Matthew Henderson

Slot labeling framework that achieves a new leap in performance for few-shot slot labeling. Introduces a new pretraining task, pairwise cloze, that allows pre-training all sequence-level layers.

• PDF
• Cite

@inproceedings{henderson-vulic-2021-convex,
    title = "{ConVEx}: Data-Efficient and Few-Shot Slot Labeling",
    author = "Henderson, Matthew  and
      Vuli{'c}, Ivan",
    booktitle = "Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies",
    month = jun,
    year = "2021",
    address = "Online",
    publisher = "Association for Computational Linguistics",
    url = "https://www.aclweb.org/anthology/2021.naacl-main.264",
    pages = "3375--3389",
}

A highly-optimized transformer architecture, pre-trained on response selection. ConveRT achieves state-of-the-art performance across widely established response selection and intent classification tasks.

• PDF
• Cite

@inproceedings{Henderson2020a,
    author      = {Matthew Henderson and I{\~{n}}igo Casanueva and  Nikola Mrk{{s}}i'c and Pei-Hao Su and Tsung-Hsien Wen and Ivan Vuli'c},
    title       = {A Repository of Conversational Datasets},
    year        = {2020},
    url         = {https://arxiv.org/abs/1911.03688},
    booktitle   = {Findings of {EMNLP}},
}

An evaluation of dual sentence encoders such as ConveRT on few-shot intent detection tasks.

• PDF
• Cite

@inproceedings{Casanueva:2020ws,
  author    = {I{\~{n}}igo Casanueva and
               Tadas Temcinas and
               Daniela Gerz and
               Matthew Henderson and
               Ivan Vuli'c},
  title     = {Efficient Intent Detection with Dual Sentence Encoders},
  booktitle   = "Proceedings of the 2nd Workshop on Natural Language Processing for Conversational AI",
  year      = {2020},
  url       = {https://arxiv.org/abs/2003.04807},
  pages = {38--45},
}

A collection of large datasets for Conversational AI, with hundreds of millions of examples, and a standardised evaluation framework.

• PDF
• Cite

@inproceedings{Henderson2019,
    author      = {Matthew Henderson and Pawe{\l} Budzianowski and I{\~{n}}igo Casanueva and Sam Coope and Daniela Gerz and Girish Kumar and Nikola Mrk{{s}}i'c and Georgios Spithourakis and Pei-Hao Su and Ivan Vuli'c and Tsung-Hsien Wen},
    title       = {A Repository of Conversational Datasets},
    year        = {2019},
    month       = {jul},
    note        = {Data available at github.com/PolyAI-LDN/conversational-datasets},
    url         = {https://arxiv.org/abs/1904.06472},
    booktitle   = {Proceedings of the Workshop on {NLP} for Conversational {AI}},
}

An evaluation of fine-tuning techniques for adapting large neural response selection models, trained on large datasets like Reddit, to task-oriented dialogue domains.

• PDF
• Cite

@inproceedings{Henderson2019b,
    author      = {Matthew Henderson and Ivan Vuli'c and Daniela Gerz and I{\~{n}}igo Casanueva and Pawe{\l} Budzianowski and Sam Coope and Georgios Spithourakis and Tsung-Hsien Wen and Nikola Mrk{{s}}i'c and Pei-Hao Su},
    title       = {Training Neural Response Selection for Task-Oriented Dialogue Systems},
    year        = {2019},
    month       = {jul},
    url         = {https://arxiv.org/abs/1906.01543},
    booktitle   = {Proceedings of ACL},
}

This paper presents a question answering system, that selects sentences from product descriptions using a neural-network ranking model. This is trained and evaluated on a dataset of 590K questions and answers from Carousell marketplace conversations.

• PDF
• Cite

@inproceedings{Kumar2018,
    author = {Kumar, Girish and Henderson, Matthew and Chan, Shannon and Nguyen, Hoang and Ngoo, Lucas},
    title = "{Question-Answer Selection in User to User Marketplace Conversations}",
    booktitle = {International Workshop on Spoken Dialog Systems (IWSDS)},
    month = {May},
    year = {2018}
}

The modelling approach behind the Smart Reply feature launched to Gmail, using hierarchical feedforward neural networks, and an efficient dot-product search. The system is 100x times faster than the original LSTM sequence-to-sequence approach and is higher quality.

• PDF
• Cite

@article{Henderson2017,
   author = {Henderson, Matthew and {Al-Rfou}, Rami and Strope, Brian and Sung, Yun-hsuan and
	 Luk{'{a}}cs, L{'{a}}szl{'{o}} and Guo, Ruiqi and Kumar, Sanjiv and Miklos, Balint and
	Kurzweil, Ray},
    title = "{Efficient Natural Language Response Suggestion for Smart Reply}",
  journal = {ArXiv e-prints},
archivePrefix = "arXiv",
   eprint = {1705.00652},
 primaryClass = "cs.CL",
     year = 2017,
    month = may,
   adsurl = {http://adsabs.harvard.edu/abs/2017arXiv170500652H},
}

A review paper on Machine Learning methods for Dialog State Tracking. An invited talk, presented at the first Machine Learning in Spoken Language Processing workshop in Aizuwakamatsu, Japan.

• PDF
• Cite

@inproceedings{Henderson2015,
    title = {Machine Learning for Dialog State Tracking: A Review},
    author  = {Matthew Henderson},
    year  = 2015,
    booktitle = {The First International Workshop on Machine Learning in Spoken Language Processing}
}

This thesis presents how discriminative machine learning methods can be used to develop high accuracy spoken language understanding and state tracking modules for spoken dialogue systems.

• PDF
• Cite

@phdthesis{MattHendersonThesis,
    Title = {{Discriminative Methods for Statistical Spoken Dialogue Systems}},
    Author = {Henderson, Matthew},
    Year = {2015},
    School = {{University of Cambridge}},
}

The Third Dialog State Tracking Challenge, with 7 research teams submitting 26 total entries, evaluated the state of the art in dialog state tracking. The focus was on adapting to expanding conversational domains. Presented as a poster at IEEE SLT 2014, in Lake Tahoe.

• PDF
• Cite

@inproceedings{Henderson2014c,
    author = {Henderson, M. and  Thomson, B. and Williams, J.},
    booktitle = {Proceedings of IEEE Spoken Language Technology},
    title = {{The Third Dialog State Tracking Challenge}},
    year = {2014}
}

This paper presents a method to robustly transfer recurrent neural networks from one dialog domain to another, including a technique for adapting the parameters online with no supervision. This was the top performing tracker in the Third Dialog State Tracking Challenge. Presented as a poster at IEEE SLT 2014, in Lake Tahoe.

• PDF
• Cite

@inproceedings{Henderson2014d,
    author = {Henderson, M. and  Thomson, B. and Young, S. J.},
    booktitle = {Proceedings of IEEE Spoken Language Technology},
    title = {{Robust Dialog State Tracking Using Delexicalised Recurrent Neural Networks and Unsupervised Adaptation}},
    year = {2014}
}

This paper presents the results of the Second Dialog State Tracking Challenge, a research challenge focused on improving the state of the art in tracking the state of spoken dialog systems. In total 9 research teams from across the world competed, with 31 total entries.

• PDF
• Cite

@inproceedings{Henderson2014a,
    author = {Henderson, M. and  Thomson, B. and Williams, J.},
    booktitle = {Proceedings of SIGdial},
    title = {{The Second Dialog State Tracking Challenge}},
    year = {2014}
}

This paper presents a new method for dialog state tracking that relies directly on the words spoken, i.e. the output of a speech recogniser, rather than a semantic representation. The method is validated in the Second Dialog State Tracking Challenge, and found to be one of the top two tracking methods. Nominated for best paper.

• PDF
• Cite

@inproceedings{Henderson2014b,
    author = {Henderson, M. and  Thomson, B. and Young, S. J.},
    booktitle = {Proceedings of SIGdial},
    title = {{Word-based Dialog State Tracking with Recurrent Neural Networks}},
    year = {2014}
}

Inspired by recent promising results using Deep Neural Networks in speech applications, this paper shows how to apply DNNs to Dialog State Tracking. This was presented as a poster at SIGdial and was entered in the Dialog State Tracking Challenge.

• PDF
• Cite

@inproceedings{Henderson2013a,
    author = { Henderson, M. and  Thomson, B. and Young, S. J.},
    booktitle = {Proceedings of SIGdial},
    title = {{Deep Neural Network Approach for the Dialog State Tracking Challenge}},
    year = {2013}
}

This paper presents a new form of semantic decoding which uses the entire distribution of speech recognition hypotheses to infer the meaning of an utterance. This was presented as a poster at SLT. The dataset used for training the decoder and for the offline evaluation is freely available: In Car SLU Corpus.

• PDF
• Cite

@inproceedings{Henderson2012a,
    author = {Henderson, Matthew and Ga{s}i'{c}, Milica and Thomson, Blaise and Tsiakoulis, Pirros and Yu, Kai and Young, Steve},
    booktitle = {Spoken Language Technology Workshop, 2012. IEEE},
    title = {{Discriminative Spoken Language Understanding Using Word Confusion Networks}},
    year = {2012}
}

A paper stemming from work done in my MSc evaluating a set of strategies a conversational (rather than task-oriented) dialogue system can use to recover from situations where it is unable to understand what the user has asked. The dialogue system concerned is for a robot tourguide which shows visitors around the Edinburgh Informatics Forum. This was an oral presentation at Semdial.

• PDF
• Cite

@inproceedings{Henderson2012a,
    author = {Henderson, Matthew and Matheson, Colin and Oberlander, Jon},
    booktitle = {Workshop on the Semantics and Pragmatics of Dialogue},
    title = {{Recovering from Non-Understanding Errors in a Conversational Dialogue System}},
    year = {2012}
}