DOCSLIB.ORG

Mitigation of Data Scarcity Issues for Semantic Classification in a Virtual

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text
Mitigation of Data Scarcity Issues for Semantic Classification in a Virtual Mitigation of Data Scarcity Issues for Semantic Classification in a Virtual Patient Dialogue Agent Dissertation Presented in Partial Fulfillment of the Requirements for the Degree Doctor of Philosophy in the Graduate School of The Ohio State University By Adam Stiff, BS Graduate Program in Computer Science and Engineering The Ohio State University 2020 Dissertation Committee: Eric Fosler-Lussier, Advisor Michael White Yu Su c Copyright by Adam Stiff 2020 Abstract We introduce a virtual patient question-answering dialogue system, used for training medical students to interview real patients, which presents many unique opportunities for research in linguistics, speech, and dialogue. Among the most challenging research topics at this point in the system’s development are issues relating to scarcity of training data. We address three main problems. The first challenge is that many questions are very rarely asked of the virtual patient, which leaves little data to learn adequate models of these questions. We validate one approach to this problem, which is to combine a statistical question classification model with a rule-based system, by deploying it in an experiment with live users. Additional work further improves rare question performance by utilizing a recurrent neural network model with a multi-headed self-attention mechanism. We contribute an analysis of the reasons for this improved performance, highlighting specialization and overlapping concerns in independent components of the model. Another data scarcity problem for the virtual patient project is the challenge of adequately characterizing questions that are deemed out-of-scope. By definition, these types of questions are infinite, so this problem is particularly challenging. We contribute a characterization of the problem as it manifests in our domain, as well as a baseline approach to handling the issue, and an analysis of the corresponding improvement in performance. ii Finally, we contribute a method for improving performance of domain-specific tasks such as ours, which use off-the-shelf speech recognition as inputs, when no in-domain speech data is available. This method augments text training data for the downstream task with inferred phonetic representations, to make the downstream task tolerant of speech recognition errors. We also see performance improvements from sampling simulated errors to replace the text inputs during training. Future enhancements to the spoken dialogue capabilities of the virtual patient are also considered. iii Dedicated to Elizabeth, without whose support—and patience— this would not have been possible. iv Acknowledgments Nothing written here will ever be adequate to acknowledge the full depth and breadth of the myriad of interactions that have contributed to me getting to this point, so I am embracing imperfection and keeping this short. Just know that if you are reading this, I undoubtedly have a reason to thank you, and I hope that I have the wisdom to appreciate what that reason is. I would like to extend special thanks to all of my collaborators whose efforts contributed to work described in this dissertation. Chapter 3 includes descriptions of some work by Kellen Maicher, Doug Danforth, Evan Jaffe, Marisa Scholl, and Mike White; Chapter 4 includes contributions from Mike White, Eric Fosler-Lussier, Lifeng Jin, Evan Jaffe, and Doug Danforth; Chapter 5 was a collaboration with Qi Song and Eric Fosler-Lussier; and Chapter 6 was coauthored with Prashant Serai and Eric Fosler-Lussier. Cheers, thanks, and best wishes to the friends made along the way: Deblin, Denis, Peter, Chaitanya, Joo-Kyung, Prashant, Ahmad, Manirupa, and the innumerable others who have also provided invaluable insight, support, humor, feedback, or otherwise made this journey a little more enjoyable. Finally, I want to express my extreme gratitude to Eric Fosler-Lussier for his mentorship and efforts in helping/pushing/dragging me over the finish line. Thanks for the guidance when I needed it, for the space to learn how to guide myself, and for the patience that I probably didn’t deserve. v Vita May 2020 . PhD, Computer Science and Engineering, The Ohio State University, USA. March 2006 . BS, Biochemistry, The Ohio State University, USA. Publications Research Publications Adam Stiff, Qi Song, and Eric Fosler-Lussier. How Self-Attention Improves Rare Class Performance. In Proceedings of the 21st Annual SIGdial Meeting on Discourse and Dialogue. Association for Computational Linguistics, 2020. Prashant Serai, Adam Stiff, and Eric Fosler-Lussier. End to end speech recognition error prediction with sequence to sequence learning. In 2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pp. 6339-6343. IEEE, 2020. Adam Stiff, Prashant Serai, and Eric Fosler-Lussier. Improving human-computer in- teraction in low-resource settings with text-to-phonetic data augmentation. In 2019 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pp. 7320-7324. IEEE, 2019. Deblin Bagchi, Peter Plantinga, Adam Stiff, and Eric Fosler-Lussier. Spectral feature mapping with mimic loss for robust speech recognition. In 2018 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pp. 5609-5613. IEEE, 2018. vi Fields of Study Major Field: Computer Science and Engineering Studies in: Artificial Intelligence Prof. Eric Fosler-Lussier Computational Linguistics Prof. Michael White Computational Neuropsychology Prof. Alexander Petrov vii Table of Contents Page Abstract . ii Dedication . iv Acknowledgments . .v Vita ........................................... vi List of Tables . xi List of Figures . xii 1. Introduction . .1 1.1 Overview and Contributions . .5 2. Background and Related Work . .8 2.1 Issues in Dialogue . .9 2.1.1 Response Generation . 10 2.1.2 Natural Language Understanding . 14 2.1.3 Spoken Dialogue . 17 2.2 Text Classification . 19 2.3 One Class Classification . 22 2.4 Speech Recognition . 23 2.5 Other Virtual Patients . 24 2.6 Conclusion . 25 viii 3. System Overview . 26 3.1 Data . 26 3.1.1 Annotation . 28 3.1.2 Core Data Set . 30 3.1.3 Enhanced Data Set . 32 3.1.4 Spoken Data Set . 32 3.1.5 Strict Relabeling . 35 3.2 System Description . 36 3.2.1 Front End Client . 37 3.2.2 Back End Web Service . 38 3.2.3 Response Model . 39 3.2.4 Miscellaneous . 43 3.2.5 Spoken Interface Extensions . 43 3.3 Challenges and Opportunities . 46 3.4 Conclusion . 49 4. Hybrid System Deployment and Out-of-Scope Handling . 50 4.1 Experiment 1: Baseline Reproduction . 51 4.1.1 Stacked CNN . 54 4.1.2 Training . 57 4.1.3 Baseline . 58 4.1.4 CNN Results . 58 4.1.5 Hybrid System . 61 4.2 Experiment 2: Prospective Replication . 64 4.2.1 Experimental Design . 65 4.2.2 Results and Discussion . 66 4.3 Experiment 3: Out-of-Scope Improvements . 74 4.3.1 Models . 77 4.3.2 Training and Evaluation . 79 4.3.3 Baseline . 82 4.3.4 Results . 83 4.3.5 Discussion . 85 4.4 Conclusion . 91 5. How Self-Attention Improves Rare Class Performance . 94 5.1 Introduction . 95 5.2 Related Work . 97 5.3 Task and Data . 99 ix 5.4 Experimental Design and Results . 100 5.5 Analysis . 104 5.5.1 Why did BERT perform less well? . 104 5.5.2 Analyzing the Self-attention RNN . 105 5.6 Conclusion . 111 6. Modality Transfer with Text-to-Phonetic Data Augmentation . 116 6.1 Introduction . 117 6.2 Related Work . 119 6.3 Model . 120 6.4 Experiments . 121 6.4.1 Data . 122 6.4.2 Experimental details . 123 6.5 Results . 124 6.6 Discussion . 126 6.7 Future Work . 127 7. Conclusion and Future Work . 129 Appendices 134 A. Label Sets . 134 A.1 Strict Relabeling Labels . 134 B. Table Definitions . 137 x List of Tables Table Page 3.1 An example conversation opening. 27 3.2 The implemented REST-like API. 40 3.3 Extensions to the back end API. 45 4.1 Mean accuracy across ten folds, with standard deviations. 58 4.2 Features used by the chooser. 61 4.3 Hybrid system results. 63 4.4 Summary statistics of collected data. 66 4.5 Raw accuracies of control and test. 67 4.6 Top changes in class frequency from core to enhanced data. 70 4.7 Adjusted accuracies of control and test. 73 4.8 New features used by the Three-way chooser. 80 4.9 Development accuracy with additional features . 83 4.10 Effects of retraining on test performance . 83 4.11 Test results for three-way choosers with retraining . 84 5.1 Dev set results comparing different models. ..
Load more

Recommended publications
  • Deep Learning for Dialogue Systems
    Deep Learning for Dialogue Systems Yun-Nung Chen Asli Celikyilmaz Dilek Hakkani-Tur¨ National Taiwan University Microsoft Research Google Research Taipei, Taiwan Redmond, WA Mountain View, CA [email protected] [email protected] [email protected] Abstract Goal-oriented spoken dialogue systems have been the most prominent component in todays vir- tual personal assistants, which allow users to speak naturally in order to finish tasks more effi- ciently. The advancement of deep learning technologies has recently risen the applications of neural models to dialogue modeling. However, applying deep learning technologies for build- ing robust and scalable dialogue systems is still a challenging task and an open research area as it requires deeper understanding of the classic pipelines as well as detailed knowledge of the prior work and the recent state-of-the-art work. Therefore, this tutorial is designed to focus on an overview of dialogue system development while describing most recent research for building dialogue systems, and summarizing the challenges, in order to allow researchers to study the po- tential improvements of the state-of-the-art dialogue systems. The tutorial material is available at http://deepdialogue.miulab.tw. 1 Tutorial Overview With the rising trend of artificial intelligence, more and more devices have incorporated goal-oriented spoken dialogue systems. Among popular virtual personal assistants, Microsoft’s Cortana, Apple’s Siri, Amazon Alexa, and Google Assistant have incorporated dialogue system modules in various devices, which allow users to speak naturally in order to finish tasks more efficiently. Traditional conversational systems have rather complex and/or modular pipelines.
    [Show full text]
  • A Survey on Dialogue Systems: Recent Advances and New Frontiers
    A Survey on Dialogue Systems: Recent Advances and New Frontiers Hongshen Cheny, Xiaorui Liuz, Dawei Yiny, and Jiliang Tangz yData Science Lab, JD.com zData Science and Engineering Lab, Michigan State University [email protected], [email protected],fxiaorui,[email protected] ABSTRACT May I know your name ? Dialogue systems have attracted more and more attention. Dialogue State NLU Recent advances on dialogue systems are overwhelmingly Tracking contributed by deep learning techniques, which have been employed to enhance a wide range of big data applications such as computer vision, natural language processing, and recommender systems. For dialogue systems, deep learning I am Robot. can leverage a massive amount of data to learn meaningful NLG Policy learning feature representations and response generation strategies, while requiring a minimum amount of hand-crafting. In this article, we give an overview to these recent advances on di- Figure 1: Traditional Pipeline for Task-oriented Systems. alogue systems from various perspectives and discuss some possible research directions. In particular, we generally di- vide existing dialogue systems into task-oriented and non- certain tasks (e.g. finding products, and booking accommo- task-oriented models, then detail how deep learning tech- dations and restaurants). The widely applied approaches to niques help them with representative algorithms and finally task-oriented systems are to treat the dialogue response as a discuss some appealing research directions that can bring pipeline as shown in Figure1. The systems first understand the dialogue system research into a new frontier. the message given by human, represent it as a internal state, then take some actions according to the policy with respect 1.
    [Show full text]
  • A Robust Dialogue System with Spontaneous Speech
    A Robust Dialogue System with Spontaneous Speech Understanding and Cooperative Response Toshihiko ITOH, Akihiro DENDA, Satoru KOGURE and Seiichi NAKAGAWA Department of Information and Computer Sciences Toyohashi University of Technology Tenpaku-cho, Toyohashi-shi, Aichi-ken, 441, Japan E-mail address • {itoh, akihiro, kogure, nakagawa}@slp.tutics.tut.ac.jp 1 Introduction 2 A Multi-Modal Dialogue System A spoken dialogue system that can understand spon- The domain of our dialogue system is "Mt. Fuji taneous speech needs to handle extensive range of sightseeing guidance (the vocabulary size is 292 speech in comparision with the read speech that has words for the recognizer and 948 words for the inter- been studied so far. The spoken language has looser preter, and the test-set word perplexity is 103)", The restriction of the grammar than the written language dialogue system is composed of 4 parts: Input by and has ambiguous phenomena such as interjections, speech recognizer and touch screen, graphical user ellipses, inversions, repairs, unknown words and so interface, interpreter, and response generator. The on. It must be noted that the recognition rate of the latter two parts are described in Sections 3 and 4. speech recognizer is limited by the trade-off between the looseness of linguistic constraints and recogni- 2.1 Spontaneous Speech Recognizer tion precision , and that the recognizer may out- The speech recognizer uses a frame-synchronous one put a sentence as recognition results which human pass Viterbi algorithm and Earley like parser for would never say. Therefore, the interpreter that re- context-free grammar, while using HMMs as sylla- ceives recognized sentences must cope not only with ble units.
    [Show full text]
  • Dialog Systems and Chatbots
    Speech and Language Processing. Daniel Jurafsky & James H. Martin. Copyright c 2018. All rights reserved. Draft of August 12, 2018. CHAPTER 25 Dialog Systems and Chatbots Les lois de la conversation sont en gen´ eral´ de ne s’y appesantir sur aucun ob- jet, mais de passer leg´ erement,` sans effort et sans affectation, d’un sujet a` un autre ; de savoir y parler de choses frivoles comme de choses serieuses´ The rules of conversation are, in general, not to dwell on any one subject, but to pass lightly from one to another without effort and without affectation; to know how to speak about trivial topics as well as serious ones; The 18th C. Encyclopedia of Diderot, start of the entry on conversation The literature of the fantastic abounds in inanimate objects magically endowed with sentience and the gift of speech. From Ovid’s statue of Pygmalion to Mary Shelley’s Frankenstein, there is something deeply moving about creating something and then having a chat with it. Legend has it that after finishing his sculpture Moses, Michelangelo thought it so lifelike that he tapped it on the knee and commanded it to speak. Per- haps this shouldn’t be surprising. Language is the mark conversation of humanity and sentience, and conversation or dialog dialog is the most fundamental and specially privileged arena of language. It is the first kind of language we learn as children, and for most of us, it is the kind of language we most commonly indulge in, whether we are ordering curry for lunch or buying spinach, participating in busi- ness meetings or talking with our families, booking air- line flights or complaining about the weather.
    [Show full text]
  • Survey on Chatbot Design Techniques in Speech Conversation Systems
    (IJACSA) International Journal of Advanced Computer Science and Applications, Vol. 6, No. 7, 2015 Survey on Chatbot Design Techniques in Speech Conversation Systems Sameera A. Abdul-Kader Dr. John Woods School of Computer Science and Electronic School of Computer Science and Electronic Engineering/University of Essex Colchester/ UK Engineering/University of Essex Colchester/ UK Diyala University/ Diyala/ Iraq Abstract—Human-Computer Speech is gaining momentum as selected works are compared with those used in Loebner-Prize a technique of computer interaction. There has been a recent Chatbots. The findings are discussed and conclusions are upsurge in speech based search engines and assistants such as drawn at the end. Siri, Google Chrome and Cortana. Natural Language Processing (NLP) techniques such as NLTK for Python can be applied to II. BACKGROUND analyse speech, and intelligent responses can be found by designing an engine to provide appropriate human like A. Human-Computer Speech interaction responses. This type of programme is called a Chatbot, which is Speech recognition is one of the most natural and sought the focus of this study. This paper presents a survey on the after techniques in computer and networked device interaction techniques used to design Chatbots and a comparison is made has only recently become possible (last two decades) with the between different design techniques from nine carefully selected advent of fast computing. papers according to the main methods adopted. These papers are representative of the significant improvements in Chatbots in the Speech is a sophisticated signal and happens at different last decade. The paper discusses the similarities and differences levels: “semantic, linguistic, articulatory, and acoustic” [3].
    [Show full text]
  • User Concepts for In-Car Speech Dialogue Systems and Their Integration Into a Multimodal Human-Machine Interface
    User Concepts for In-Car Speech Dialogue Systems and their Integration into a Multimodal Human-Machine Interface Von der Philosophisch-Historischen Fakultät der Universität Stuttgart zur Erlangung der Würde eines Doktors der Philosophie (Dr. phil.) genehmigte Abhandlung Vorgelegt von Sandra Mann aus Aalen Hauptberichter: Prof. Dr. Grzegorz Dogil Mitberichter: Apl. Prof. Dr. Bernd Möbius Tag der mündlichen Prüfung: 02.02.2010 Institut für Maschinelle Sprachverarbeitung der Universität Stuttgart 2010 3 Acknowledgements This dissertation developed during my work at Daimler AG Group Research and Advanced Engineering in Ulm, formerly DaimlerChrysler AG. Graduation was accomplished at the University of Stuttgart, Institute for Natural Language Processing (IMS) at the chair of Experimental Phonetics. I would like to thank Prof. Dr. Grzegorz Dogil from the IMS of the University of Stuttgart for supervising this thesis and supporting me in scientific matters. At this point I would also like to thank Apl. Prof. Dr. Bernd Möbius for being secondary supervisor. I also wish to particularly thank my mentors at Daimler AG, Dr. Ute Ehrlich and Dr. Susanne Kronenberg, for valuable advice on speech dialogue systems. They were always available to discuss matters concerning my research and gave constructive comments. Besides I would like to thank Paul Heisterkamp whose long-time experience in the field of human-machine interaction was very valuable to me. Special thanks go to all colleagues from the speech dialogue team, the recognition as well as the acoustics team. I very much acknowledge the good atmosphere as well as fruitful discussions, advice and criticism contributing to this thesis. I would especially like to mention Dr.
    [Show full text]
  • Influencing an Artificial Conversational Entity by Information Fusion
    University of West Bohemia Department of Computer Science and Engineering Univerzitn´ı8 30614 Plzeˇn Czech Republic Influencing an artificial conversational entity by information fusion Technical report Ing. Jarom´ırSalamon Technical report No. DCSE/TR-2020-04 June 2020 Distribution: public Technical report No. DCSE/TR-2020-04 June 2020 Influencing an artificial conversational entity by information fusion Technical report Ing. Jarom´ırSalamon Abstract The dissertation thesis proposes a new method of using an artificial conversational enti(later also dialogue system or chatbot) influenced by information or information fusion. This new method could potentially serve various purposes of use like fitness and well-being support, mental health support, mental illness treatment, and a like. To propose such a new method, one has to investigate two main topics. Whether it is possible to influence the dialogue system by information (fusion) and what kind of data needs to be collected and prepared for such influence. The dialogue system uses textual data from conversation to determine the context of human interaction and decide about next response. It presents correct behavior without external influence with data, but with the influence, the dialogue system needs to react without hiccups in the conversation adequately. The data which could be used for the dialogue system influencing can be a combina- tion of qualitative measure (from text extracted sentiment, from voice determined tone, from face revealed emotion), and measured quantitative values (wearable mea- sured heartbeat, on-camera correctly performed exercise, based on EEG found focus on activity). All the research found in the relation of those two topics is described in next more than 200 pages.
    [Show full text]
  • A Generative Dialogue System for Reminiscence Therapy
    Universitat Polit`ecnicade Catalunya Escola T`ecnicaSuperior d'Enginyeria de Telecomunicaci´ode Barcelona A Generative Dialogue System for Reminiscence Therapy Mariona Car´osRoca Advisors: Xavier Gir´o-i-Nieto,Petia Radeva A thesis submitted in fulfillment of the requirements for the Master in Telecommunications Engineering Barcelona, September 2019 Abstract With people living longer than ever, the number of cases with neurodegenerative diseases such as Alzheimer's or cognitive impairment increases steadily. In Spain it affects more than 1.2 million patients and it is estimated that in 2050 more than 100 million people will be affected. While there are not effective treatments for this terminal disease, therapies such as reminiscence, that stimulate memories of the patient's past are recommended, as they encourage the communication and produce mental and emotional benefits on the patient. Currently, reminiscence therapy takes place in hospitals or residences, where the therapists are located. Since people that receive this therapy are old and may have mobility difficulties, we present an AI solution to guide older adults through reminiscence sessions by using their laptop or smartphone. Our solution consists in a generative dialogue system composed of two deep learning archi- tectures to recognize image and text content. An Encoder-Decoder with Attention is trained to generate questions from photos provided by the user, which is composed of a pretrained Con- volution Neural Network to encode the picture, and a Long Short-Term Memory to decode the image features and generate the question. The second architecture is a sequence-to-sequence model that provides feedback to engage the user in the conversation.
    [Show full text]
  • A Context-Aware Conversational Agent in the Rehabilitation Domain
    future internet Article A Context-Aware Conversational Agent in the Rehabilitation Domain Thanassis Mavropoulos 1,*, Georgios Meditskos 1,*, Spyridon Symeonidis 1, Eleni Kamateri 1, Maria Rousi 1, Dimitris Tzimikas 2, Lefteris Papageorgiou 2, Christos Eleftheriadis 3, George Adamopoulos 4, Stefanos Vrochidis 1 and Ioannis Kompatsiaris 1 1 Centre for Research and Technology Hellas, 57001 Thessaloniki, Greece; [email protected] (S.S.); [email protected] (E.K.); [email protected] (M.R.); [email protected] (S.V.); [email protected] (I.K.) 2 Entranet Ltd., 54249 Thessaloniki, Greece; [email protected] (D.T.); [email protected] (L.P.) 3 ELBIS P.C., 57001 Thessaloniki, Greece; [email protected] 4 Evexia S.A., 63080 Thessaloniki, Greece; [email protected] * Correspondence: [email protected] (T.M.); [email protected] (G.M.) Received: 30 September 2019; Accepted: 30 October 2019; Published: 1 November 2019 Abstract: Conversational agents are reshaping our communication environment and have the potential to inform and persuade in new and effective ways. In this paper, we present the underlying technologies and the theoretical background behind a health-care platform dedicated to supporting medical stuff and individuals with movement disabilities and to providing advanced monitoring functionalities in hospital and home surroundings. The framework implements an intelligent combination of two research areas: (1) sensor- and camera-based monitoring to collect, analyse, and interpret people behaviour and (2) natural machine–human interaction through an apprehensive virtual assistant benefiting ailing patients. In addition, the framework serves as an important assistant to caregivers and clinical experts to obtain information about the patients in an intuitive manner.
    [Show full text]
  • A Dialogue-Act Taxonomy for a Virtual Coach Designed to Improve the Life of Elderly
    Multimodal Technologies and Interaction Article A Dialogue-Act Taxonomy for a Virtual Coach Designed to Improve the Life of Elderly César Montenegro 1,†, Asier López Zorrilla 2,†, Javier Mikel Olaso 2,†, Roberto Santana 1,*, Raquel Justo 2, Jose A. Lozano 1,3 and María Inés Torres 2,* 1 ISG-UPV/EHU, Faculty of Computer Science , Paseo Manuel sde Lardizabal 1, 20018 Donostia-San Sebastian, Gipuzkoa, Spain 2 SPIN-UPV/EHU, Department of Electrical and Electronics, Faculty of Science and Technology, Campus de Leioa, 48940 Leioa, Bizkaia, Spain 3 BCAM, Alameda de Mazarredo 14, 48009 Bilbao, Spain * Correspondence: [email protected] (R.S.); [email protected] (M.I.T.) † These authors contributed equally to this work. Received: 11 June 2019; Accepted: 4 July 2019; Published: 11 July 2019 Abstract: This paper presents a dialogue act taxonomy designed for the development of a conversational agent for elderly. The main goal of this conversational agent is to improve life quality of the user by means of coaching sessions in different topics. In contrast to other approaches such as task-oriented dialogue systems and chit-chat implementations, the agent should display a pro-active attitude, driving the conversation to reach a number of diverse coaching goals. Therefore, the main characteristic of the introduced dialogue act taxonomy is its capacity for supporting a communication based on the GROW model for coaching. In addition, the taxonomy has a hierarchical structure between the tags and it is multimodal. We use the taxonomy to annotate a Spanish dialogue corpus collected from a group of elder people.
    [Show full text]
  • Response Selection for End-To-End Retrieval-Based Dialogue Systems Basma El Amel Boussaha
    Response Selection for End-to-End Retrieval-Based Dialogue Systems Basma El Amel Boussaha To cite this version: Basma El Amel Boussaha. Response Selection for End-to-End Retrieval-Based Dialogue Systems. Computation and Language [cs.CL]. Université de Nantes (UN), 2019. English. tel-02926608 HAL Id: tel-02926608 https://hal.archives-ouvertes.fr/tel-02926608 Submitted on 31 Aug 2020 HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non, lished or not. The documents may come from émanant des établissements d’enseignement et de teaching and research institutions in France or recherche français ou étrangers, des laboratoires abroad, or from public or private research centers. publics ou privés. THÈSE DE DOCTORAT DE L’UNIVERSITE DE NANTES COMUE UNIVERSITE BRETAGNE LOIRE Ecole Doctorale N°601 Mathèmatique et Sciences et Technologies de l’Information et de la Communication Spécialité : Informatique Par Basma El Amel BOUSSAHA Response Selection for End-to-End Retrieval-Based Dialogue Systems Thèse présentée et soutenue à NANTES le 23 Octobre 2019 Unité de recherche : Laboratoire des Sciences du Numérique de Nantes (LS2N) Rapporteurs avant soutenance : Frédéric BÉCHET Professeur des universités, Aix Marseille Université, LIS Sophie ROSSET Directrice de recherche, Université Paris-Sud, LIMSI Composition du jury : Président : Yannick ESTÈVE Professeur des universités, Université d’Avignon, LIA Examinateurs : Frédéric BÉCHET Professeur des universités, Aix Marseille Université, LIS Sophie ROSSET Directrice de recherche, Université Paris-Sud, LIMSI Yannick ESTÈVE Professeur des universités, Université d’Avignon, LIA Dir.
    [Show full text]
  • Challenges in Designing Natural Language Interfaces for Complex Visual Models
    Challenges in Designing Natural Language Interfaces for Complex Visual Models Henrik Voigt1,2, Monique Meuschke1, Kai Lawonn1 and Sina Zarrieß2 1University of Jena 2University of Bielefeld [email protected] [email protected] Abstract In the VIS community, interaction with visual models plays an important role and natural lan- Intuitive interaction with visual models be- guage interaction represents a big part of it (Bacci comes an increasingly important task in the et al., 2020; Kumar et al., 2020; Srinivasan et al., field of Visualization (VIS) and verbal interac- 2020). Natural Language Interfaces (NLIs) that tion represents a significant aspect of it. Vice versa, modeling verbal interaction in visual support interactive visualizations based on lan- environments is a major trend in ongoing re- guage queries have found increasing interest in search in NLP. To date, research on Language recent research (Narechania et al., 2020; Yu and & Vision, however, mostly happens at the in- Silva, 2020; Fu et al., 2020). However, from an tersection of NLP and Computer Vision (CV), NLP point of view, the methods applied in these and much less at the intersection of NLP and recent interfaces, mostly rely on established meth- Visualization, which is an important area in ods for implementing semantic parsers that map Human-Computer Interaction (HCI). This pa- natural language instructions to symbolic data base per presents a brief survey of recent work on interactive tasks and set-ups in NLP and Vi- queries, which are consecutively visualized by a sualization. We discuss the respective meth- visualization pipeline. In this paper, we argue that ods, show interesting gaps and conclude by there is space for further support of intuitive interac- suggesting neural, visually grounded dialogue tion with visual models using state-of-the-art NLP modeling as a promising potential for NLIs for methods that would also pose novel and interesting visual models.
    [Show full text]