Dialogue act modeling for automatic tagging and recognition of conversational speech
By Mark H. Burstein, Jerry R. Hobbs, Ora Lassila, David L. Martin, Drew V. McDermott, Sheila A. McIlraith, S. Narayanan, Massimo Paolucci, Terry R. Payne, Katia P. Sycara. International Conference on Computational Logic, 2000
2,914 citations recorded by the source corpus.
Abstract We describe a statistical approach for modeling dialogue acts in conversational speech, i.e., speech-act-like units such as STATEMENT, Question, BACKCHANNEL, Agreement, Disagreement, and Apology. Our model detects and predicts dialogue acts based on lexical, collocational, and prosodic cues, as well as on the discourse coherence of the dialogue act sequence. The dialogue model is based on treating the discourse structure of a conversation as a hidden Markov model and the individual dialogue acts as observations emanating from the model states. Constraints on the likely sequence of dialogue acts are modeled via a dialogue act n-gram. The statistical dialogue grammar is combined with word n-grams, decision trees, and neural networks modeling the idiosyncratic lexical and prosodic manifestations of each dialogue act. We develop a probabilistic integration of speech recognition with dialogue modeling, to improve both speech recognition and dialogue act classification accuracy. Models are trained and evaluated using a large hand-labeled database of 1,155 conversations from the Switchboard corpus of spontaneous human-to-human telephone speech. We achieved good dialogue act labeling accuracy (65% based on errorful, automatically recognized words and prosody, and 71% based on word transcripts, compared to a chance baseline accuracy of 35% and human accuracy of 84%) and a small reduction in word recognition error.
Subfields: Artificial Intelligence, Language and Linguistics
View Paper: https://www.semanticscholar.org/paper/22d45dadde6b5837eff11dc031045754bc5901c3
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Artificial Intelligence
- Mohit Bansal
- Yoshua Bengio
- Jianfeng Gao
- Percy Liang
- Luke S. Zettlemoyer
- Noah A. Smith
- Kyunghyun Cho
- Quoc V. Le
In Language and Linguistics
- Phil Blunsom
- Daniel Klein
- David Leo Wright Hall
- Nal Kalchbrenner
- Sam Thomson
- Dawn Song
- Dan Roth
- A. P. Sarath Chandar
Citations in the corpus
- On lattices, learning with errors, random linear codes, and cryptography (2009, background): “A much more significant improvement is described by Peikert, Vaikuntanathan, and Waters in [34].”
- Advanced and authenticated marking schemes for IP traceback (2001, methodology): “Dean, Franklin and Stubblefield [21] propose a nice alternat ive marking scheme using noisy polynomial reconstruction.”
- Change-point monitoring for the detection of DoS attacks (2004, background): “To identify the compromised end-hosts that directly generate flooding packets and the network path that these packets take, various traceback techniques [3], [7], [11], [47], [50], [51] have been proposed.”
- Recurrent Convolutional Neural Networks for Discourse Compositionality (2013, positive): “In addition, the recognition of dialogue acts within dialogues has largely been treated in non-compositional ways by way of language models coupled to hidden Markov sequence models (Stolcke et al., 2000).”
- Trapdoors for hard lattices and new cryptographic constructions (2008, methodology): “We thank Vadim Lyubashevsky for comments on an earlier draft and for bringing Klein’s paper [32] to our attention, Phong Nguyen and Oded Regev for helpful comments and advice, and Brent Waters for useful observations about our IBE system.”
- Security Against Covert Adversaries: Efficient Protocols for Realistic Adversaries (2007, methodology): “A highly efficient protocol for this task, with only a constant number of exponentiations per transfer, appears in [29].”
- Lattice-Based Cryptography (2006, methodology): “By following an argument similar to the one in [71, 65], one can show that a sufficient condition for this is that the number of possibilities for a is much larger than the number of elements in our range, i.”
- SIFF: a stateless Internet flow filter to mitigate DDoS flooding attacks (2004, background): “Several researchers propose to embed traceback information within the IP packet [2, 12, 19, 26, 35, 38, 43].”
- Pi: a path identification mechanism to defend against DDoS attacks (2003, background): “Dean, Franklin, and Stubblefield propose using algebraic codes to encode the upstream router path for IP traceback [8, 9].”
- Bonsai Trees, or How to Delegate a Lattice Basis (2010, background): “A similar issue arises in other recent cryptographic schemes [GPV08, PVW08, Pei09b], but in a simpler setting involving only a single lattice and short basis (not a hierarchy).”
- Controlled user evaluations of information visualization interfaces for text retrieval: Literature review and meta-analysis (2008, neutral): “The latter approach has been widely studied by researchers working in the areas of interactive information retrieval (IR) and interactive question answering (QA; Sitter & Stein, 1992; Spink, Goodrum, & Robins, 1998; Ellis et al., 2002; Wu & Liu, 2003; Small, Strzalkowski, Liu, Shimizu, & Yamrom, 2”
- A Deep Reinforcement Learning Chatbot (2017, positive): “Dialogue act: A one-hot vector, indicating whether the last user utterance’s dialogue is a request, a question, a statement or contains profanity (Stolcke et al. 2000).”
- Using Linguistic Cues for the Automatic Recognition of Personality in Conversation and Text (2007, neutral): “This could possibly arise from the small size of the datasets, or from the relatively low accuracy of our hand-crafted automatic tagger, compared to other work using supervised learning methods (Stolcke et al., 2000; Webb et al., 2005).”
- Lossy trapdoor functions and their applications (2008, background): “We also note that while NIZK proofs for certain lattice problems are known [51], they do not appear to suffice for CCA security.”
- Practical Covertly Secure MPC for Dishonest Majority - Or: Breaking the SPDZ Limits (2013, neutral): “It was proven for the (standard) LWE setting with n = 3 in [23]; here we need a version with ring-LWE for any n.”
- A Survey on Conversational Recommender Systems (2020, neutral): “ding components for this task, e.g., [54]. In principle, intentdetectioncanbeseenasaclassificationtask(dialogueactclassification),whereuserutterances are assigned to one or multiple intent categories [131]. Named entity recognition aims at the identification of entities in a given utterance into pre”
- Efficient Public Key Encryption Based on Ideal Lattices (2009, background): “Most of the cryptosystems based on general lattices [33, 30, 31, 9, 28] rely on the average-case hardness of the Learning With Errors (LWE) problem introduced in [33].”