ABSTRACT
This chapter presents a review of studies on the problem of automatic analysis of language from human dialogues in meetings, mainly carried out in the IM2 NCCR and in related initiatives, during the past decade. The words pronounced during human interactions in meetings have the potential to convey the major part of the information exchanges, although – as argued in Part II of this book – the non-verbal content can play a significant role as well. Still, pieces of information such as the topic of a discussion, the assigned tasks, deadlines, or budget, can only be determined automatically if some form of understanding of the language-based substance of a dialogue is achieved. These pieces of information can then be used for indexing and retrieval through meeting browsers, as presented in Chapters 4-6 above.
