Challenges of Discourse Processing

Discourse analysis remains an unresolved challenge in Computational Linguistics, in spite of the numerous theoretical works that have been developed in the past two decades. This situation is mainly due to the complexity of discourse constructions whose recognition often involves language analysis associated with domain knowledge and reasoning.Technical documents, such as procedures, requirements, and product manuals, must be relatively constrained in terms of language diversity and complexity: the goal is to make sure that users can efficiently and accurately understand these documents. For that purpose, these documents often follow authoring guidelines. These constraints make it possible to develop an accurate discourse analysis of technical documents which can be used to model their contents and to improve their overall quality.This book shows that linguistic analysis and natural language processing methods can efficiently be used to automatically recognize the discourse structures of technical documents, independently of the industrial sector and activity that is considered. Furthermore, the book presents well-founded and concrete solutions, which can be deployed in industrial contexts for various types of applications. This book begins with a presentation of the different types of technical texts. Their structure is then developed in conjunction with a survey of a number of authoring guidelines developed in the industry. The TextCoop platform and the Dislog language, designed for discourse analysis, are then presented with a large number of concrete examples, allowing readers to develop their own applications. The book concludes with an in-depth investigation of the structure of procedures and requirements.