We present the dialogue module of the speech-to-speech translation system VERBMOBIL. We follow

In this paper, we present a statistical approach for dialogue act processing in the dialogue component of the speech-to-speech translation system VERBMOBIL. Statistics in dialogue processing is used to predict follow-up dialogue acts. As an application example we show how it supports repair when unexpected dialogue states occur.

Automatic summarization of open-domain spoken dialogues is a relatively new research area. This article introduces the task and the challenges involved and motivates and presents an approach for obtaining automatic-extract summaries for human transcripts of multiparty dialogues of four different genres, without any restriction on domain. We address the following issues, which are intrinsic to spoken-dialogue summarization and typically can be ignored when summarizing written text such as news wire data: (1) detection and removal of speech disfluencies; (2) detection and insertion of sentence boundaries; and (3) detection and linking of cross-speaker information units (question-answer pairs). A system evaluation is performed using a corpus of 23 dialogue excerpts with an average duration of about 10 minutes, comprising 80 topical segments and about 47,000 words total. The corpus was manually annotated for relevant text spans by six human annotators. The global evaluation shows that for the two more informal genres, our summarization system using dialoguespecific components significantly outperforms two baselines: (1) a maximum-marginal-relevance ranking algorithm using TF*IDF term weighting, and (2) a LEAD baseline that extracts the first n words from a text. 1.

New and challenging requirements arise for the dialogue processing component in the speech-to-speech translation system Verbmobilcomponent. It has to cope with both unexpected and vague input as well as gaps in the input. The design is based on a large corpus of transliterated dialogues which provide the data to start from. A careful analysis of this corpus and of the requirements from other components of verbmobil resulted in a hybrid approach consisting of both knowledge based as well as statistic based processing. In this paper, we present the design process and the resulting architecture. Using the corpus, we made various experiments to evaluate the first design of the component.

We describe a novel method for coping with ungrammatical input based on the use of chart-like data structures, which permit anytime processing. Priority is given to deep syntactic analysis. Should this fail, the best partial analyses are selected, according to a shortest-paths algorithm, and assembled in a robust processing phase. The method has been applied in a speech translation project with large HPSG grammars.

This paper presents a novel multi-lingual progress protocol generation module. The module is used within the speech-to-speech translation system VERBMOBIL. The task of the protocol is to give the dialogue partners a brief description of the content of their dialogue. We utilize an abstract representation describing, for instance, thematic information and dialogue acts of the dialogue utterances. From this representation we generate simplified paraphrases of the individual turns of the dialogue which together make up the protocol. Instead of writing completely new software, the protocol generation component is almost exclusively composed 'of already existing modules in the system which are extended by planning and formatting routines for protocol formulations. We describe how the abstract information is extracted from user utterances in different languages and how the abstract thematic representation is used to generate a protocol in one specific language. Future directions are given.

This paper revisits word lattice search whose task is to find a plausible semantic interpretation for a given utterance. Our approach of interleaved search and analysis is designed to break the frontier of "toy" applications. The framework is implemented in two interacting modules, running in parallel. Instead of simply parsing a word lattice, we rather do tree decoding with a probabilistic approximation of a given grammar, employing a beam search strategy. Logical form is build up in tandem according to the decoded derivation histories, using a codescriptive HPSG grammar for dialog turns. The proposed architecture only uses the knowledge necessary in every processing step, the key aspect being an asynchronous coupling of the two specialized modules.

Unification-based grammar formalisms have become the predominant paradigm in natural language processing (NLP) and computational linguistics (CL). Their success stems from the fact that they can be seen as high-level declarative programming languages for linguists, which allow them to express linguistic knowledge in a monotonic fashion. Moreover, such formalisms can be given a precise, set-theoretical semantics. This paper presents TDL, a typed feature-based language and inference system, which is specifically designed to support highly lexicalized grammar theories like HPSG, FUG, or CUG. TDL allows the user to define (possibly recursive) hierarchically-ordered types, consisting of type constraints and feature constraints over the boolean connectives , , and :. TDL distinguishes between avm types (open-world reasoning), sort types (closed-world reasoning), built-in types and atoms, and allows the declaration of partitions and incompatible types. Working with partially as ...

Over the last few years, constraint-based grammar formalisms have become the predominant paradigm in natural language processing and computational linguistics From the viewpoint of computer science typed feature structures can be seen as data structures that allow the representation of linguistic knowledge in a uniform fashion Type expansion is an operation that makes constraints of a typed feature structure explicit and determines its satisfiability We describe an efficient expansion algorithm that takes care of recursive type definitions and permits the exploration of different expansion strategies through the use of control knowledge This knowledge is specified on a separate layer independent of grammatical information The algorithm as presented in the paper, has been full> implemented in COMMON LISP and is an integrated part of the typed feature formallsm TDC that is employed in several large NL projects

This paper presents a neural network approach to the problem of finding the dialogue act for a given utterance. So far only symbolic, decision tree and statistical approaches were utilized to deal with a corpus as large as the VERBMOBIL corpus. We propose solutions to the questions of representing speech, network architecture and training in this context. We argue that, when using neural networks, a task like this can only be solved in a modular approach where training data is split and processed by different components of a larger network. Special care must be taken in constructing a feeding mechanism that avoids oscillatory behaviour due to the heterogeneous data. We were successful in constructing a modular neural network that yielded interesting time-sensitive properties as well as recognition rates superior to most other methods. A first attempt at devising a hybrid system got very close to the best results of this field which suggests further enhancement in future architectures. 1