Elamur Research⌕
☰
home / papers / 2012 / “Love ya, jerkface”: Using Sparse Log-Linear Models to Build Positive and Impolite Relationships with Teens

“Love ya, jerkface”: Using Sparse Log-Linear Models to Build Positive and Impolite Relationships with Teens

Paper · 2012 · William Yang Wang, Samantha Finkelstein, Amy Ogan et al.

Summary

The paper asks how a tutoring dialogue system could read and signal rapport with teenagers, given that closeness shows up through both praise and friendly rudeness. The authors use chat logs from an algebra peer-tutoring study with high-school students. Most pairs were friends and a few were strangers. Utterances were hand-coded for linguistic devices such as insults, challenges, brags, complaints and emoticons, and for their social function (positivity, impoliteness). Automatically extracted lexical and syntactic features were added. The task is to predict whether the partner's next turn will be impolite or positive, without seeing that turn. Because socially charged cues are rare, they compare sparse log-linear models (Lasso, elastic net) with ridge and plain logistic regression. Models were trained on friend dyads and tested separately on friends and strangers. The sparse models beat the non-sparse ones. Positivity was easier to predict for strangers, while impoliteness was easier for friends, and longer dialogue-history windows helped. The broader lesson is that the same device does different social work depending on the relationship and the speaker's role.

Key points

  • Data: 130 students in grades 7–12 (81 female) from one American high school, taking part in a 60-minute chat-based algebra peer-tutoring session. There were 54 friend dyads and 6 stranger dyads.
  • Data split:
    • 48 friend dyads for training (4538 utterances).
    • 6 friend dyads held out for testing (468 utterances).
    • 6 stranger dyads as a second test set (402 utterances).
  • Annotation:
    • Devices coded include insults, challenges, condescension/brags, message enforcers and dismissals/curses.
    • Excessive punctuation and capitalization were detected automatically.
    • Inter-rater kappa ranged from .71 to 1, with .79 for positivity and .76 for impoliteness.
  • Models compared: logistic regression, ridge, Lasso and elastic net. Elastic net and Lasso outperformed the logistic regression and ridge baselines.
  • Findings: friends' impoliteness and strangers' positivity were the more predictable targets, and wider feature windows helped most for these. Tutee bragging predicted tutor positivity, while tutor bragging negatively predicted tutee positivity.

Interesting ideas

  • I1. The authors label each surface device (insult, emoticon, brag) separately from its social function in context, since a rude remark can strengthen a bond and a smiley can soften a command. This keeps form and effect from being conflated.
  • I2. The model predicts the reaction in the partner's next turn using only earlier turns, never peeking at the target. This frames social understanding as anticipating effects rather than classifying what is already visible.
  • I3. Training on one population (friends) and testing on a structurally different one (strangers) exposes cues that do not transfer, instead of hiding them in a random split.
  • I4. Sparse penalties (Lasso, elastic net) keep rare but meaningful cues from being overweighted, and they produce a small, readable feature set.
  • I5. Per-turn binary features are built by OR-ing over all utterances of one speaker, so fragmented chat messages merge into a coherent turn.

For Mova

How these ideas could serve Mova, our small open language model — its files, stages and tests:

  • I1 (frames): prag could tag speech acts and politeness devices in one layer and their contextual function (affiliative vs. hostile) in another, so irony and teasing get explicit, explainable readings.
  • I2 (world): the world model could track a relational state between dialogue participants and forecast the likely next move, which would make dialogue answers about "how will she respond" checkable.
  • I3 (learn): evaluation could keep a held-out "different register/relationship" set as a control, revealing repair rules or features that only work on the training genre.
  • I4 (learn): when learning weights for rare pragmatic or repair features, L1/elastic-net regularization would keep models compact and their selected features inspectable, matching the explainability goal.
  • I5 (tokens): when handling chat-style input, merging consecutive same-speaker messages into one turn before tagging would reduce noise from split messages and emphatic punctuation or capitals.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

📚 Shelves 4The field of science this work belongs to (OpenAlex fields).Computer Science · 9,086 worksArtificial Intelligence · 8,889 worksArts and Humanities · 9,296 worksLanguage and Linguistics · 9,206 works

🧭 Topics 1Research subjects and languages the work is about.Dialogue systems · 401 works

Cite

Cite as William Yang Wang, Samantha Finkelstein, Amy Ogan, Alan W Black and Justine Cassell. 2012. “Love ya, jerkface”: Using Sparse Log-Linear Models to Build Positive and Impolite Relationships with Teens. Proceedings of the 13th Annual Meeting of the Special Interest Group on Discourse and Dialogue.

The paper itself, not this page. BibTeX from the library's record of the work.

@inproceedings{wang2012love,
  title = {{“Love ya, jerkface”: Using Sparse Log-Linear Models to Build Positive and Impolite Relationships with Teens}},
  author = {William Yang Wang, Samantha Finkelstein, Amy Ogan, Alan W Black and Justine Cassell},
  booktitle = {Proceedings of the 13th Annual Meeting of the Special Interest Group on Discourse and Dialogue},
  year = {2012},
  url = {https://aclanthology.org/W12-1603/}
}

References

47 references, 0 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Hua Ai, Diane J. Litman, Kate Forbes-Riley, Mihai Rotaru, Joel Tetreault, and Amruta Purandare. 2006. Using system and user performance features to improve emotion detection in spoken tutoring dialogs. In Proceedings of the Ninth International Conference on Spoken Language Processing (Interspeech 2006). find
  2. Angela M. Ardington. 2006. Playfully negotiated activity in girls talk. Journal of Pragmatics, 38(1):73 – 95. ↗
  3. Rachel E. Baker, Alastair J. Gill, and Justine Cassell. 2008. Reactive redundancy and listener comprehension in direction-giving. In Proceedings of the 9th SIGdial Workshop on Discourse and Dialogue. find
  4. Timothy W. Bickmore and Rosalind W. Picard. 2005. Establishing and maintaining long-term human-computer relationships. ACM Transactions on Computer-Human Interaction. ↗
  5. Timothy Bickmore, Laura Pfeifer, and Daniel Schulman. 2011. Relational agents improve engagement and learning in science museum visitors. In Proceedings of the 10th international conference on Intelligent virtual agents, IVA’11. find
  6. Kristy Elizabeth Boyer, Robert Phillips, Michael Wallis, Mladen Vouk, and James Lester. 2008. Balancing cognitive and motivational scaffolding in tutorial dialogue. In Proceedings of the 9th international conference on Intelligent Tutoring Systems, ITS ’08. find
  7. Penelope Brown and Stephen Levinson. 1978. Universals in language usage: Politeness phenomena. In Questions and politeness: Strategies in social interaction. find
  8. Justine Cassell, Alastair J. Gill, and Paul A. Tepper. 2007. Coordination in conversation and rapport. In Proceedings of the Workshop on Embodied Language Processing, EmbodiedNLP ’07, pages 41–50, Stroudsburg, PA, USA. Association for Computational Linguistics. find
  9. Jan Chovanec. 2009. Simulation of spoken interaction in written online media texts. Brno Studies in English. find
  10. David Crystal. 2001. Language and the internet. Cambridge University Press. ↗
  11. Jonathan Culpeper. 1996. Towards an anatomy of impoliteness. In Journal of Pragmatics. ↗
  12. Jonathan Culpeper. 2011. Impoliteness: Using language to cause offence. find
  13. Laurence Devillers and Laurence Vidrascu. 2006. Real-life emotions detection with lexical and paralinguistic cues on human-human call center dialogs. In Proceedings of the Ninth International Conference on Spoken Language Processing (Interspeech 2006). find
  14. A Gartner, M Kohler, and F Riessman. 1971. Children teach children: Learning by teaching. In New York and London: Harper and Row. find
  15. José González-Brenes and Jack Mostow. 2011. Which system differences matter? using l1/l2 regularization to compare dialogue systems. In Proceedings of the SIGDIAL 2011 Conference, pages 8–17, Portland, Oregon, June. Association for Computational Linguistics. find
  16. Jonathan Gratch, Anna Okhmatovskaia, Francois Lamothe, Stacy Marsella, Mathieu Morales, Rick J. van der Werf, and Louis-Philippe Morency. 2006. Virtual rapport. In Proceedings of the International Conference on Intelligent Virtual Agents (IVA 2006). find
  17. M. Grimm, E. Mower K. Kroschel, and S. Narayanan. 2007. Primitives-based evaluation and estimation of emotions in speech. In Speech Communication. ↗
  18. P. Gupta and R. Nitendra. 2007. Two-stream emotion recognition for call center monitoring. In Proceedings of the 8th Annual Conference of the International Speech Communication Association (Interspeech 2007). find
  19. Susan C. Herring and Asta Zelenkauskaite. 2009. Symbolic capital in a virtual heterosexual market. In Written Communication. ↗
  20. Je Hun Jeon, Rui Xia, and Yang Liu. 2010. Level of interest sensing in spoken dialog using multi-level fusion of acoustic and lexical evidence. In Proceedings of the 11th Annual Conference of the International Speech Communication Association (Interspeech 2010), IN-TERSPEECH 2010. find
  21. W. Lewis Johnson and Paola Rizzo. 2004. Politeness in tutoring dialogs: run the factory, thats what id do. In Intelligent Tutoring Systems, Lecture Notes in Computer Science. find
  22. Manfred Kienpointner. 1997. Varieties of rudeness: types and functions of impolite utterances. In Functions of Language. find
  23. S. le Cessie and J.C. van Houwelingen. 1992. Ridge estimators in logistic regression. Applied Statistics, 41(1):191–201. ↗
  24. Jackson Liscombe, Julia Hirschberg, and Jennifer J. Venditti. 2005. Detecting certainness in spoken tutorial dialogues. In Proceedings of the 6th Annual Conference of the International Speech Communication Association (Interspeech 2005). find
  25. D. Litman and K. Forbes-Riley. 2004. Predicting student emotions in computer-human tutoring dialogues. In Proceedings of the 42nd Annual Meeting of the Association for Computational Linguistics (ACL 2004). find
  26. Dong C. Liu and Jorge Nocedal. 1989. On the limited memory bfgs method for large scale optimization. Mathematical Programming, 45:503–528. ↗
  27. Andre Martins, Noah Smith, Mario Figueiredo, and Pedro Aguiar. 2011. Structured sparsity in structured prediction. In Proceedings of the 2011 Conference on Empirical Methods in Natural Language Processing, pages 1500–1511, Edinburgh, Scotland, UK., July. Association for Computational Linguistics. find
  28. Bruce M. McLaren, Sung-Joo Lim, David Yaron, and Ken Koedinger. 2007. Can a polite intelligent tutoring system lead to improved learning outside of the lab? In Proceedings of the 2007 conference on Artificial Intelligence in Education: Building Technology Rich Learning Contexts That Work. find
  29. Bruce McLaren, DeLeeuwm Krista E., and Richard E. Mayer. 2011. Polite web-based intelligent tutors: Can they im-prove learning in classrooms? In Computers and Education. find
  30. Angeliki Metallinou, Martin Wollmer, Athanasios Katsamanis, Florian Eyben, Bjorn Schuller, and Shrikanth S. Narayanan. 2011. Context-sensitive learning for enhanced audiovisual emotion classification. IEEE Transactions on Affective Computing. ↗
  31. Amy Ogan, Samantha Finkelstein, Erin Walker, Ryan Carlson, and Justine Cassell. 2012. Rudeness and rapport: Insults and learning gains in peer tutoring. In Proceedings of the 11 International Conference on Intelligence Tutoring Systems (ITS 2012). find
  32. J. Alfredo Sánchez, Norma P. Hernández, Julio C. Penagos, and Yulia Ostróvskaya. 2006. Conveying mood and emotion in instant messaging by using a two-dimensional model for affective states. In Proceedings of VII Brazilian symposium on Human factors in computing systems, IHC ’06, pages 66–72, New York, NY, USA. ACM. find
  33. A. Sharpley, J. Irvine, and C. Sharpley. 1983. An examination of the effectiveness of a cross-age tutoring program in mathematics for elementary school children. In American Educational Research Journal. ↗
  34. Helen Spencer-Oatey. 2008. Face (im)politeness and rapport. In Culturally Speaking: Culture, Communication and Politeness Theory. find
  35. Carolyn A. Straehle. 1993. ”samuel?” ”yes dear?” teasing and conversatrion rapport. In Framing in Discourse. find
  36. Bas Stronks, Anton Nijholt, Paul van Der Vet, Dirk Heylen, and Aaron Machado. 2002. Designing for friendship: Becoming friends with your eca. In Proceedings of Embodied conversational agents - let’s specify and evaluate (AAMAS). find
  37. Marina Terkourafi. 2008. Toward a unified theory of politeness, impoliteness, and rudeness. Impoliteness in language: studies on its interplay with power in theory and practice. find
  38. Robert Tibshirani. 1994. Regression shrinkage and selection via the lasso. Journal of the Royal Statistical Society, Series B, 58:267–288. find
  39. Linda Tickle-Degnen and Robert Rosenthal. 1990. The nature of rapport and its nonverbal correlates. In Psychological Inquiry. ↗
  40. Kevin E. Vowles and Miles Thompson. 2012. The patient-provider relationship in chronic pain. In Psychiatric Management of Pain. ↗
  41. Erin Walker, Nikol Rummel, and Kenneth R. Koedinger. 2011. Is it feedback relevance or increased accountability that matters? In Proceedings of the 10th International Conference on Computer-Supported Collaborative Learning (CSCL 2011). find
  42. William Yang Wang and Julia Hirschberg. 2011. Detecting levels of interest from spoken dialog with multi-stream prediction feedback and similarity based hierarchical fusion learning. In Proceedings of the 12th annual SIGdial Meeting on Discourse and Dialogue (SIGDIAL 2011), Portland, OR., USA, June. ACL. find
  43. William Yang Wang and Kathleen McKeown. 2010. ”got you!”: Automatic vandalism detection in wikipedia with web-based shallow syntactic-semantic modeling. In Proceedings of the 23rd International Conference on Computational Linguistics (Coling 2010), pages 1146–1154, Beijing, China, August. Coling 2010 Organizing Committee. ↗
  44. William Yang Wang, Fadi Biadsy, Andrew Rosenberg, and Julia Hirschberg. 2012a. Automatic detection of speaker state: Lexical, prosodic, and phonetic approaches to level-of-interest and intoxication classification. Computer Speech & Language. ↗
  45. William Yang Wang, Elijah Mayfield, Suresh Naidu, and Jeremiah Dittmar. 2012b. Historical analysis of legal opinions with a sparse mixed-effects latent variable model. In Proceedings of the 50th Annual Meeting of the Association for Computational Linguistics (ACL 2012). find
  46. Richard J. Watts. 2003. Politeness. Cambridge University Press. find
  47. Hui Zou and Trevor Hastie. 2005. Regularization and variable selection via the elastic net. Journal of the Royal Statistical Society, Series B, 67:301–320. ↗

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/wang-2012-love-ya-jerkface-sparse-log/