Behavioral testing in NLP allows fine-grained evaluation of systems by examining their linguistic capabilities through the analysis of input-output behavior. Unfortunately, existing work on behavioral testing in Machine Translation (MT) is currently restricted to largely handcrafted tests covering a limited range of capabilities and languages. To address this limitation, we propose to use Large Language Models (LLMs) to generate a diverse set of source sentences tailored to test the behavior of MT models in a range of situations. We can then verify whether the MT model exhibits the expected behavior through matching candidate sets that are also generated using LLMs. Our approach aims to make behavioral testing of MT systems practical while requiring only minimal human effort. In our experiments, we apply our proposed evaluation framework to assess multiple available MT systems, revealing that while in general pass-rates follow the trends observable from traditional accuracy-based metrics, our method was able to uncover several important differences and potential bugs that go unnoticed when relying only on accuracy.

Code-switching (CS), i.e. mixing different languages in a single sentence, is a common phenomenon in communication and can be challenging in many Natural Language Processing (NLP) settings. Previous studies on CS speech have shown promising results for end-to-end speech translation (ST), but have been limited to offline scenarios and to translation to one of the languages present in the source monolingual transcription). In this paper, we focus on two essential yet unexplored areas for real-world CS speech translation: streaming settings, and translation to a third language (i.e., a language not included in the source). To this end, we extend the Fisher and Miami test and validation datasets to include new targets in Spanish and German. Using this data, we train a model for both offline and streaming ST and we establish baseline results for the two settings mentioned earlier.

Code switching (CS) refers to the phenomenon of interchangeably using words and phrases from different languages. CS can pose significant accuracy challenges to NLP, due to the often monolingual nature of the underlying systems. In this work, we focus on CS in the context of English/Spanish conversations for the task of speech translation (ST), generating and evaluating both transcript and translation. To evaluate model performance on this task, we create a novel ST corpus derived from existing public data sets. We explore various ST architectures across two dimensions: cascaded (transcribe then translate) vs end-to-end (jointly transcribe and translate) and unidirectional (source -> target) vs bidirectional (source <-> target). We show that our ST architectures, and especially our bidirectional end-to-end architecture, perform well on CS speech, even when no CS training data is used.

The Interdisciplinary Longitudinal Study on Adult Development and Aging (ILSE) was created to facilitate the study of challenges posed by rapidly aging societies in developed countries such as Germany. ILSE contains over 8,000 hours of biographic interviews recorded from more than 1,000 participants over the course of 20 years. Investigations on various aspects of aging, such as cognitive decline, often rely on the analysis of linguistic features which can be derived from spoken content like these interviews. However, transcribing speech is a time and cost consuming manual process and so far only 380 hours of ILSE interviews have been transcribed. Thus, it is the aim of our work to establish technical systems to fully automatically transcribe the ILSE interview data. The joint occurrence of poor recording quality, long audio segments, erroneous transcriptions, varying speaking styles & crosstalk, and emotional & dialectal speech in these interviews presents challenges for automatic speech recognition (ASR). We describe our ongoing work towards the fully automatic transcription of all ILSE interviews and the steps we implemented in preparing the transcriptions to meet the interviews’ challenges. Using a recursive long audio alignment procedure 96 hours of the transcribed data have been made accessible for ASR training.