Link to the University of Pittsburgh Homepage
Link to the University Library System Homepage Link to the Contact Us Form

LINGUISTIC ENTRAINMENT IN MULTI-PARTY SPOKEN DIALOGUES

Rahimi, Zahra (2019) LINGUISTIC ENTRAINMENT IN MULTI-PARTY SPOKEN DIALOGUES. Doctoral Dissertation, University of Pittsburgh. (Unpublished)

This is the latest version of this item.

[img]
Preview
PDF (corrected format)
Updated Version

Download (1MB) | Preview

Abstract

Entrainment is the propensity of speakers to begin behaving like one another in conversations. Evidence of entrainment has been found in multiple aspects of speech, including acoustic-prosodic and lexical. More interestingly, the strength of entrainment has been shown to be associated with numerous conversational qualities, such as social variables. These two characteristics make entrainment an interesting research area for multiple disciplines, such as natural language processing and psychology. To date, mainly simple methods such as unweighted averaging have been used to move from pairs to groups, and the focus of prior multi-party work has been on text rather than speech (e.g., Wikipedia, Twitter, online forums, and corporate emails). The focus of this research, unlike previous studies, is multi-party spoken dialogues. The goal of this work is to develop, validate, and evaluate multi-party entrainment measures that incorporate characteristics of multi-party interactions, and are associated with measures of team outcomes.
In this thesis, first, I explore the relation between entrainment on acoustic-prosodic and lexical features and show that they correlate. In addition, I show that a multi-modal model using entrainment features from both of these modalities outperforms the uni-modal model at predicting team outcomes. Moreover, I present enhanced multi-party entrainment measures which utilize dynamics of entrainment in groups for both global and local settings. As for the global entrainment, I present a weighted convergence based on group dynamics. As the first step toward the development of local multi-party measures, I investigate whether local entrainment occurs within a time-lag in groups using a temporal window approach. Next, I propose a novel approach to learn a vector representation of multi-party local entrainment by encoding the structure of the presented multi-party entrainment graphs. The positive results of both the global and local settings indicate the importance of incorporating entrainment dynamics in groups. Finally, I propose a novel approach to incorporate a team-level factor of gender-composition to enhance multi-party entrainment measures. All of the proposed works are in the direction of enhancing multi-party entrainment measures with the focus on spoken dialogues although they can also be employed on text-based communications.


Share

Citation/Export:
Social Networking:
Share |

Details

Item Type: University of Pittsburgh ETD
Status: Unpublished
Creators/Authors:
CreatorsEmailPitt UsernameORCID
Rahimi, Zahrazrahimi11@gmail.comzar100000-0002-4135-0336
ETD Committee:
TitleMemberEmail AddressPitt UsernameORCID
Thesis AdvisorLitman, Dianelitman@cs.pitt.edu
Committee MemberHwa, Rebecca
Committee MemberAshley, Kevin
Committee MemberMorency, Louis-Philippe
Date: 30 August 2019
Date Type: Publication
Defense Date: 28 May 2019
Approval Date: 30 August 2019
Submission Date: 22 June 2019
Access Restriction: No restriction; Release the ETD for access worldwide immediately.
Number of Pages: 137
Institution: University of Pittsburgh
Schools and Programs: School of Computing and Information > Intelligent Systems Program
Degree: PhD - Doctor of Philosophy
Thesis Type: Doctoral Dissertation
Refereed: Yes
Uncontrolled Keywords: Alignment, Dialogue, Entrainment embedding, Multi-party, Natural language processing, Spoken conversation
Date Deposited: 30 Aug 2019 15:45
Last Modified: 30 Aug 2019 15:45
URI: http://d-scholarship.pitt.edu/id/eprint/37265

Available Versions of this Item


Metrics

Monthly Views for the past 3 years

Plum Analytics


Actions (login required)

View Item View Item