Publications
Search
1.
Georgila, Kallirroi
Comparing Pre-Trained Embeddings and Domain-Independent Features for Regression-Based Evaluation of Task-Oriented Dialogue Systems Proceedings Article
In: Proceedings of the 25th Annual Meeting of the Special Interest Group on Discourse and Dialogue, pp. 610–623, Association for Computational Linguistics, Kyoto, Japan, 2024.
@inproceedings{georgila_comparing_2024,
title = {Comparing Pre-Trained Embeddings and Domain-Independent Features for Regression-Based Evaluation of Task-Oriented Dialogue Systems},
author = {Kallirroi Georgila},
url = {https://aclanthology.org/2024.sigdial-1.52},
doi = {10.18653/v1/2024.sigdial-1.52},
year = {2024},
date = {2024-09-01},
urldate = {2024-10-15},
booktitle = {Proceedings of the 25th Annual Meeting of the Special Interest Group on Discourse and Dialogue},
pages = {610–623},
publisher = {Association for Computational Linguistics},
address = {Kyoto, Japan},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
2.
Frummet, Alexander; Speggiorin, Alessandro; Elsweiler, David; Leuski, Anton; Dalton, Jeff
Cooking with Conversation: Enhancing User Engagement and Learning with a Knowledge-Enhancing Assistant Journal Article
In: ACM Trans. Inf. Syst., pp. 3649500, 2024, ISSN: 1046-8188, 1558-2868.
@article{frummet_cooking_2024,
title = {Cooking with Conversation: Enhancing User Engagement and Learning with a Knowledge-Enhancing Assistant},
author = {Alexander Frummet and Alessandro Speggiorin and David Elsweiler and Anton Leuski and Jeff Dalton},
url = {https://dl.acm.org/doi/10.1145/3649500},
doi = {10.1145/3649500},
issn = {1046-8188, 1558-2868},
year = {2024},
date = {2024-03-01},
urldate = {2024-04-16},
journal = {ACM Trans. Inf. Syst.},
pages = {3649500},
abstract = {We present two empirical studies to investigate users’ expectations and behaviours when using digital assistants, such as Alexa and Google Home, in a kitchen context: First, a survey (N=200) queries participants on their expectations for the kinds of information that such systems should be able to provide. While consensus exists on expecting information about cooking steps and processes, younger participants who enjoy cooking express a higher likelihood of expecting details on food history or the science of cooking. In a follow-up Wizard-of-Oz study (N = 48), users were guided through the steps of a recipe either by an
active
wizard that alerted participants to information it could provide or a
passive
wizard who only answered questions that were provided by the user. The
active
policy led to almost double the number of conversational utterances and 1.5 times more knowledge-related user questions compared to the
passive
policy. Also, it resulted in 1.7 times more knowledge communicated than the
passive
policy. We discuss the findings in the context of related work and reveal implications for the design and use of such assistants for cooking and other purposes such as DIY and craft tasks, as well as the lessons we learned for evaluating such systems.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
We present two empirical studies to investigate users’ expectations and behaviours when using digital assistants, such as Alexa and Google Home, in a kitchen context: First, a survey (N=200) queries participants on their expectations for the kinds of information that such systems should be able to provide. While consensus exists on expecting information about cooking steps and processes, younger participants who enjoy cooking express a higher likelihood of expecting details on food history or the science of cooking. In a follow-up Wizard-of-Oz study (N = 48), users were guided through the steps of a recipe either by an
active
wizard that alerted participants to information it could provide or a
passive
wizard who only answered questions that were provided by the user. The
active
policy led to almost double the number of conversational utterances and 1.5 times more knowledge-related user questions compared to the
passive
policy. Also, it resulted in 1.7 times more knowledge communicated than the
passive
policy. We discuss the findings in the context of related work and reveal implications for the design and use of such assistants for cooking and other purposes such as DIY and craft tasks, as well as the lessons we learned for evaluating such systems.
active
wizard that alerted participants to information it could provide or a
passive
wizard who only answered questions that were provided by the user. The
active
policy led to almost double the number of conversational utterances and 1.5 times more knowledge-related user questions compared to the
passive
policy. Also, it resulted in 1.7 times more knowledge communicated than the
passive
policy. We discuss the findings in the context of related work and reveal implications for the design and use of such assistants for cooking and other purposes such as DIY and craft tasks, as well as the lessons we learned for evaluating such systems.
3.
Lukin, Stephanie M.; Pollard, Kimberly A.; Bonial, Claire; Hudson, Taylor; Arstein, Ron; Voss, Clare; Traum, David
Navigating to Success in Multi-Modal Human-Robot Collaboration: Analysis and Corpus Release Miscellaneous
2023, (arXiv:2310.17568 [cs]).
@misc{lukin_navigating_2023,
title = {Navigating to Success in Multi-Modal Human-Robot Collaboration: Analysis and Corpus Release},
author = {Stephanie M. Lukin and Kimberly A. Pollard and Claire Bonial and Taylor Hudson and Ron Arstein and Clare Voss and David Traum},
url = {http://arxiv.org/abs/2310.17568},
year = {2023},
date = {2023-10-01},
urldate = {2023-12-07},
publisher = {arXiv},
abstract = {Human-guided robotic exploration is a useful approach to gathering information at remote locations, especially those that might be too risky, inhospitable, or inaccessible for humans. Maintaining common ground between the remotely-located partners is a challenge, one that can be facilitated by multi-modal communication. In this paper, we explore how participants utilized multiple modalities to investigate a remote location with the help of a robotic partner. Participants issued spoken natural language instructions and received from the robot: text-based feedback, continuous 2D LIDAR mapping, and upon-request static photographs. We noticed that different strategies were adopted in terms of use of the modalities, and hypothesize that these differences may be correlated with success at several exploration sub-tasks. We found that requesting photos may have improved the identification and counting of some key entities (doorways in particular) and that this strategy did not hinder the amount of overall area exploration. Future work with larger samples may reveal the effects of more nuanced photo and dialogue strategies, which can inform the training of robotic agents. Additionally, we announce the release of our unique multi-modal corpus of human-robot communication in an exploration context: SCOUT, the Situated Corpus on Understanding Transactions.},
note = {arXiv:2310.17568 [cs]},
keywords = {},
pubstate = {published},
tppubtype = {misc}
}
Human-guided robotic exploration is a useful approach to gathering information at remote locations, especially those that might be too risky, inhospitable, or inaccessible for humans. Maintaining common ground between the remotely-located partners is a challenge, one that can be facilitated by multi-modal communication. In this paper, we explore how participants utilized multiple modalities to investigate a remote location with the help of a robotic partner. Participants issued spoken natural language instructions and received from the robot: text-based feedback, continuous 2D LIDAR mapping, and upon-request static photographs. We noticed that different strategies were adopted in terms of use of the modalities, and hypothesize that these differences may be correlated with success at several exploration sub-tasks. We found that requesting photos may have improved the identification and counting of some key entities (doorways in particular) and that this strategy did not hinder the amount of overall area exploration. Future work with larger samples may reveal the effects of more nuanced photo and dialogue strategies, which can inform the training of robotic agents. Additionally, we announce the release of our unique multi-modal corpus of human-robot communication in an exploration context: SCOUT, the Situated Corpus on Understanding Transactions.
4.
Gilani, Setareh Nasihati; Pollard, Kimberly; Traum, David
Multimodal Prediction of User's Performance in High-Stress Dialogue Interactions Proceedings Article
In: International Cconference on Multimodal Interaction, pp. 71–75, ACM, Paris France, 2023, ISBN: 979-8-4007-0321-8.
@inproceedings{nasihati_gilani_multimodal_2023,
title = {Multimodal Prediction of User's Performance in High-Stress Dialogue Interactions},
author = {Setareh Nasihati Gilani and Kimberly Pollard and David Traum},
url = {https://dl.acm.org/doi/10.1145/3610661.3617166},
doi = {10.1145/3610661.3617166},
isbn = {979-8-4007-0321-8},
year = {2023},
date = {2023-10-01},
urldate = {2023-12-07},
booktitle = {International Cconference on Multimodal Interaction},
pages = {71–75},
publisher = {ACM},
address = {Paris France},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
5.
Gainer, Alesia; Aptaker, Allison; Artstein, Ron; Cobbins, David; Core, Mark; Gordon, Carla; Leuski, Anton; Li, Zongjian; Merchant, Chirag; Nelson, David; Soleymani, Mohammad; Traum, David
DIVIS: Digital Interactive Victim Intake Simulator Proceedings Article
In: Proceedings of the 23rd ACM International Conference on Intelligent Virtual Agents, pp. 1–2, ACM, Würzburg Germany, 2023, ISBN: 978-1-4503-9994-4.
@inproceedings{gainer_divis_2023,
title = {DIVIS: Digital Interactive Victim Intake Simulator},
author = {Alesia Gainer and Allison Aptaker and Ron Artstein and David Cobbins and Mark Core and Carla Gordon and Anton Leuski and Zongjian Li and Chirag Merchant and David Nelson and Mohammad Soleymani and David Traum},
url = {https://dl.acm.org/doi/10.1145/3570945.3607328},
doi = {10.1145/3570945.3607328},
isbn = {978-1-4503-9994-4},
year = {2023},
date = {2023-09-01},
urldate = {2024-02-20},
booktitle = {Proceedings of the 23rd ACM International Conference on Intelligent Virtual Agents},
pages = {1–2},
publisher = {ACM},
address = {Würzburg Germany},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
6.
Pal, Debaditya; Leuski, Anton; Traum, David
Comparing Statistical Models for Retrieval based Question-answering Dialogue: BERT vs Relevance Models Journal Article
In: FLAIRS, vol. 36, 2023, ISSN: 2334-0762.
@article{pal_comparing_2023,
title = {Comparing Statistical Models for Retrieval based Question-answering Dialogue: BERT vs Relevance Models},
author = {Debaditya Pal and Anton Leuski and David Traum},
url = {https://journals.flvc.org/FLAIRS/article/view/133386},
doi = {10.32473/flairs.36.133386},
issn = {2334-0762},
year = {2023},
date = {2023-05-01},
urldate = {2023-08-23},
journal = {FLAIRS},
volume = {36},
abstract = {In this paper, we compare the performance of four models in a retrieval based question answering dialogue task on two moderately sized corpora (textasciitilde 10,000 utterances). One model is a statistical model and uses cross language relevance while the others are deep neural networks utilizing the BERT architecture along with different retrieval methods. The statistical model has previously outperformed LSTM based neural networks in a similar task whereas BERT has been proven to perform well on a variety of NLP tasks, achieving state-of-the-art results in many of them. Results show that the statistical cross language relevance model outperforms the BERT based architectures in learning question-answer mappings. BERT achieves better results by mapping new questions to existing questions.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
In this paper, we compare the performance of four models in a retrieval based question answering dialogue task on two moderately sized corpora (textasciitilde 10,000 utterances). One model is a statistical model and uses cross language relevance while the others are deep neural networks utilizing the BERT architecture along with different retrieval methods. The statistical model has previously outperformed LSTM based neural networks in a similar task whereas BERT has been proven to perform well on a variety of NLP tasks, achieving state-of-the-art results in many of them. Results show that the statistical cross language relevance model outperforms the BERT based architectures in learning question-answer mappings. BERT achieves better results by mapping new questions to existing questions.
7.
Georgila, Kallirroi
Considerations for Child Speech Synthesis for Dialogue Systems Proceedings Article
In: Los Angeles, CA, 2023.
@inproceedings{georgila_considerations_2023,
title = {Considerations for Child Speech Synthesis for Dialogue Systems},
author = {Kallirroi Georgila},
url = {https://kgeorgila.github.io/publications/georgila_aiaic23.pdf},
year = {2023},
date = {2023-03-01},
address = {Los Angeles, CA},
abstract = {We present a number of important issues for consideration with regard to child speech synthesis for dialogue systems. We specifically discuss challenges in building child synthetic voices compared to adult synthetic voices, synthesizing expressive conversational speech, and evaluating speech synthesis quality.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
We present a number of important issues for consideration with regard to child speech synthesis for dialogue systems. We specifically discuss challenges in building child synthetic voices compared to adult synthetic voices, synthesizing expressive conversational speech, and evaluating speech synthesis quality.
8.
Speggiorin, Alessandro; Dalton, Jeffrey; Leuski, Anton
TaskMAD: A Platform for Multimodal Task-Centric Knowledge-Grounded Conversational Experimentation Proceedings Article
In: Proceedings of the 45th International ACM SIGIR Conference on Research and Development in Information Retrieval, pp. 3240–3244, ACM, Madrid Spain, 2022, ISBN: 978-1-4503-8732-3.
@inproceedings{speggiorin_taskmad_2022,
title = {TaskMAD: A Platform for Multimodal Task-Centric Knowledge-Grounded Conversational Experimentation},
author = {Alessandro Speggiorin and Jeffrey Dalton and Anton Leuski},
url = {https://dl.acm.org/doi/10.1145/3477495.3531679},
doi = {10.1145/3477495.3531679},
isbn = {978-1-4503-8732-3},
year = {2022},
date = {2022-07-01},
urldate = {2022-09-22},
booktitle = {Proceedings of the 45th International ACM SIGIR Conference on Research and Development in Information Retrieval},
pages = {3240–3244},
publisher = {ACM},
address = {Madrid Spain},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
9.
Chaffey, Patricia; Traum, David
Identity models for role-play dialogue characters Proceedings Article
In: 2021.
@inproceedings{chaffey_identity_2021,
title = {Identity models for role-play dialogue characters},
author = {Patricia Chaffey and David Traum},
url = {http://semdial.org/anthology/papers/Z/Z21/Z21-4022/},
year = {2021},
date = {2021-09-01},
urldate = {2022-09-23},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
10.
Bonial, Claire; Abrams, Mitchell; Baker, Anthony L.; Hudson, Taylor; Lukin, Stephanie; Traum, David; Voss, Clare
Context is key: Annotating situated dialogue relations in multi-floor dialogue Proceedings Article
In: 2021.
@inproceedings{bonial_context_2021,
title = {Context is key: Annotating situated dialogue relations in multi-floor dialogue},
author = {Claire Bonial and Mitchell Abrams and Anthony L. Baker and Taylor Hudson and Stephanie Lukin and David Traum and Clare Voss},
url = {http://semdial.org/anthology/papers/Z/Z21/Z21-3006/},
year = {2021},
date = {2021-09-01},
urldate = {2022-09-23},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
11.
Johnson, Emmanuel; Gratch, Jonathan; Boberg, Jill; DeVault, David; Kim, Peter; Lucas, Gale
Using Intelligent Agents to Examine Gender in Negotiations Proceedings Article
In: Proceedings of the 21th ACM International Conference on Intelligent Virtual Agents, pp. 90–97, ACM, Virtual Event Japan, 2021, ISBN: 978-1-4503-8619-7.
@inproceedings{johnson_using_2021,
title = {Using Intelligent Agents to Examine Gender in Negotiations},
author = {Emmanuel Johnson and Jonathan Gratch and Jill Boberg and David DeVault and Peter Kim and Gale Lucas},
url = {https://dl.acm.org/doi/10.1145/3472306.3478348},
doi = {10.1145/3472306.3478348},
isbn = {978-1-4503-8619-7},
year = {2021},
date = {2021-09-01},
urldate = {2022-09-28},
booktitle = {Proceedings of the 21th ACM International Conference on Intelligent Virtual Agents},
pages = {90–97},
publisher = {ACM},
address = {Virtual Event Japan},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
12.
Bonial, Claire; Abrams, Mitchell; Traum, David; Voss, Clare
Builder, we have done it: Evaluating & Extending Dialogue-AMR NLU Pipeline for Two Collaborative Domains Proceedings Article
In: Proceedings of the 14th International Conference on Computational Semantics (IWCS), pp. 173–183, Association for Computational Linguistics, Groningen, The Netherlands (online), 2021.
@inproceedings{bonial_builder_2021,
title = {Builder, we have done it: Evaluating & Extending Dialogue-AMR NLU Pipeline for Two Collaborative Domains},
author = {Claire Bonial and Mitchell Abrams and David Traum and Clare Voss},
url = {https://aclanthology.org/2021.iwcs-1.17},
year = {2021},
date = {2021-06-01},
urldate = {2022-09-23},
booktitle = {Proceedings of the 14th International Conference on Computational Semantics (IWCS)},
pages = {173–183},
publisher = {Association for Computational Linguistics},
address = {Groningen, The Netherlands (online)},
abstract = {We adopt, evaluate, and improve upon a two-step natural language understanding (NLU) pipeline that incrementally tames the variation of unconstrained natural language input and maps to executable robot behaviors. The pipeline first leverages Abstract Meaning Representation (AMR) parsing to capture the propositional content of the utterance, and second converts this into “Dialogue-AMR,” which augments standard AMR with information on tense, aspect, and speech acts. Several alternative approaches and training datasets are evaluated for both steps and corresponding components of the pipeline, some of which outperform the original. We extend the Dialogue-AMR annotation schema to cover a different collaborative instruction domain and evaluate on both domains. With very little training data, we achieve promising performance in the new domain, demonstrating the scalability of this approach.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
We adopt, evaluate, and improve upon a two-step natural language understanding (NLU) pipeline that incrementally tames the variation of unconstrained natural language input and maps to executable robot behaviors. The pipeline first leverages Abstract Meaning Representation (AMR) parsing to capture the propositional content of the utterance, and second converts this into “Dialogue-AMR,” which augments standard AMR with information on tense, aspect, and speech acts. Several alternative approaches and training datasets are evaluated for both steps and corresponding components of the pipeline, some of which outperform the original. We extend the Dialogue-AMR annotation schema to cover a different collaborative instruction domain and evaluate on both domains. With very little training data, we achieve promising performance in the new domain, demonstrating the scalability of this approach.
13.
Gervits, Felix; Leuski, Anton; Bonial, Claire; Gordon, Carla; Traum, David
A Classification-Based Approach to Automating Human-Robot Dialogue Book Section
In: Marchi, Erik; Siniscalchi, Sabato Marco; Cumani, Sandro; Salerno, Valerio Mario; Li, Haizhou (Ed.): Increasing Naturalness and Flexibility in Spoken Dialogue Interaction: 10th International Workshop on Spoken Dialogue Systems, pp. 115–127, Springer, Singapore, 2021, ISBN: 978-981-15-9323-9.
@incollection{gervits_classification-based_2021,
title = {A Classification-Based Approach to Automating Human-Robot Dialogue},
author = {Felix Gervits and Anton Leuski and Claire Bonial and Carla Gordon and David Traum},
editor = {Erik Marchi and Sabato Marco Siniscalchi and Sandro Cumani and Valerio Mario Salerno and Haizhou Li},
url = {https://doi.org/10.1007/978-981-15-9323-9_10},
doi = {10.1007/978-981-15-9323-9_10},
isbn = {978-981-15-9323-9},
year = {2021},
date = {2021-01-01},
urldate = {2022-09-23},
booktitle = {Increasing Naturalness and Flexibility in Spoken Dialogue Interaction: 10th International Workshop on Spoken Dialogue Systems},
pages = {115–127},
publisher = {Springer},
address = {Singapore},
series = {Lecture Notes in Electrical Engineering},
abstract = {We present a dialogue system based on statistical classification which was used to automate human-robot dialogue in a collaborative navigation domain. The classifier was trained on a small corpus of multi-floor Wizard-of-Oz dialogue including two wizards: one standing in for dialogue capabilities and another for navigation. Below, we describe the implementation details of the classifier and show how it was used to automate the dialogue wizard. We evaluate our system on several sets of source data from the corpus and find that response accuracy is generally high, even with very limited training data. Another contribution of this work is the novel demonstration of a dialogue manager that uses the classifier to engage in multi-floor dialogue with two different human roles. Overall, this approach is useful for enabling spoken dialogue systems to produce robust and accurate responses to natural language input, and for robots that need to interact with humans in a team setting.},
keywords = {},
pubstate = {published},
tppubtype = {incollection}
}
We present a dialogue system based on statistical classification which was used to automate human-robot dialogue in a collaborative navigation domain. The classifier was trained on a small corpus of multi-floor Wizard-of-Oz dialogue including two wizards: one standing in for dialogue capabilities and another for navigation. Below, we describe the implementation details of the classifier and show how it was used to automate the dialogue wizard. We evaluate our system on several sets of source data from the corpus and find that response accuracy is generally high, even with very limited training data. Another contribution of this work is the novel demonstration of a dialogue manager that uses the classifier to engage in multi-floor dialogue with two different human roles. Overall, this approach is useful for enabling spoken dialogue systems to produce robust and accurate responses to natural language input, and for robots that need to interact with humans in a team setting.
14.
Kawano, Seiya; Yoshino, Koichiro; Traum, David; Nakamura, Satoshi
Dialogue Structure Parsing on Multi-Floor Dialogue Based on Multi-Task Learning Proceedings Article
In: 1st RobotDial Workshop on Dialogue Models for Human-Robot Interaction, pp. 21–29, ISCA, 2021.
@inproceedings{kawano_dialogue_2021,
title = {Dialogue Structure Parsing on Multi-Floor Dialogue Based on Multi-Task Learning},
author = {Seiya Kawano and Koichiro Yoshino and David Traum and Satoshi Nakamura},
url = {http://www.isca-speech.org/archive/RobotDial_2021/abstracts/4.html},
doi = {10.21437/RobotDial.2021-4},
year = {2021},
date = {2021-01-01},
urldate = {2021-04-15},
booktitle = {1st RobotDial Workshop on Dialogue Models for Human-Robot Interaction},
pages = {21–29},
publisher = {ISCA},
abstract = {A multi-floor dialogue consists of multiple sets of dialogue participants, each conversing within their own floor, but also at least one multicommunicating member who is a participant of multiple floors and coordinating each to achieve a shared dialogue goal. The structure of such dialogues can be complex, involving intentional structure and relations that are within or across floors. In this study, we propose a neural dialogue structure parser based on multi-task learning and an attention mechanism on multi-floor dialogues in a collaborative robot navigation domain. Our experimental results show that our proposed model improved the dialogue structure parsing performance more than those of single models, which are trained on each dialogue structure parsing task in multi-floor dialogues.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
A multi-floor dialogue consists of multiple sets of dialogue participants, each conversing within their own floor, but also at least one multicommunicating member who is a participant of multiple floors and coordinating each to achieve a shared dialogue goal. The structure of such dialogues can be complex, involving intentional structure and relations that are within or across floors. In this study, we propose a neural dialogue structure parser based on multi-task learning and an attention mechanism on multi-floor dialogues in a collaborative robot navigation domain. Our experimental results show that our proposed model improved the dialogue structure parsing performance more than those of single models, which are trained on each dialogue structure parsing task in multi-floor dialogues.
15.
Georgila, Kallirroi; Leuski, Anton; Yanov, Volodymyr; Traum, David
Evaluation of Off-the-shelf Speech Recognizers Across Diverse Dialogue Domains Proceedings Article
In: Proceedings of the Twelfth Language Resources and Evaluation Conference, pp. 6469–6476, European Language Resources Association, Sapporo, Japan, 2020.
@inproceedings{georgila_evaluation_2020,
title = {Evaluation of Off-the-shelf Speech Recognizers Across Diverse Dialogue Domains},
author = {Kallirroi Georgila and Anton Leuski and Volodymyr Yanov and David Traum},
url = {https://aclanthology.org/2020.lrec-1.797.pdf},
year = {2020},
date = {2020-01-01},
booktitle = {Proceedings of the Twelfth Language Resources and Evaluation Conference},
pages = {6469–6476},
publisher = {European Language Resources Association},
address = {Sapporo, Japan},
abstract = {We evaluate several publicly available off-the-shelf (commercial and research) automatic speech recognition (ASR) systems across diverse dialogue domains (in US-English). Our evaluation is aimed at non-experts with limited experience in speech recognition. Our goal is not only to compare a variety of ASR systems on several diverse data sets but also to measure how much ASR technology has advanced since our previous large-scale evaluations on the same data sets. Our results show that the performance of each speech recognizer can vary significantly depending on the domain. Furthermore, despite major recent progress in ASR technology, current state-of-the-art speech recognizers perform poorly in domains that require special vocabulary and language models, and under noisy conditions. We expect that our evaluation will prove useful to ASR consumers and dialogue system designers.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
We evaluate several publicly available off-the-shelf (commercial and research) automatic speech recognition (ASR) systems across diverse dialogue domains (in US-English). Our evaluation is aimed at non-experts with limited experience in speech recognition. Our goal is not only to compare a variety of ASR systems on several diverse data sets but also to measure how much ASR technology has advanced since our previous large-scale evaluations on the same data sets. Our results show that the performance of each speech recognizer can vary significantly depending on the domain. Furthermore, despite major recent progress in ASR technology, current state-of-the-art speech recognizers perform poorly in domains that require special vocabulary and language models, and under noisy conditions. We expect that our evaluation will prove useful to ASR consumers and dialogue system designers.
16.
Artstein, Ron; Chen, Elizabeth
Augmenting Training Data for a Virtual Character Using GPT-3.5 Proceedings Article
In: Tyhe Florida Artificial Intelligence Research Society, 0000.
@inproceedings{artstein_augmenting_nodate,
title = {Augmenting Training Data for a Virtual Character Using GPT-3.5},
author = {Ron Artstein and Elizabeth Chen},
url = {https://journals.flvc.org/FLAIRS/article/view/135552},
volume = {37},
publisher = {Tyhe Florida Artificial Intelligence Research Society},
abstract = {This paper compares different methods of using a large lan-guage model (GPT-3.5) for creating synthetic training datafor a retrieval-based conversational character. The trainingdata are in the form of linked questions and answers, whichallow a classifier to retrieve a pre-recorded answer to an un-seen question; the intuition is that a large language modelcould predict what human users might ask, thus saving theeffort of collecting real user questions as training data. Re-sults show small improvements in test performance for allsynthetic datasets. However, a classifier trained on only smallamounts of collected user data resulted in a higher F-scorethan the classifiers trained on much larger amounts of syn-thetic data generated using GPT-3.5. Based on these results,we see a potential in using large language models for gener-ating training data, but at this point it is not as valuable ascollecting actual user data for training.},
keywords = {},
pubstate = {published},
tppubtype = {inproceedings}
}
This paper compares different methods of using a large lan-guage model (GPT-3.5) for creating synthetic training datafor a retrieval-based conversational character. The trainingdata are in the form of linked questions and answers, whichallow a classifier to retrieve a pre-recorded answer to an un-seen question; the intuition is that a large language modelcould predict what human users might ask, thus saving theeffort of collecting real user questions as training data. Re-sults show small improvements in test performance for allsynthetic datasets. However, a classifier trained on only smallamounts of collected user data resulted in a higher F-scorethan the classifiers trained on much larger amounts of syn-thetic data generated using GPT-3.5. Based on these results,we see a potential in using large language models for gener-ating training data, but at this point it is not as valuable ascollecting actual user data for training.
Filter
2024
Georgila, Kallirroi
Comparing Pre-Trained Embeddings and Domain-Independent Features for Regression-Based Evaluation of Task-Oriented Dialogue Systems Proceedings Article
In: Proceedings of the 25th Annual Meeting of the Special Interest Group on Discourse and Dialogue, pp. 610–623, Association for Computational Linguistics, Kyoto, Japan, 2024.
Links | BibTeX | Tags: Dialogue, DTIC, Natural Language
@inproceedings{georgila_comparing_2024,
title = {Comparing Pre-Trained Embeddings and Domain-Independent Features for Regression-Based Evaluation of Task-Oriented Dialogue Systems},
author = {Kallirroi Georgila},
url = {https://aclanthology.org/2024.sigdial-1.52},
doi = {10.18653/v1/2024.sigdial-1.52},
year = {2024},
date = {2024-09-01},
urldate = {2024-10-15},
booktitle = {Proceedings of the 25th Annual Meeting of the Special Interest Group on Discourse and Dialogue},
pages = {610–623},
publisher = {Association for Computational Linguistics},
address = {Kyoto, Japan},
keywords = {Dialogue, DTIC, Natural Language},
pubstate = {published},
tppubtype = {inproceedings}
}
Frummet, Alexander; Speggiorin, Alessandro; Elsweiler, David; Leuski, Anton; Dalton, Jeff
Cooking with Conversation: Enhancing User Engagement and Learning with a Knowledge-Enhancing Assistant Journal Article
In: ACM Trans. Inf. Syst., pp. 3649500, 2024, ISSN: 1046-8188, 1558-2868.
Abstract | Links | BibTeX | Tags: DTIC, Natural Language, UARC
@article{frummet_cooking_2024,
title = {Cooking with Conversation: Enhancing User Engagement and Learning with a Knowledge-Enhancing Assistant},
author = {Alexander Frummet and Alessandro Speggiorin and David Elsweiler and Anton Leuski and Jeff Dalton},
url = {https://dl.acm.org/doi/10.1145/3649500},
doi = {10.1145/3649500},
issn = {1046-8188, 1558-2868},
year = {2024},
date = {2024-03-01},
urldate = {2024-04-16},
journal = {ACM Trans. Inf. Syst.},
pages = {3649500},
abstract = {We present two empirical studies to investigate users’ expectations and behaviours when using digital assistants, such as Alexa and Google Home, in a kitchen context: First, a survey (N=200) queries participants on their expectations for the kinds of information that such systems should be able to provide. While consensus exists on expecting information about cooking steps and processes, younger participants who enjoy cooking express a higher likelihood of expecting details on food history or the science of cooking. In a follow-up Wizard-of-Oz study (N = 48), users were guided through the steps of a recipe either by an
active
wizard that alerted participants to information it could provide or a
passive
wizard who only answered questions that were provided by the user. The
active
policy led to almost double the number of conversational utterances and 1.5 times more knowledge-related user questions compared to the
passive
policy. Also, it resulted in 1.7 times more knowledge communicated than the
passive
policy. We discuss the findings in the context of related work and reveal implications for the design and use of such assistants for cooking and other purposes such as DIY and craft tasks, as well as the lessons we learned for evaluating such systems.},
keywords = {DTIC, Natural Language, UARC},
pubstate = {published},
tppubtype = {article}
}
We present two empirical studies to investigate users’ expectations and behaviours when using digital assistants, such as Alexa and Google Home, in a kitchen context: First, a survey (N=200) queries participants on their expectations for the kinds of information that such systems should be able to provide. While consensus exists on expecting information about cooking steps and processes, younger participants who enjoy cooking express a higher likelihood of expecting details on food history or the science of cooking. In a follow-up Wizard-of-Oz study (N = 48), users were guided through the steps of a recipe either by an
active
wizard that alerted participants to information it could provide or a
passive
wizard who only answered questions that were provided by the user. The
active
policy led to almost double the number of conversational utterances and 1.5 times more knowledge-related user questions compared to the
passive
policy. Also, it resulted in 1.7 times more knowledge communicated than the
passive
policy. We discuss the findings in the context of related work and reveal implications for the design and use of such assistants for cooking and other purposes such as DIY and craft tasks, as well as the lessons we learned for evaluating such systems.
active
wizard that alerted participants to information it could provide or a
passive
wizard who only answered questions that were provided by the user. The
active
policy led to almost double the number of conversational utterances and 1.5 times more knowledge-related user questions compared to the
passive
policy. Also, it resulted in 1.7 times more knowledge communicated than the
passive
policy. We discuss the findings in the context of related work and reveal implications for the design and use of such assistants for cooking and other purposes such as DIY and craft tasks, as well as the lessons we learned for evaluating such systems.
2023
Lukin, Stephanie M.; Pollard, Kimberly A.; Bonial, Claire; Hudson, Taylor; Arstein, Ron; Voss, Clare; Traum, David
Navigating to Success in Multi-Modal Human-Robot Collaboration: Analysis and Corpus Release Miscellaneous
2023, (arXiv:2310.17568 [cs]).
Abstract | Links | BibTeX | Tags: DTIC, Natural Language, UARC
@misc{lukin_navigating_2023,
title = {Navigating to Success in Multi-Modal Human-Robot Collaboration: Analysis and Corpus Release},
author = {Stephanie M. Lukin and Kimberly A. Pollard and Claire Bonial and Taylor Hudson and Ron Arstein and Clare Voss and David Traum},
url = {http://arxiv.org/abs/2310.17568},
year = {2023},
date = {2023-10-01},
urldate = {2023-12-07},
publisher = {arXiv},
abstract = {Human-guided robotic exploration is a useful approach to gathering information at remote locations, especially those that might be too risky, inhospitable, or inaccessible for humans. Maintaining common ground between the remotely-located partners is a challenge, one that can be facilitated by multi-modal communication. In this paper, we explore how participants utilized multiple modalities to investigate a remote location with the help of a robotic partner. Participants issued spoken natural language instructions and received from the robot: text-based feedback, continuous 2D LIDAR mapping, and upon-request static photographs. We noticed that different strategies were adopted in terms of use of the modalities, and hypothesize that these differences may be correlated with success at several exploration sub-tasks. We found that requesting photos may have improved the identification and counting of some key entities (doorways in particular) and that this strategy did not hinder the amount of overall area exploration. Future work with larger samples may reveal the effects of more nuanced photo and dialogue strategies, which can inform the training of robotic agents. Additionally, we announce the release of our unique multi-modal corpus of human-robot communication in an exploration context: SCOUT, the Situated Corpus on Understanding Transactions.},
note = {arXiv:2310.17568 [cs]},
keywords = {DTIC, Natural Language, UARC},
pubstate = {published},
tppubtype = {misc}
}
Human-guided robotic exploration is a useful approach to gathering information at remote locations, especially those that might be too risky, inhospitable, or inaccessible for humans. Maintaining common ground between the remotely-located partners is a challenge, one that can be facilitated by multi-modal communication. In this paper, we explore how participants utilized multiple modalities to investigate a remote location with the help of a robotic partner. Participants issued spoken natural language instructions and received from the robot: text-based feedback, continuous 2D LIDAR mapping, and upon-request static photographs. We noticed that different strategies were adopted in terms of use of the modalities, and hypothesize that these differences may be correlated with success at several exploration sub-tasks. We found that requesting photos may have improved the identification and counting of some key entities (doorways in particular) and that this strategy did not hinder the amount of overall area exploration. Future work with larger samples may reveal the effects of more nuanced photo and dialogue strategies, which can inform the training of robotic agents. Additionally, we announce the release of our unique multi-modal corpus of human-robot communication in an exploration context: SCOUT, the Situated Corpus on Understanding Transactions.
Gilani, Setareh Nasihati; Pollard, Kimberly; Traum, David
Multimodal Prediction of User's Performance in High-Stress Dialogue Interactions Proceedings Article
In: International Cconference on Multimodal Interaction, pp. 71–75, ACM, Paris France, 2023, ISBN: 979-8-4007-0321-8.
Links | BibTeX | Tags: DTIC, Natural Language, UARC
@inproceedings{nasihati_gilani_multimodal_2023,
title = {Multimodal Prediction of User's Performance in High-Stress Dialogue Interactions},
author = {Setareh Nasihati Gilani and Kimberly Pollard and David Traum},
url = {https://dl.acm.org/doi/10.1145/3610661.3617166},
doi = {10.1145/3610661.3617166},
isbn = {979-8-4007-0321-8},
year = {2023},
date = {2023-10-01},
urldate = {2023-12-07},
booktitle = {International Cconference on Multimodal Interaction},
pages = {71–75},
publisher = {ACM},
address = {Paris France},
keywords = {DTIC, Natural Language, UARC},
pubstate = {published},
tppubtype = {inproceedings}
}
Gainer, Alesia; Aptaker, Allison; Artstein, Ron; Cobbins, David; Core, Mark; Gordon, Carla; Leuski, Anton; Li, Zongjian; Merchant, Chirag; Nelson, David; Soleymani, Mohammad; Traum, David
DIVIS: Digital Interactive Victim Intake Simulator Proceedings Article
In: Proceedings of the 23rd ACM International Conference on Intelligent Virtual Agents, pp. 1–2, ACM, Würzburg Germany, 2023, ISBN: 978-1-4503-9994-4.
Links | BibTeX | Tags: DTIC, MxR, UARC, Virtual Humans
@inproceedings{gainer_divis_2023,
title = {DIVIS: Digital Interactive Victim Intake Simulator},
author = {Alesia Gainer and Allison Aptaker and Ron Artstein and David Cobbins and Mark Core and Carla Gordon and Anton Leuski and Zongjian Li and Chirag Merchant and David Nelson and Mohammad Soleymani and David Traum},
url = {https://dl.acm.org/doi/10.1145/3570945.3607328},
doi = {10.1145/3570945.3607328},
isbn = {978-1-4503-9994-4},
year = {2023},
date = {2023-09-01},
urldate = {2024-02-20},
booktitle = {Proceedings of the 23rd ACM International Conference on Intelligent Virtual Agents},
pages = {1–2},
publisher = {ACM},
address = {Würzburg Germany},
keywords = {DTIC, MxR, UARC, Virtual Humans},
pubstate = {published},
tppubtype = {inproceedings}
}
Pal, Debaditya; Leuski, Anton; Traum, David
Comparing Statistical Models for Retrieval based Question-answering Dialogue: BERT vs Relevance Models Journal Article
In: FLAIRS, vol. 36, 2023, ISSN: 2334-0762.
Abstract | Links | BibTeX | Tags: DTIC, Natural Language, UARC
@article{pal_comparing_2023,
title = {Comparing Statistical Models for Retrieval based Question-answering Dialogue: BERT vs Relevance Models},
author = {Debaditya Pal and Anton Leuski and David Traum},
url = {https://journals.flvc.org/FLAIRS/article/view/133386},
doi = {10.32473/flairs.36.133386},
issn = {2334-0762},
year = {2023},
date = {2023-05-01},
urldate = {2023-08-23},
journal = {FLAIRS},
volume = {36},
abstract = {In this paper, we compare the performance of four models in a retrieval based question answering dialogue task on two moderately sized corpora (textasciitilde 10,000 utterances). One model is a statistical model and uses cross language relevance while the others are deep neural networks utilizing the BERT architecture along with different retrieval methods. The statistical model has previously outperformed LSTM based neural networks in a similar task whereas BERT has been proven to perform well on a variety of NLP tasks, achieving state-of-the-art results in many of them. Results show that the statistical cross language relevance model outperforms the BERT based architectures in learning question-answer mappings. BERT achieves better results by mapping new questions to existing questions.},
keywords = {DTIC, Natural Language, UARC},
pubstate = {published},
tppubtype = {article}
}
In this paper, we compare the performance of four models in a retrieval based question answering dialogue task on two moderately sized corpora (textasciitilde 10,000 utterances). One model is a statistical model and uses cross language relevance while the others are deep neural networks utilizing the BERT architecture along with different retrieval methods. The statistical model has previously outperformed LSTM based neural networks in a similar task whereas BERT has been proven to perform well on a variety of NLP tasks, achieving state-of-the-art results in many of them. Results show that the statistical cross language relevance model outperforms the BERT based architectures in learning question-answer mappings. BERT achieves better results by mapping new questions to existing questions.
Georgila, Kallirroi
Considerations for Child Speech Synthesis for Dialogue Systems Proceedings Article
In: Los Angeles, CA, 2023.
Abstract | Links | BibTeX | Tags: DTIC, UARC, Virtual Humans
@inproceedings{georgila_considerations_2023,
title = {Considerations for Child Speech Synthesis for Dialogue Systems},
author = {Kallirroi Georgila},
url = {https://kgeorgila.github.io/publications/georgila_aiaic23.pdf},
year = {2023},
date = {2023-03-01},
address = {Los Angeles, CA},
abstract = {We present a number of important issues for consideration with regard to child speech synthesis for dialogue systems. We specifically discuss challenges in building child synthetic voices compared to adult synthetic voices, synthesizing expressive conversational speech, and evaluating speech synthesis quality.},
keywords = {DTIC, UARC, Virtual Humans},
pubstate = {published},
tppubtype = {inproceedings}
}
We present a number of important issues for consideration with regard to child speech synthesis for dialogue systems. We specifically discuss challenges in building child synthetic voices compared to adult synthetic voices, synthesizing expressive conversational speech, and evaluating speech synthesis quality.
2022
Speggiorin, Alessandro; Dalton, Jeffrey; Leuski, Anton
TaskMAD: A Platform for Multimodal Task-Centric Knowledge-Grounded Conversational Experimentation Proceedings Article
In: Proceedings of the 45th International ACM SIGIR Conference on Research and Development in Information Retrieval, pp. 3240–3244, ACM, Madrid Spain, 2022, ISBN: 978-1-4503-8732-3.
Links | BibTeX | Tags: Dialogue, DTIC, UARC
@inproceedings{speggiorin_taskmad_2022,
title = {TaskMAD: A Platform for Multimodal Task-Centric Knowledge-Grounded Conversational Experimentation},
author = {Alessandro Speggiorin and Jeffrey Dalton and Anton Leuski},
url = {https://dl.acm.org/doi/10.1145/3477495.3531679},
doi = {10.1145/3477495.3531679},
isbn = {978-1-4503-8732-3},
year = {2022},
date = {2022-07-01},
urldate = {2022-09-22},
booktitle = {Proceedings of the 45th International ACM SIGIR Conference on Research and Development in Information Retrieval},
pages = {3240–3244},
publisher = {ACM},
address = {Madrid Spain},
keywords = {Dialogue, DTIC, UARC},
pubstate = {published},
tppubtype = {inproceedings}
}
2021
Chaffey, Patricia; Traum, David
Identity models for role-play dialogue characters Proceedings Article
In: 2021.
Links | BibTeX | Tags: Dialogue, DTIC, UARC
@inproceedings{chaffey_identity_2021,
title = {Identity models for role-play dialogue characters},
author = {Patricia Chaffey and David Traum},
url = {http://semdial.org/anthology/papers/Z/Z21/Z21-4022/},
year = {2021},
date = {2021-09-01},
urldate = {2022-09-23},
keywords = {Dialogue, DTIC, UARC},
pubstate = {published},
tppubtype = {inproceedings}
}
Bonial, Claire; Abrams, Mitchell; Baker, Anthony L.; Hudson, Taylor; Lukin, Stephanie; Traum, David; Voss, Clare
Context is key: Annotating situated dialogue relations in multi-floor dialogue Proceedings Article
In: 2021.
Links | BibTeX | Tags: Dialogue, DTIC
@inproceedings{bonial_context_2021,
title = {Context is key: Annotating situated dialogue relations in multi-floor dialogue},
author = {Claire Bonial and Mitchell Abrams and Anthony L. Baker and Taylor Hudson and Stephanie Lukin and David Traum and Clare Voss},
url = {http://semdial.org/anthology/papers/Z/Z21/Z21-3006/},
year = {2021},
date = {2021-09-01},
urldate = {2022-09-23},
keywords = {Dialogue, DTIC},
pubstate = {published},
tppubtype = {inproceedings}
}
Johnson, Emmanuel; Gratch, Jonathan; Boberg, Jill; DeVault, David; Kim, Peter; Lucas, Gale
Using Intelligent Agents to Examine Gender in Negotiations Proceedings Article
In: Proceedings of the 21th ACM International Conference on Intelligent Virtual Agents, pp. 90–97, ACM, Virtual Event Japan, 2021, ISBN: 978-1-4503-8619-7.
Links | BibTeX | Tags: DTIC, UARC, Virtual Humans
@inproceedings{johnson_using_2021,
title = {Using Intelligent Agents to Examine Gender in Negotiations},
author = {Emmanuel Johnson and Jonathan Gratch and Jill Boberg and David DeVault and Peter Kim and Gale Lucas},
url = {https://dl.acm.org/doi/10.1145/3472306.3478348},
doi = {10.1145/3472306.3478348},
isbn = {978-1-4503-8619-7},
year = {2021},
date = {2021-09-01},
urldate = {2022-09-28},
booktitle = {Proceedings of the 21th ACM International Conference on Intelligent Virtual Agents},
pages = {90–97},
publisher = {ACM},
address = {Virtual Event Japan},
keywords = {DTIC, UARC, Virtual Humans},
pubstate = {published},
tppubtype = {inproceedings}
}
Bonial, Claire; Abrams, Mitchell; Traum, David; Voss, Clare
Builder, we have done it: Evaluating & Extending Dialogue-AMR NLU Pipeline for Two Collaborative Domains Proceedings Article
In: Proceedings of the 14th International Conference on Computational Semantics (IWCS), pp. 173–183, Association for Computational Linguistics, Groningen, The Netherlands (online), 2021.
Abstract | Links | BibTeX | Tags: Dialogue, DTIC
@inproceedings{bonial_builder_2021,
title = {Builder, we have done it: Evaluating & Extending Dialogue-AMR NLU Pipeline for Two Collaborative Domains},
author = {Claire Bonial and Mitchell Abrams and David Traum and Clare Voss},
url = {https://aclanthology.org/2021.iwcs-1.17},
year = {2021},
date = {2021-06-01},
urldate = {2022-09-23},
booktitle = {Proceedings of the 14th International Conference on Computational Semantics (IWCS)},
pages = {173–183},
publisher = {Association for Computational Linguistics},
address = {Groningen, The Netherlands (online)},
abstract = {We adopt, evaluate, and improve upon a two-step natural language understanding (NLU) pipeline that incrementally tames the variation of unconstrained natural language input and maps to executable robot behaviors. The pipeline first leverages Abstract Meaning Representation (AMR) parsing to capture the propositional content of the utterance, and second converts this into “Dialogue-AMR,” which augments standard AMR with information on tense, aspect, and speech acts. Several alternative approaches and training datasets are evaluated for both steps and corresponding components of the pipeline, some of which outperform the original. We extend the Dialogue-AMR annotation schema to cover a different collaborative instruction domain and evaluate on both domains. With very little training data, we achieve promising performance in the new domain, demonstrating the scalability of this approach.},
keywords = {Dialogue, DTIC},
pubstate = {published},
tppubtype = {inproceedings}
}
We adopt, evaluate, and improve upon a two-step natural language understanding (NLU) pipeline that incrementally tames the variation of unconstrained natural language input and maps to executable robot behaviors. The pipeline first leverages Abstract Meaning Representation (AMR) parsing to capture the propositional content of the utterance, and second converts this into “Dialogue-AMR,” which augments standard AMR with information on tense, aspect, and speech acts. Several alternative approaches and training datasets are evaluated for both steps and corresponding components of the pipeline, some of which outperform the original. We extend the Dialogue-AMR annotation schema to cover a different collaborative instruction domain and evaluate on both domains. With very little training data, we achieve promising performance in the new domain, demonstrating the scalability of this approach.
Gervits, Felix; Leuski, Anton; Bonial, Claire; Gordon, Carla; Traum, David
A Classification-Based Approach to Automating Human-Robot Dialogue Book Section
In: Marchi, Erik; Siniscalchi, Sabato Marco; Cumani, Sandro; Salerno, Valerio Mario; Li, Haizhou (Ed.): Increasing Naturalness and Flexibility in Spoken Dialogue Interaction: 10th International Workshop on Spoken Dialogue Systems, pp. 115–127, Springer, Singapore, 2021, ISBN: 978-981-15-9323-9.
Abstract | Links | BibTeX | Tags: Dialogue, DTIC
@incollection{gervits_classification-based_2021,
title = {A Classification-Based Approach to Automating Human-Robot Dialogue},
author = {Felix Gervits and Anton Leuski and Claire Bonial and Carla Gordon and David Traum},
editor = {Erik Marchi and Sabato Marco Siniscalchi and Sandro Cumani and Valerio Mario Salerno and Haizhou Li},
url = {https://doi.org/10.1007/978-981-15-9323-9_10},
doi = {10.1007/978-981-15-9323-9_10},
isbn = {978-981-15-9323-9},
year = {2021},
date = {2021-01-01},
urldate = {2022-09-23},
booktitle = {Increasing Naturalness and Flexibility in Spoken Dialogue Interaction: 10th International Workshop on Spoken Dialogue Systems},
pages = {115–127},
publisher = {Springer},
address = {Singapore},
series = {Lecture Notes in Electrical Engineering},
abstract = {We present a dialogue system based on statistical classification which was used to automate human-robot dialogue in a collaborative navigation domain. The classifier was trained on a small corpus of multi-floor Wizard-of-Oz dialogue including two wizards: one standing in for dialogue capabilities and another for navigation. Below, we describe the implementation details of the classifier and show how it was used to automate the dialogue wizard. We evaluate our system on several sets of source data from the corpus and find that response accuracy is generally high, even with very limited training data. Another contribution of this work is the novel demonstration of a dialogue manager that uses the classifier to engage in multi-floor dialogue with two different human roles. Overall, this approach is useful for enabling spoken dialogue systems to produce robust and accurate responses to natural language input, and for robots that need to interact with humans in a team setting.},
keywords = {Dialogue, DTIC},
pubstate = {published},
tppubtype = {incollection}
}
We present a dialogue system based on statistical classification which was used to automate human-robot dialogue in a collaborative navigation domain. The classifier was trained on a small corpus of multi-floor Wizard-of-Oz dialogue including two wizards: one standing in for dialogue capabilities and another for navigation. Below, we describe the implementation details of the classifier and show how it was used to automate the dialogue wizard. We evaluate our system on several sets of source data from the corpus and find that response accuracy is generally high, even with very limited training data. Another contribution of this work is the novel demonstration of a dialogue manager that uses the classifier to engage in multi-floor dialogue with two different human roles. Overall, this approach is useful for enabling spoken dialogue systems to produce robust and accurate responses to natural language input, and for robots that need to interact with humans in a team setting.
Kawano, Seiya; Yoshino, Koichiro; Traum, David; Nakamura, Satoshi
Dialogue Structure Parsing on Multi-Floor Dialogue Based on Multi-Task Learning Proceedings Article
In: 1st RobotDial Workshop on Dialogue Models for Human-Robot Interaction, pp. 21–29, ISCA, 2021.
Abstract | Links | BibTeX | Tags: ARL, Dialogue, DTIC, Natural Language, Virtual Humans
@inproceedings{kawano_dialogue_2021,
title = {Dialogue Structure Parsing on Multi-Floor Dialogue Based on Multi-Task Learning},
author = {Seiya Kawano and Koichiro Yoshino and David Traum and Satoshi Nakamura},
url = {http://www.isca-speech.org/archive/RobotDial_2021/abstracts/4.html},
doi = {10.21437/RobotDial.2021-4},
year = {2021},
date = {2021-01-01},
urldate = {2021-04-15},
booktitle = {1st RobotDial Workshop on Dialogue Models for Human-Robot Interaction},
pages = {21–29},
publisher = {ISCA},
abstract = {A multi-floor dialogue consists of multiple sets of dialogue participants, each conversing within their own floor, but also at least one multicommunicating member who is a participant of multiple floors and coordinating each to achieve a shared dialogue goal. The structure of such dialogues can be complex, involving intentional structure and relations that are within or across floors. In this study, we propose a neural dialogue structure parser based on multi-task learning and an attention mechanism on multi-floor dialogues in a collaborative robot navigation domain. Our experimental results show that our proposed model improved the dialogue structure parsing performance more than those of single models, which are trained on each dialogue structure parsing task in multi-floor dialogues.},
keywords = {ARL, Dialogue, DTIC, Natural Language, Virtual Humans},
pubstate = {published},
tppubtype = {inproceedings}
}
A multi-floor dialogue consists of multiple sets of dialogue participants, each conversing within their own floor, but also at least one multicommunicating member who is a participant of multiple floors and coordinating each to achieve a shared dialogue goal. The structure of such dialogues can be complex, involving intentional structure and relations that are within or across floors. In this study, we propose a neural dialogue structure parser based on multi-task learning and an attention mechanism on multi-floor dialogues in a collaborative robot navigation domain. Our experimental results show that our proposed model improved the dialogue structure parsing performance more than those of single models, which are trained on each dialogue structure parsing task in multi-floor dialogues.
2020
Georgila, Kallirroi; Leuski, Anton; Yanov, Volodymyr; Traum, David
Evaluation of Off-the-shelf Speech Recognizers Across Diverse Dialogue Domains Proceedings Article
In: Proceedings of the Twelfth Language Resources and Evaluation Conference, pp. 6469–6476, European Language Resources Association, Sapporo, Japan, 2020.
Abstract | Links | BibTeX | Tags: DTIC
@inproceedings{georgila_evaluation_2020,
title = {Evaluation of Off-the-shelf Speech Recognizers Across Diverse Dialogue Domains},
author = {Kallirroi Georgila and Anton Leuski and Volodymyr Yanov and David Traum},
url = {https://aclanthology.org/2020.lrec-1.797.pdf},
year = {2020},
date = {2020-01-01},
booktitle = {Proceedings of the Twelfth Language Resources and Evaluation Conference},
pages = {6469–6476},
publisher = {European Language Resources Association},
address = {Sapporo, Japan},
abstract = {We evaluate several publicly available off-the-shelf (commercial and research) automatic speech recognition (ASR) systems across diverse dialogue domains (in US-English). Our evaluation is aimed at non-experts with limited experience in speech recognition. Our goal is not only to compare a variety of ASR systems on several diverse data sets but also to measure how much ASR technology has advanced since our previous large-scale evaluations on the same data sets. Our results show that the performance of each speech recognizer can vary significantly depending on the domain. Furthermore, despite major recent progress in ASR technology, current state-of-the-art speech recognizers perform poorly in domains that require special vocabulary and language models, and under noisy conditions. We expect that our evaluation will prove useful to ASR consumers and dialogue system designers.},
keywords = {DTIC},
pubstate = {published},
tppubtype = {inproceedings}
}
We evaluate several publicly available off-the-shelf (commercial and research) automatic speech recognition (ASR) systems across diverse dialogue domains (in US-English). Our evaluation is aimed at non-experts with limited experience in speech recognition. Our goal is not only to compare a variety of ASR systems on several diverse data sets but also to measure how much ASR technology has advanced since our previous large-scale evaluations on the same data sets. Our results show that the performance of each speech recognizer can vary significantly depending on the domain. Furthermore, despite major recent progress in ASR technology, current state-of-the-art speech recognizers perform poorly in domains that require special vocabulary and language models, and under noisy conditions. We expect that our evaluation will prove useful to ASR consumers and dialogue system designers.
0000
Artstein, Ron; Chen, Elizabeth
Augmenting Training Data for a Virtual Character Using GPT-3.5 Proceedings Article
In: Tyhe Florida Artificial Intelligence Research Society, 0000.
Abstract | Links | BibTeX | Tags: Dialogue, DTIC, Natural Language
@inproceedings{artstein_augmenting_nodate,
title = {Augmenting Training Data for a Virtual Character Using GPT-3.5},
author = {Ron Artstein and Elizabeth Chen},
url = {https://journals.flvc.org/FLAIRS/article/view/135552},
volume = {37},
publisher = {Tyhe Florida Artificial Intelligence Research Society},
abstract = {This paper compares different methods of using a large lan-guage model (GPT-3.5) for creating synthetic training datafor a retrieval-based conversational character. The trainingdata are in the form of linked questions and answers, whichallow a classifier to retrieve a pre-recorded answer to an un-seen question; the intuition is that a large language modelcould predict what human users might ask, thus saving theeffort of collecting real user questions as training data. Re-sults show small improvements in test performance for allsynthetic datasets. However, a classifier trained on only smallamounts of collected user data resulted in a higher F-scorethan the classifiers trained on much larger amounts of syn-thetic data generated using GPT-3.5. Based on these results,we see a potential in using large language models for gener-ating training data, but at this point it is not as valuable ascollecting actual user data for training.},
keywords = {Dialogue, DTIC, Natural Language},
pubstate = {published},
tppubtype = {inproceedings}
}
This paper compares different methods of using a large lan-guage model (GPT-3.5) for creating synthetic training datafor a retrieval-based conversational character. The trainingdata are in the form of linked questions and answers, whichallow a classifier to retrieve a pre-recorded answer to an un-seen question; the intuition is that a large language modelcould predict what human users might ask, thus saving theeffort of collecting real user questions as training data. Re-sults show small improvements in test performance for allsynthetic datasets. However, a classifier trained on only smallamounts of collected user data resulted in a higher F-scorethan the classifiers trained on much larger amounts of syn-thetic data generated using GPT-3.5. Based on these results,we see a potential in using large language models for gener-ating training data, but at this point it is not as valuable ascollecting actual user data for training.