scispace - formally typeset
Search or ask a question
Proceedings Article

Lightly Supervised Learning of Procedural Dialog Systems

01 Jan 2013-pp 1669-1679
TL;DR: A novel approach is described that first automatically extracts task knowledge from instructions, then learns a dialog manager over this task knowledge to provide assistance, and can be integrated into a dialog system that provides effective help to users.
Abstract: Procedural dialog systems can help users achieve a wide range of goals. However, such systems are challenging to build, currently requiring manual engineering of substantial domain-specific task knowledge and dialog management strategies. In this paper, we demonstrate that it is possible to learn procedural dialog systems given only light supervision, of the type that can be provided by non-experts. We consider domains where the required task knowledge exists in textual form (e.g., instructional web pages) and where system builders have access to statements of user intent (e.g., search query logs or dialog interactions). To learn from such textual resources, we describe a novel approach that first automatically extracts task knowledge from instructions, then learns a dialog manager over this task knowledge to provide assistance. Evaluation in a Microsoft Office domain shows that the individual components are highly accurate and can be integrated into a dialog system that provides effective help to users.

Content maybe subject to copyright    Report

Citations
More filters
Proceedings ArticleDOI
01 Jun 2015
TL;DR: This paper extends the constrained lattice training of T¨¨ om et al. (2013) to non-linear conditional random fields in which latent variables mediate between observations and labels and shows that this method gives significant improvement over strong supervised and weakly-supervised baselines.
Abstract: In this paper, we apply a weakly-supervised learning approach for slot tagging using conditional random fields by exploiting web search click logs. We extend the constrained lattice training of T¨¨ om et al. (2013) to non-linear conditional random fields in which latent variables mediate between observations and labels. When combined with a novel initialization scheme that leverages unlabeled data, we show that our method gives significant improvement over strong supervised and weakly-supervised baselines.

37 citations


Cites background from "Lightly Supervised Learning of Proc..."

  • ...Web click logs present an opportunity to learn semantic tagging models from large-scale and naturally occurring user interaction data (Volkova et al., 2013)....

    [...]

Proceedings ArticleDOI
13 Sep 2014
TL;DR: This investigation sought to make good on a proof-of-concept where people interact with a social robot whereby the robot helps people to be more humanly creative, and results supported the proof of concept.
Abstract: This research builds on the UbiComp vision of systems that do not do things for people but engage people in their computational environment so that people can do things for themselves better. In this investigation, we sought to make good on a proof-of-concept where people interact with a social robot whereby the robot helps people to be more humanly creative. Twenty seven participants interacted with ATR's humanoid robot Robovie (through a WoZ interface) in a creativity task. Results supported our proof of concept insofar as 100% of the participants generated creative ideas, and 63% incorporated the robot's ideas into their own ideas for their creative output. Of the participants who had the highest creativity scores, 83% incorporated the robot's ideas into their own. Discussion focuses on next steps toward building the Natural Language Processing system, and integrating the system into a more extensive networked UbiComp environment.

9 citations


Cites background or methods from "Lightly Supervised Learning of Proc..."

  • ...This approach has been shown to be successful for Zettlemoyer’s work in the Microsoft Office domain [19], where the task knowledge came from Windows help pages that describe how to solve the users’ problems, including for example what information needs to be gathered from the user, and the dialogue manager used search query logs to learn how to best understand user statements and respond to them....

    [...]

  • ...Building on Zettlemoyer’s research [19], a promising approach would be to build a NLP architecture that has two main components: a library of task knowledge that includes examples of possible conversational interactions and a dialogue manager that uses the task knowledge to decide how to best interact with users....

    [...]

Patent
Hao Chen1, Qi Cheng Li1, Shao Chun Li1, Jie Ma1, Li Jun Mei1 
10 Feb 2017
TL;DR: In this paper, a machine learning model is used to improve procedure dialogs through knowledge mining within a reinforcement learning framework, and user interactions with the model are monitored and used to update the model.
Abstract: Procedure dialogs are improved through knowledge mining within a reinforcement learning framework. Taking an existing procedure dialog as input, a machine learning model is generated. User interactions with the machine learning model are monitored and used to update the machine learning model. The updates to the machine learning model are applied to the existing procedure dialog for review and revision by subject matter experts.

3 citations

Dissertation
28 Apr 2017
TL;DR: L'intelligence artificielle est la discipline de recherche d'imitation ou de remplacement de fonctions cognitives humaines, mais aussi l'inconvenient de n'aquerir qu'un nombre tres limite d'exemples, en analysant et reduisant progressivement les biais induits.
Abstract: L'intelligence artificielle est la discipline de recherche d'imitation ou de remplacement de fonctions cognitives humaines. A ce titre, l'une de ses branches s'inscrit dans l'automatisation progressive du processus de programmation. Il s'agit alors de transferer de l'intelligence ou, a defaut de definition, de transferer de la charge cognitive depuis l'humain vers le systeme, qu'il soit autonome ou guide par l'utilisateur. Dans le cadre de cette these, nous considerons les conditions de l'evolution depuis un systeme guide par son utilisateur vers un systeme autonome, en nous appuyant sur une autre branche de l'intelligence artificielle : l'apprentissage artificiel. Notre cadre applicatif est celui de la conception d'un assistant operationnel incremental, c'est-a-dire d'un systeme capable de reagir a des requetes formulees par l'utilisateur en adoptant les actions appropriees, et capable d'apprendre a le faire. Pour nos travaux, les requetes sont exprimees en francais, et les actions sont designees par les commandes correspondantes dans un langage de programmation (ici, R ou bash). L'apprentissage du systeme est effectue a l'aide d'un ensemble d'exemples constitue par les utilisateurs eux-memes lors de leurs interactions. Ce sont donc ces derniers qui definissent, progressivement, les actions qui sont appropriees pour chaque requete, afin de rendre le systeme de plus en plus autonome. Nous avons collecte plusieurs ensembles d'exemples pour l'evaluation des methodes d'apprentissage, en analysant et reduisant progressivement les biais induits. Le protocole que nous proposons est fonde sur l'amorcage incremental des connaissances du systeme a partir d'un ensemble vide ou tres restreint. Cela presente l'avantage de constituer une base de connaissances tres representative des besoins des utilisateurs, mais aussi l'inconvenient de n'aquerir qu'un nombre tres limite d'exemples. Nous utilisons donc, apres examen des performances d'une methode naive, une methode de raisonnement a partir de cas : le raisonnement par analogie formelle. Nous montrons que cette methode permet une precision tres elevee dans les reponses du systeme, mais egalement une couverture relativement faible. L'extension de la base d'exemples par analogie est exploree afin d'augmenter la couverture des reponses donnees. Dans une autre perspective, nous explorons egalement la piste de rendre l'analogie plus tolerante au bruit et aux faibles differences en entree en autorisant les approximations, ce qui a egalement pour effet la production de reponses incorrectes plus nombreuses. La duree d'execution de l'approche par analogie, deja de l'ordre de la seconde, souffre beaucoup de l'extension de la base et de l'approximation. Nous avons explore plusieurs methodes de segmentation des sequences en entree afin de reduire cette duree, mais elle reste encore le principal obstacle a contourner pour l'utilisation de l'analogie formelle dans le traitement automatique de la langue. Enfin, l'assistant operationnel incremental fonde sur le raisonnement analogique a ete teste en condition incrementale simulee, afin d'etudier la progression de l'apprentissage du systeme au cours du temps. On en retient que le modele permet d'atteindre un taux de reponse stable apres une dizaine d'exemples vus en moyenne pour chaque type de commande. Bien que la performance effective varie selon le nombre total de commandes considerees, cette propriete ouvre sur des applications interessantes dans le cadre incremental du transfert depuis un domaine riche (la langue naturelle) vers un domaine moins riche (le langage de programmation).

2 citations

Journal ArticleDOI
TL;DR: A proposal to automatically generate the dialogue rules from a dialogue corpus through the use of evolving algorithms and adapt the rules according to the detected user intention, which is an efficient way for adapting a set of dialogue rules considering user utterance clusters.
Abstract: Conversational systems have become an element of everyday life for billions of users who use speech‐based interfaces to services, engage with personal digital assistants on smartphones, social media chatbots, or smart speakers. One of the most complex tasks in the development of these systems is to design the dialogue model, the logic that provided a user input selects the next answer. The dialogue model must also consider mechanisms to adapt the response of the system and the interaction style according to different groups and user profiles. Rule‐based systems are difficult to adapt to phenomena that were not taken into consideration at design‐time. However, many of the systems that are commercially available are based on rules, and so are the most widespread tools for the development of chatbots and speech interfaces. In this article, we present a proposal to: (a) automatically generate the dialogue rules from a dialogue corpus through the use of evolving algorithms, (b) adapt the rules according to the detected user intention. We have evaluated our proposal with several conversational systems of different application domains, from which our approach provided an efficient way for adapting a set of dialogue rules considering user utterance clusters.

2 citations

References
More filters
Book
01 Nov 2008
TL;DR: Numerical Optimization presents a comprehensive and up-to-date description of the most effective methods in continuous optimization, responding to the growing interest in optimization in engineering, science, and business by focusing on the methods that are best suited to practical problems.
Abstract: Numerical Optimization presents a comprehensive and up-to-date description of the most effective methods in continuous optimization. It responds to the growing interest in optimization in engineering, science, and business by focusing on the methods that are best suited to practical problems. For this new edition the book has been thoroughly updated throughout. There are new chapters on nonlinear interior methods and derivative-free methods for optimization, both of which are used widely in practice and the focus of much current research. Because of the emphasis on practical methods, as well as the extensive illustrations and exercises, the book is accessible to a wide audience. It can be used as a graduate text in engineering, operations research, mathematics, computer science, and business. It also serves as a handbook for researchers and practitioners in the field. The authors have strived to produce a text that is pleasant to read, informative, and rigorous - one that reveals both the beautiful nature of the discipline and its practical side.

17,420 citations

Proceedings Article
19 Jun 2011
TL;DR: A novel data collection framework is presented that produces highly parallel text data relatively inexpensively and on a large scale that allows for simple n-gram comparisons to measure both the semantic adequacy and lexical dissimilarity of paraphrase candidates.
Abstract: A lack of standard datasets and evaluation metrics has prevented the field of paraphrasing from making the kind of rapid progress enjoyed by the machine translation community over the last 15 years. We address both problems by presenting a novel data collection framework that produces highly parallel text data relatively inexpensively and on a large scale. The highly parallel nature of this data allows us to use simple n-gram comparisons to measure both the semantic adequacy and lexical dissimilarity of paraphrase candidates. In addition to being simple and efficient to compute, experiments show that these metrics correlate highly with human judgments.

934 citations


"Lightly Supervised Learning of Proc..." refers background in this paper

  • ..., 2009), and learning paraphrases from crowdsourced captions of video snippets (Chen and Dolan, 2011)....

    [...]

Proceedings ArticleDOI
03 Oct 2010
TL;DR: S soylent, a word processing interface that enables writers to call on Mechanical Turk workers to shorten, proofread, and otherwise edit parts of their documents on demand, and the Find-Fix-Verify crowd programming pattern, which splits tasks into a series of generation and review stages.
Abstract: This paper introduces architectural and interaction patterns for integrating crowdsourced human contributions directly into user interfaces. We focus on writing and editing, complex endeavors that span many levels of conceptual and pragmatic activity. Authoring tools offer help with pragmatics, but for higher-level help, writers commonly turn to other people. We thus present Soylent, a word processing interface that enables writers to call on Mechanical Turk workers to shorten, proofread, and otherwise edit parts of their documents on demand. To improve worker quality, we introduce the Find-Fix-Verify crowd programming pattern, which splits tasks into a series of generation and review stages. Evaluation studies demonstrate the feasibility of crowdsourced editing and investigate questions of reliability, cost, wait time, and work time for edits.

814 citations


"Lightly Supervised Learning of Proc..." refers background in this paper

  • ...Crucially, though, this annotation work can be carried out by non-specialists, and could even be crowdsourced (Bernstein et al., 2010)....

    [...]

Posted Content
Eric Horvitz1, Jack Breese1, David Heckerman1, David O. Hovel1, Koos Rommelse1 
TL;DR: This work reviews work on Bayesian user models that can be employed to infer a user's needs by considering a users' background, actions, and queries and proposes an overall architecture for an intelligent user interface.
Abstract: The Lumiere Project centers on harnessing probability and utility to provide assistance to computer software users. We review work on Bayesian user models that can be employed to infer a users needs by considering a user's background, actions, and queries. Several problems were tackled in Lumiere research, including (1) the construction of Bayesian models for reasoning about the time-varying goals of computer users from their observed actions and queries, (2) gaining access to a stream of events from software applications, (3) developing a language for transforming system events into observational variables represented in Bayesian user models, (4) developing persistent profiles to capture changes in a user expertise, and (5) the development of an overall architecture for an intelligent user interface. Lumiere prototypes served as the basis for the Office Assistant in the Microsoft Office '97 suite of productivity applications.

751 citations


"Lightly Supervised Learning of Proc..." refers background in this paper

  • ..., 2011), modeling users goals in a Windows domain (Horvitz et al., 1998), learning from conversational interaction (Artzi and Zettlemoyer, 2011), learning to sportscast (Chen and Mooney, 2011), learning from event streams (Liang et al....

    [...]

  • ...…other grounded language problems, including understanding game strategy guides (Branavan et al., 2011), modeling users goals in a Windows domain (Horvitz et al., 1998), learning from conversational interaction (Artzi and Zettlemoyer, 2011), learning to sportscast (Chen and Mooney, 2011),…...

    [...]

Proceedings Article
Eric Horvitz1, Jack Breese1, David Heckerman1, David O. Hovel1, Koos Rommelse1 
24 Jul 1998
TL;DR: The Lumiere Project as discussed by the authors harnesses probability and utility to provide assistance to computer software users by considering a user's background, actions, and queries, and develops persistent profiles to capture changes in user's expertise.
Abstract: The Lumiere Project centers on harnessing probability and utility to provide assistance to computer software users. We review work on Bayesian user models that can be employed to infer a user's needs by considering a user's background, actions, and queries. Several problems were tackled in Lumiere research, including (1) the construction of Bayesian models for reasoning about the time-varying goals of computer users from their observed actions and queries, (2) gaining access to a stream of events from software applications, (3) developing a language for transforming system events into observational variables represented in Bayesian user models, (4) developing persistent profiles to capture changes in a user's expertise, and (5) the development of an overall architecture for an intelligent user interface. Lumiere prototypes served as the basis for the Ofice Assistant in the Microsoft Office '97 suite of productivity applications.

723 citations