders. The possibility to process texts in various languages and to derive factual information in a language independent, formatted form shall should allow for the rapid dissemination of information across borders.

The aim of the TAMIC-P project was to improve the access to pension-related data for office clerks to provide a better service for the citizen. Previously, person-specific relevant data were stored in a number of databases which had to be accessed separately. Additionally, information on laws and regulations was available on paper and on-line. In TAMIC-P an application demonstrator was implemented that allows uniform access to all relevant data via natural language. At the heart of the system is a module that allows access to heterogeneous databases. The user interacts with the system via archives of "electronic cards" that can be retrieved either by their "name" or via links from other cards and a hierarchy of pension terms. Additionally, on-line information on relevant regulatory and legal texts is provided. The consortium comprised partners from Austria, France, Germany and Italy and is centered around the Italian user Instituto Nationale per la Providenza Sociale (INPS) and the Austrian user Sozialversicherung der Bauern (SVB). Accordingly, prototypes were developed for Italian and German. ÖFAI cooperated with SVB and the DFKI to implement the German version of TAMIC-P.

In this nationally sponsored project (Austrian Science Foundation), a prototype of a concept-to-speech system was developed able to produce German utterances from a symbolic representation that contains word-specific information (phonemes, word accent, syllable structure) as well as abstract prosodic information. The representation is based on the autosegmental phonologic theory of Pierrehumbert which can be considered a de-facto standard of prosodic notation. The project focussed on developing a mapping from this symbolic description of prosody to the corresponding phonetic/acoustic parameter values (fundamental frequency, duration, intensity). An algorithm for realizing this mapping was developed and implemented. The description arrived at forms the input for a parametrizable synthesizer. The synthesizer used in the project is based on an adaptation of a demi-syllable synthesizer developed at the Institute für Nachrichtentechnik und Hochfrequenztechnik at the Vienna University of Technology.

GIST aimed at the multilingual generation of instructional texts, in particular forms. The project was sponsored within the Linguistic Research and Engineering (LRE) Sector of the Telematics Application Programme of the European Commission, the Austrian contribution was nationally sponsored. OFAI's main responsibility in this project was the development of the German tactical generator. For this aim we adapted the FUF generator by replacing the original morphological component with a version of X2MorF to enable it to cope with the rich inflectional morphology of German and by providing it with an HPSG-inspired grammar of German. We also contributed to the development of a text structuring component and the user interface.