The objective of this project is to develop automated methods for modeling and recognizing participants’ social actions and social roles in small groups based on the linguistic evidence in their interactions. These methods will be based on detecting a wide range of linguistic features (lexical, syntactic, discourse, etc.) so that they can apply effectively to different communication media and, with appropriate modifications, to different languages. When fully functional, the DSARMD system will provide an efficient and effective means for understanding the social dynamics of small groups across languages and cultures, as well as computer-mediated communication.

For more information, please contact:

Prof. Tomek Strzalkowski, Principal Investigator
University at Albany, SUNY
ILS Room 262B, Social Science Building
Albany, NY 12222.
Email: tomek [at] albany [dot] edu
Phone: 518-442-2608; Fax: 518-442-2606

Executive Summary

The DSARMD project (Detecting Social Actions and Roles in Multiparty Dialogue) investigates the dynamics of small group interactions across media, cultures and languages. Specifically, we are developing computational models of how certain social phenomena such as leadership, power, and conflict are signaled and reflected in language through the choice of lexical, syntactic, semantic and conversational forms by discourse participants in face-to-face discourse and in on-line interaction. Our objective is to build a prototype system that given a representative segment of multiparty task-oriented dialogue would automatically detect, with a high degree of accuracy, those social and cultural elements in language and discourse structure that are commonly associated with (a) conflict and disagreement, (b) power and dominance and (c) leadership. In this work we adopt an empirical approach based on observable linguistic features that can be automatically extracted from available sources. This approach leads directly to tools which will increase the ability of social science researchers to collect and analyze data automatically, without the current requirement for intensive human data mark-up tasks, and to additional tools which enable automated analysis of collection for intelligence purposes.

DSARMD Detecting Social Actions and Roles in Multiparty Dialogue

The immediate objective of our research is to recognize and understand small group interactions through their use of language in both face-to-face and on-line discussion settings. We are exploring and comparing behaviors in English, Urdu, and Chinese speaking groups, initially with respect to power and conflict and their manifestation through such social actions as disagreement, power projection, and non-cooperation; and subsequently moving on to higher-level phenomena of leadership and opposition by detecting patterns of social actions in task oriented dialogues. It is our hypothesis that social actions are reflected in the language and structure of dialogue and that they can be recognized automatically from characteristic patterns of linguistic and dialogic features, which we call social action signatures (SAS). We further hypothesize that more complex social roles such as leadership or opposition within a group can be circumscribed by sequences of social actions. Our research will also address how face-to-face linguistic behaviors are modified by the switch to on-line interactions via chat where, among other things, means of expression are more limited, and how the same social actions may be signaled by different linguistic phenomena in different languages and cultures.

The project is funded by the Intelligence Advanced Research Projects Activity (IARPA) as part of the Socio-Cultural Issues in Language (SCIL) program. This is a joint work with Lockheed Martin Corporation.