Elamur Research⌕
☰
home / papers / 2018 / Detecting Institutional Dialog Acts in Police Traffic Stops

Detecting Institutional Dialog Acts in Police Traffic Stops

Paper · 2018 · Vinodkumar Prabhakaran, Camilla Griffiths, Hang Su et al.

Summary

Police body-worn cameras capture large amounts of traffic-stop conversation, but most of it is never analysed. The authors treat these encounters as institutional talk, where the setting determines which conversational moves are expected. They design a scheme of officer-side institutional dialog acts, including giving the Reason for the stop, Searches, Offering Help, greetings, sanctions and closings.

To build the scheme, they first ran an unsupervised mixed membership Markov model to find recurring topic segments, then turned those segments into labeled categories. Using the Voigt et al. (2017) collection of 981 Oakland Police Department stops, they trained supervised taggers with lexical, positional, contextual and institution-specific features. They also built speech recognition and segmentation components so the system can run on raw camera audio.

On transcripts, the tagger reaches about 78% F-score per turn and 89% per stop. From raw audio it reaches 81% per stop. The resulting act sequences show whether officers follow procedural-justice norms such as introducing themselves, stating the reason for the stop, and asking consent before a search.

Key points

  • Data: 981 Oakland PD vehicle stops from April 2014, totalling 35 hours of audio, hand-transcribed into 94K speaker turns and 757K words. About 7K sentences were annotated, which the authors consider too few to train in-domain recurrent taggers.
  • The typology was seeded by a 10-topic mixed membership Markov model (Paul, 2012) trained on 541 transcripts. It segmented stops into reason, documents, driver details, sanction and closing.
  • Results: about 78% F-score at turn level and 89% at stop level from gold transcripts. From raw audio, results are 60% at turn level and 81% at stop level, with higher accuracy for crucial acts like the Reason.
  • Annotation covers only officer speech, because the officer holds conversational initiative. Raw annotations are not released for sensitivity reasons, but the labeling scheme and annotation manual are.
  • Stated limitations: the data comes from a single department, the models may encode dataset biases, driver audio is poor, and night-time video is mostly unusable.

Interesting ideas

  • I1. Generic acts like "request" are too coarse for this setting, so the authors define acts specific to the institution, such as "requests documentation", that link a speech act to its topical purpose.
  • I2. The label inventory was bootstrapped from unsupervised topic segmentation and then refined by humans, rather than invented from scratch.
  • I3. Evaluation runs at two granularities: per turn, and per whole conversation ("does this stop contain a Reason?"). The coarser question can be answered reliably even when turn-level tagging is noisy.
  • I4. The tagger uses positional and contextual features, such as relative place in the dialog and the preceding utterance, alongside the words themselves. Act identity depends strongly on sequence position.
  • I5. An act can be performed jointly across speakers: an officer's "yes" to "did you stop me for the red light?" counts as giving the reason.

For Mova

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

  • I1 (frames): crate prag could add domain-specialised speech-act subtypes, e.g. a teacher asking a question in a school text, sitting above the generic request/assert layer.
  • I2 (learn): unsupervised clustering of frame sequences over a corpus could propose candidate act or frame categories for an LLM teacher or human to name and validate.
  • I3 (answer): QA on fables could score both sentence-level frame accuracy and document-level questions like "does the fox ever lie?", showing where errors actually matter.
  • I4 (frames): speech-act classification in prag could explicitly use discourse position and the previous turn's act, keeping the rule explainable instead of purely lexical.
  • I5 (frames): prag should let answers to polar questions inherit the propositional content of the question, so the world records the confirmed fact.

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

 Institutions

🎓Stanford UniversityStanford · United States310 works · 284,480 citations🎓University of California, BerkeleyBerkeley · United States127 works · 212,523 citations🔬International Computer Science InstituteBerkeley · United States12 works · 8,813 citations

📚 Shelves 6The 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 worksSocial Sciences · 744 worksPolitical Science and International Relations · 72 works

🧭 Topics 2Research subjects and languages the work is about.Dialogue systems · 401 worksSpeech act theory · 160 works

Cite

Cite as Vinodkumar Prabhakaran, Camilla Griffiths, Hang Su, Prateek Verma, Nelson Morgan, Jennifer L. Eberhardt and Dan Jurafsky. 2018. Detecting Institutional Dialog Acts in Police Traffic Stops. Transactions of the Association for Computational Linguistics.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.1162/tacl_a_00031

@article{prabhakaran2018detecting,
  title = {{Detecting Institutional Dialog Acts in Police Traffic Stops}},
  author = {Vinodkumar Prabhakaran, Camilla Griffiths, Hang Su, Prateek Verma, Nelson Morgan, Jennifer L. Eberhardt and Dan Jurafsky},
  journal = {Transactions of the Association for Computational Linguistics},
  year = {2018},
  doi = {10.1162/tacl\_a\_00031},
  url = {https://doi.org/10.1162/tacl_a_00031}
}

References

61 references, 12 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Tim Althoff, Kevin Clark, and Jure Leskovec. 2016. Large-scale analysis of counseling conversations: An application of natural language processing to mental health. Transactions of the Association for Computational Linguistics, 4:463–476. ↗
  2. Jeremy Ang, Yang Liu, and Elizabeth Shriberg. 2005. Automatic dialog act segmentation and classification in multiparty meetings. In Proceedings of IEEE International Conference on Acoustics, Speech, and Signal Processing, volume 1, pages 1061–1064. IEEE. find
  3. J. Maxwell Atkinson and Paul Drew. 1979. Order in Court. Springer. ↗
  4. John Langshaw Austin. 1975. How To Do Things With Words. Oxford University Press. ↗
  5. Srinivas Bangalore, Giuseppe Di Fabbrizio, and Amanda Stent. 2006. Learning the structure of task-driven human-human dialogs. In Proceedings of the 21st International Conference on Computational Linguistics and 44th Annual Meeting of the Association for Computational Linguistics, pages 201–208. Association for Computational Linguistics. find
  6. David M. Blei and Pedro J. Moreno. 2001. Topic segmentation with an aspect hidden Markov model. In Proceedings of the 24th Annual International ACM SI-GIR conference on Research and Development in Information Retrieval, pages 343–348. ACM. find
  7. Rod K. Brunson. 2007. “Police Don’t Like Black people”: African-American Young Men’s Accumulated Police Experiences. Criminology & Public Policy, 6(1):71–101. find
  8. Muthu Kumar Chandrasekaran, Carrie Epp, Min-Yen Kan, and Diane Litman. 2017. Using discourse signals for robust instructor intervention prediction. In Proceedings of the AAAI Conference on Artificial Intelligence. ↗
  9. Christopher Cieri, David Miller, and Kevin Walker. 2004. The Fisher corpus: A resource for the next generations of speech-to-text. In Proceedings of the Fourth International Conference on Language Resources and Evaluation (LREC’04). European Language Resources Association (ELRA).
  10. William W. Cohen, Vitor R. Carvalho, and Tom M. Mitchell. 2004. Learning to classify email into “speech acts”. In Proceedings of the Conference on Empirical Methods in Natural Language Processing, volume 4, pages 309–316. Association for Computational Linguistics.
  11. Corinna Cortes and Vladimir Vapnik. 1995. Support-vector networks. Machine learning, 20(3):273–297. ↗
  12. Justine Coupland, Nikolas Coupland, and Howard Giles. 1991. Accommodation theory, communication, context and consequences. Contexts of Accommodation, pages 1–68. find
  13. Jacob Eisenstein and Regina Barzilay. 2008. Bayesian unsupervised topic segmentation. In Proceedings of the Conference on Empirical Methods in Natural Language Processing, pages 334–343. Association for Computational Linguistics.
  14. Robin S. Engel. 2005. Citizens’ perceptions of distributive and procedural injustice during traffic stops with police. Journal of Research in Crime and Delinquency, 42(4):445–481. ↗
  15. Charles R. Epp, Steven Maynard-Moody, and Donald P. Haider-Markel. 2014. Pulled Over: How Police Stops Define Race and Citizenship. University of Chicago Press. find
  16. Mark J. F. Gales. 1998. Maximum likelihood linear transformations for HMM-based speech recognition. Computer Speech & Language, 12(2):75–98. ↗
  17. Felix A. Gers, Nicol N. Schraudolph, and Jürgen Schmidhuber. 2002. Learning precise timing with LSTM recurrent networks. Journal of Machine Learning Research, 3(Aug):115–143. find
  18. Pierre Geurts, Damien Ernst, and Louis Wehenkel. 2006. Extremely randomized trees. Machine Learning, 63(1):3–42. ↗
  19. Howard Giles, Jennifer Fortman, René Dailey, Valerie Barker, Christopher Hajek, Michelle Chernikoff Anderson, and Nicholas O. Rule. 2006. Communication accommodation: Law enforcement and the public. Applied Interpersonal Communication Matters: Family, Health, and Community Relations, 5:241–269. find
  20. Howard Giles, Christopher Hajek, Valerie Barker, Mei-Chen Lin, Yan Bing Zhang, Mary Lee Hummert, and Michelle C. Anderson. 2007. Accommodation and institutional talk: Communicative dimensions of po-licecivilian interactions. In Language, Discourse and Social Psychology, pages 131–159. Springer. find
  21. Augusto Gnisci. 2005. Sequential strategies of accommodation: A new method in courtroom. British Journal of Social Psychology, 44(4):621–643. ↗
  22. John J. Godfrey and Edward Holliman. 1997. Switchboard-1 release 2. Linguistic Data Consortium, Philadelphia, 926:927. find
  23. Alex Graves and Navdeep Jaitly. 2014. Towards end-to-end speech recognition with recurrent neural networks. In International Conference on Machine Learning, pages 1764–1772. find
  24. Alex Graves, Abdel-rahman Mohamed, and Geoffrey Hinton. 2013. Speech recognition with deep recurrent neural networks. In 2013 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pages 6645–6649. IEEE. ↗
  25. Barbara J. Grosz. 1977. The representation and use of focus in dialogue understanding. Technical report, SRI International Menlo Park United States. find
  26. John Heritage. 2005. Conversation analysis and institutional talk. Handbook of Language and Social Interaction, pages 103–147. find
  27. Daniel J. Hopkins and Gary King. 2010. A method of automated nonparametric content analysis for social science. American Journal of Political Science, 54(1):229–247. ↗
  28. Daniel Jurafsky, Rebecca Bates, Noah Coccaro, Rachel Martin, Marie Meteer, Klaus Ries, Elizabeth Shriberg, Andreas Stolcke, Paul Taylor, and Carol Van Ess-Dykema. 1997. Automatic detection of discourse structure for speech recognition and understanding. In Proceedings of the 1997 IEEE Workshop on Automatic Speech Recognition and Understanding, pages 88–95. IEEE. ↗
  29. Hamed Khanpour, Nishitha Guntakandla, and Rodney Nielsen. 2016. Dialogue act classification in domain-independent conversations using a deep recurrent neural network. In Proceedings of the 26th International Conference on Computational Linguistics: Technical Papers, pages 2012–2021.
  30. Su Nam Kim, Lawrence Cavedon, and Timothy Baldwin. 2010. Classifying dialogue acts in one-on-one live chats. In Proceedings of the 2010 Conference on Empirical Methods in Natural Language Processing, pages 862–871. Association for Computational Linguistics.
  31. Su Nam Kim, Lawrence Cavedon, and Timothy Baldwin. 2012. Classifying dialogue acts in multi-party live chats. In Proceedings of the 26th Pacific Asia Conference on Language, Information, and Computation, pages 463–472.
  32. Yoon Kim. 2014. Convolutional neural networks for sentence classification. In Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 1746–1751. Association for Computational Linguistics.
  33. Reinhard Kneser and Hermann Ney. 1995. Improved backing-off for M-gram language modeling. In Proceedings of the International Conference on Acoustics, Speech, and Signal Processing, volume 1, pages 181– 184. IEEE. find
  34. Tom Ko, Vijayaditya Peddinti, Daniel Povey, and Sanjeev Khudanpur. 2015. Audio augmentation for speech recognition. In Proceedings of Sixteenth Annual Conference of the International Speech Communication Association (INTERSPEECH), pages 3586–3589. find
  35. Lynn Langton and Matthew R. Durose. 2013. Police behavior during traffic and street stops, 2011. US Department of Justice, Office of Justice Programs, Bureau of Justice Statistics Washington, DC. find
  36. Wei Li and Yunfang Wu. 2016. Multi-level gated recurrent neural network for dialog act classification. In Proceedings of the 26th International Conference on Computational Linguistics: Technical papers, pages 1970–1979.
  37. Yang Liu, Kun Han, Zhao Tan, and Yun Lei. 2017. Using context information for dialog act classification in DNN framework. In Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing, pages 2170–2178.
  38. Richard J. Lundman and Robert L. Kaufman. 2003. Driving while black: Effects of race, ethnicity, and gender on citizen self-reports of traffic stops and police actions. Criminology, 41(1):195–220. ↗
  39. Abdel-rahman Mohamed, Frank Seide, Dong Yu, Jasha Droppo, Andreas Stoicke, Geoffrey Zweig, and Gerald Penn. 2015. Deep bi-directional recurrent networks over spectral windows. In Proceedings of 2015 IEEE Workshop on Automatic Speech Recognition and Understanding (ASRU), pages 78–83. IEEE. find
  40. Viet-An Nguyen, Jordan Boyd-Graber, and Philip Resnik. 2012. SITS: A hierarchical nonparametric model using speaker identity for topic segmentation in multiparty conversations. In Proceedings of the 50th Annual Meeting of the Association for Computational Linguistics: Long Papers-Volume 1, pages 78–87. Association for Computational Linguistics. find
  41. Adinoyi Omuya, Vinodkumar Prabhakaran, and Owen Rambow. 2013. Improving the quality of minority class identification in dialog act tagging. In Proceedings of the 2013 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pages 802–807. ↗
  42. Michael J. Paul. 2012. Mixed membership Markov models for unsupervised conversation modeling. In Proceedings of the 2012 Joint Conference on Empirical Methods in Natural Language Processing and Computational Natural Language Learning, pages 94–104. Association for Computational Linguistics. find
  43. Daniel Povey, Arnab Ghoshal, Gilles Boulianne, Lukas Burget, Ondrej Glembek, Nagendra Goel, Mirko Hannemann, Petr Motlı́ček, Yanmin Qian, Petr Schwarz, Jan Silovský, Georg Stemmer, and Karel Veselý. 2011. The Kaldi speech recognition toolkit. In IEEE 2011 Workshop on Automatic Speech Recognition and Understanding. IEEE Signal Processing Society. find
  44. Vinodkumar Prabhakaran, Owen Rambow, and Mona Diab. 2012. Predicting overt display of power in written dialogs. In Proceedings of the 2012 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pages 518–522. Association for Computational Linguistics. find
  45. Vinodkumar Prabhakaran, Emily E. Reid, and Owen Rambow. 2014. Gender and power: How gender and gender environment affect manifestations of power. In Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 1965–1976. Association for Computational Linguistics. find
  46. Charles H. Ramsey and Laurie O. Robinson. 2015. Final report of the President’s task force on 21st century policing. Washington, DC: Office of Community Oriented Policing Services. find
  47. Alan Ritter, Colin Cherry, and Bill Dolan. 2010. Unsupervised modeling of Twitter conversations. In Human Language Technologies: The 2010 Annual Conference of the North American Chapter of the Association for Computational Linguistics, pages 172–180. Association for Computational Linguistics.
  48. Harvey Sacks, Emanuel A. Schegloff, and Gail Jefferson. 1974. A simplest systematics for the organization of turn-taking for conversation. Language, pages 696– 735. ↗
  49. Haşim Sak, Andrew Senior, and Françoise Beaufays. 2014. Long short-term memory recurrent neural network architectures for large scale acoustic modeling. In Fifteenth Annual Conference of the International Speech Communication Association. find
  50. Ruslan Salakhutdinov, Andriy Mnih, and Geoffrey Hinton. 2007. Restricted Boltzmann machines for collaborative filtering. In Proceedings of the 24th International Conference on Machine Learning, pages 791– 798. ACM. find
  51. Emanuel A. Schegloff and Harvey Sacks. 1973. Opening up closings. Semiotica, 8(4):289–327. ↗
  52. Emanuel A. Schegloff. 1979. Identification and recognition in telephone conversation openings. Everyday-Language: Studies in Ethnomethodology, New York, Irvington, pages 23–78. find
  53. Nitish Srivastava, Geoffrey Hinton, Alex Krizhevsky, Ilya Sutskever, and Ruslan Salakhutdinov. 2014. find
  54. Dropout: A simple way to prevent neural networks from overfitting. The Journal of Machine Learning Research, 15(1):1929–1958. find
  55. Andreas Stolcke, Klaus Ries, Noah Coccaro, Elizabeth Shriberg, Rebecca Bates, Daniel Jurafsky, Paul Taylor, Rachel Martin, Carol Van Ess-Dykema, and Marie Meteer. 2006. Dialogue act modeling for automatic tagging and recognition of conversational speech. Dialogue, 26(3). find
  56. Prateek Verma, T. P. Vinutha, Parthe Pandit, and Preeti Rao. 2015. Structural segmentation of Hindustani concert audio with posterior features. In Proceedings of the 2015 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pages 136–140. IEEE. find
  57. Karel Veselý, Mirko Hannemann, and Lukas Burget. 2013. Semi-supervised training of deep neural networks. In Proceedings of the 2013 IEEE Workshop on Automatic Speech Recognition and Understanding (ASRU). IEEE. find
  58. Rob Voigt, Nicholas P. Camp, Vinodkumar Prabhakaran, William L. Hamilton, Rebecca C. Hetey, Camilla M. Griffiths, David Jurgens, Dan Jurafsky, and Jennifer L. Eberhardt. 2017. Language from police body camera footage shows racial disparities in officer respect. Proceedings of the National Academy of Sciences, 114(25):6521–6526. ↗
  59. Marilyn R. Whalen and Don H. Zimmerman. 1987. Sequential and institutional contexts in calls for help. Social Psychology Quarterly, pages 172–185. ↗
  60. Gethin Williams and Daniel P.W. Ellis. 1999. Speech/music discrimination based on posterior probability features. In Proceedings of the Sixth European Conference on Speech Communication and Technology. ↗
  61. Diyi Yang, Tanmay Sinha, David Adamson, and Carolyn P. Rose. 2013. Turn on, tune in, drop out: Anticipating student dropouts in massive open online courses. In Proceedings of the 2013 NIPS Data-Driven Education Workshop, volume 10, pages 13–20. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/prabhakaran-2018-detecting-institutional-dialog-acts-police/