Collaborative Research: What's the question? A cross-linguistic investigation into compositional and pragmatic constraints on the question under discussion

Ohio State University, The, Columbus OH

Investigators

Abstract

The greatest value of language resides in its meaning, in the information that is exchanged in discourse. This project is part of a broad scholarly effort to find out how that meaning is determined, and to identify what inferences can be drawn from what is said. Recent research on meaning in language, including much published work by the four PIs on this project, shows that the meaning of a sentence is dramatically affected by the context in which the sentence is uttered. Specifically, the three-institution project team has shown in past work how the meaning of a sentence is greatly affected by what question the speaker seeks to address. The fact that a sentence meaning can only be fully grasped in terms of the question the uttered sentence answers creates a problem because the question addressed by a given utterance is typically implicit. So, often a listener can only fully understand the meaning of an uttered sentence by recreating the question that was answered. To do this, the listener must reason from many available clues, for example the intonation used in the utterance, and what is known of the speaker's intentions. The goal of this project is to conduct empirical and theoretical work which will identify the process by which implicit questions are revealed. The work to be undertaken in this project is of intrinsic interest in understanding human culture and human communication, and also has significant practical applications, for example in the field of Natural Language Processing. To build computers that can understand and use language, the features that inform human language understanding must be identified. The project also offers advanced training opportunities for young researchers at US institutions, develops experimental and other research methodologies that can impact a broad range of fields, and analyzes a foundational problem in language which is related to strategically valuable natural language technology. The project is set in the context of a Question Under Discussion model in which the context incorporates a dynamically evolving stack of questions. The problem for this model is to figure out what question is addressed by any given stretch of discourse. Three types of constraint on the question will be explored cross-linguistically: lexical constraints contributed by, for instance, focus sensitive expressions and factive predicates, information-structural constraints, including those imposed by intonation and cleft constructions, and contextual constraints, such as Gricean principles. This goal is pursued by expanding methodologies which the three-institution team has jointly pioneered in previous work: a mix of experimental research, corpus-based studies of naturally occurring utterances, and cross-linguistic fieldwork. The development effort is spearheaded with English and Paraguayan Guaraní (Tupí-Guaraní) and selective work is performed on other languages to test aspects of the work that are particularly germane, including K'iche' and Kaqchikel (both Mayan), Hungarian (Ugric) and Tagalog (Austronesian). Throughout the project, there will be an emphasis on the development of data-collection methods appropriate for use with theoretically untrained native speakers