Latent Semantic Similarity in Unstructured, Initial Dyadic Interactions
Download as pptx, pdf0 likes455 views
This document summarizes a research study that examined how latent semantic similarity develops in initial interactions between strangers. The study hypothesized that latent semantic similarity, as measured by a text analysis program, would be positively correlated with how much dyad members talked to each other, looked at each other, and acknowledged each other nonverbally. The results supported this, finding significant correlations between latent semantic similarity and various behavioral and self-report measures of the interaction. A factor analysis revealed that aspects like looking, acknowledging, gesturing while talking, and smiling mediated the relationship between latent semantic similarity and factors like the percentage of repeated words and overall word count.
Latent Semantic Similarity in Unstructured, Initial Dyadic Interactions
- 1. Latent Semantic Similarity in Unstructured, Initial Dyadic Interactions Vivian Ta, Meghan Babcock, Dr. William Ickes Oral Presentation for ACES March 26, 2014
- 2. Introduction How do people develop a basis of understanding each other? Developing a common-ground understanding1 or an intersubjective meaning context2 Using the same words in essentially the same way Involved behaviors: 1) How much the interaction partners talk to each other 2) How much the interaction partners look at each other 3) How much the interaction partners acknowledge each other 1Abbeduto, Short-Meyerson, Benson, Dolish, & Weissman, 1998; Kecskes & Zhang, 2009; Krauss & Fussell, 1991; Schober & Clark, 1989; Wilkes-Gibbs & Clark, 1992 2Gesn & Ickes, 1999; Morganti, 2008
- 3. Latent Semantic Analysis Program that determines the contextual meaning of any text Exams the relationships among words (Landauer & Dumais, 1997; Landauer et al., 1998) Word/word pattern usage LSA Pairwise Comparison program Analyzes the similarity of 2 blocks of text Produces an index from -1 to 1 to indicate overall degree of latent semantic similarity (LSS index) Measures overall semantic similarity
- 4. Hypothesis LSS will be positively correlated with 1. How much dyad members talk to each other 2. How much dyad members look at each other 3. How much dyad members acknowledge each other
- 5. Participants Archival data from Ickes, Tooke, Stinson, Baker, Bissonnette (1988) 46 dyads (92 students) Male-Male (20) & Female-Female (26) Strangers Undergraduate students
- 6. Procedure Participants came into Social Interaction Lab Experimenter left to retrieve important items Covertly audio-videotaped for 6 minutes Completed a post-interaction questionnaire (e.g., did you like your partner, how smooth was the interaction, how awkward was the interaction, etc.) Coded various behaviors (e.g., duration of mutual gazes, expressive gestures, head nods, etc.) Transcripts of conversations were created
- 7. Procedure Transcripts were separated into 2 electronic text files; each file contained only one dyad members portion of the conversation These files were then transferred into the LSA program
- 9. Results Significant Correlations of LSS with Dyad-Level Behavioral & Post- Interaction Measures Total number of conversation sequences 0.46** Total number of speaking turns 0.45** Total duration of speaking turns 0.60*** Total number of directed gazes 0.35* Total number of mutual gazes 0.47** Total number of nonverbal acknowledgements (head nods) 0.43** Note: * = p < .025; ** = p < .01; *** = p < .001
- 10. Results Unique Significant Correlations of LSS with Dyad-Level Behavioral and Post- Interaction Measures Total word count 0.61* Total number of questions asked 0.43* Number of times dyad members talked about others -0.42* Total percentage of positive entries 0.38* Total frequency of positive affect 0.47* Total frequency of expressive gestures 0.58** Total duration of expressive gestures 0.44* How smooth did you think the interaction was for your partner? 0.36* How awkward did you think the interaction was for your partner? -0.38* How much did you like your partner? 0.35* Note: * = p < .025; ** = p < .001
- 11. Mediation Analysis Factor analysis revealed 4 factors: 1- Looking and acknowledging 2- Gesturing while talking 3- Using questions to advance/sustain conversation 4- Smiling and laughing Factors 1-4 % of repeated words & overall word count LSS + + +
- 12. Conclusion LSS develops from of a highly involving interaction in which a lot of verbal information and nonverbal cues are exchanged by both dyad members % of repeated words and total word count as mediator between behaviors and LSS