This paper presents the semi-automatically created Corpus of Aligned Read Speech Including Annotations (CARInA), a speech corpus based on the German Spoken Wikipedia Corpus (GSWC). CARInA tokenizes, consolidates and organizes the vast, but rather unstructured material contained in GSWC. The contents are grouped by annotation completeness, and extended by canonic, morphosyntactic and prosodic annotations. The annotations are provided in BPF and TextGrid format. It contains 194 hours of speech material from 327 speakers, of which 124 hours are fully phonetically aligned and 30 hours are fully aligned at all annotation levels. CARInA is freely available1, designed to grow and improve over time, and suitable for large-scale speech analyses or machine learning tasks as illustrated by two examples shown in this paper.

This paper describes the handling of multimodality in the Embassi system. Here, multimodality is treated in two modules. Firstly, a modality fusion component merges speech, video traced pointing gestures, and input from a graphical user interface. Secondly, a presentation planning component decides upon the modality to be used for the output, i.e., speech, an animated life-like character (ALC) and/or the graphical user interface, and ensures that the presentation is coherent and cohesive. We describe how these two components work and emphasize one particular feature of our system architecture: All modality analysis components generate output in a common semantic description format and all render components process input in a common output language. This makes it particularly easy to add or remove modality analyzers or renderer components, even dynamically while the system is running. This plug and play of modalities can be used to adjust the system's capabilities to different demands of users and their situative context. In this paper we give details about the implementations of the models, protocols and modules that are necessary to realize those features.

In this paper, we investigate the practical applicability of Co-Training for the task of building a classifier for reference resolution. We are concerned with the question if Co-Training can significantly reduce the amount of manual labeling work and still produce a classifier with an acceptable performance.

We report on experiments in reference resolution using a decision tree approach. We started with a standard feature set used in previous work, which led to moderate results. A closer examination of the performance of the features for different forms of anaphoric expressions showed good results for pronouns, moderate results for proper names, and poor results for definite noun phrases. We then included a cheap, language and domain independent feature based on the minimum edit distance between strings. This feature yielded a significant improvement for data sets consisting of definite noun phrases and proper names, respectively. When applied to the whole data set the feature produced a smaller but still significant improvement.

Multimodality is likely to play a key role in household technology interfaces of the future offering as it can, enhanced efficiency, improved flexibility and increased user preference. These benefits are unlikely to be realized however unless such interfaces are well designed specifically with regard to modality allocation and configuration. We report on a methodology aimed at evaluating modality usage, which involves a combination of two sets of heuristics, one derived from a description of modality properties, the other concerned with issues of usability. We describe how modality properties can be reformulated into a procedural style checklist and then describe the implementation of this methodology and the issues we were able to highlight in the context of the EMBASSI 'Home' system, a multimodal system which aims to provide a natural and intuitive interface to a potentially open-ended array of appliances within the home.

One limitation in prosody research is the lack of sufficient prosodically labelled speech data. In this paper, we present research on an automatic labelling system that is able to produce a phonological tonal labelling according to the ToBI like intonation model for German developed by Fery. The system is not totally dependent on the specific language and/or labelling system, as it uses corpus based techniques such as an HMM based word, syllable and phoneme segmentation and a decision tree learning algorithm (C4.5) for the phonetic-phonology mapping. The current system was trained on about 1 hour of expert prosodically labelled speech from a single male radio news announcer. We present experiments for finding a suitable feature set drawn from features that describe the prosodic correlates fundamental frequency, duration and intensity as well as some lexical and syntactic features. With the best feature set, we achieve a recognition rate of 78.7% for speaker dependent recognition of ToBI labels (simultaneously predicting prominence and phrasing) and 86.9% for the simpler accented/not accented decision. Although the system’s accuracy is well below that of human transcribers, it is a useful tool actively used in our laboratory due to it’s ability to process large amounts of speech data at low costs.