CHI 99 special interest group on natural language in computer-human interaction

Abstract

With the growing interest in human-computer interfaces that use natural language in some way, researchers and practitioners who work on these interfaces are finding that two general fields of research, CHI and natural language processing (NLP), are complementary and converging. In the CHI research community, there have been investigations on a number of related issues such as usability of text and graphics in on-line documentation, hypertext, spoken-dialogue interfaces, and language/audio resources. In the NLP research community, there is increasing interest in use of natural language, both spoken and written, in intelligent multimodal and multimedia interfaces, e.g., International Symposium on Spoken Dialogue (ISSD-96), COOP 98 Workshop on The Use of Herbert H. Clark's Models of Language Use for the Design of Cooperative Systems, 1998 AAAI Workshop on Representations for Multi-Modal Human-Computer Interaction, and Coling-ACL'98 Workshop on Content Visualizations and Intermedia Representations (CVIR'98).Some technical issues of possible interest to both communities are:1. For what and under what conditions is NL effective in the human-computer interface? For what types of tasks or communication? How does modality influence its effectiveness? How does its effectiveness in computer media differ from that in traditional forms of communication such as face-to-face conversation and print media? How do performance limitations of NLP technologies (e.g., .speech recognition errors) influence effectiveness?2. What are the critical technical requirements for NLP to be effective in the human-computer interface, e.g., coordination of generated text and graphics, incremental and robust interpretation, and modeling turn-taking and initiative in dialogue? What technical requirements arise in transferring technology developed for one language to systems for users of another language (e.g., languages using different writing systems)? How should effectiveness of NLP technologies be evaluated?