Showing papers by "Amélie Cordier published in 2013"

Trace-Based Reasoning - Modeling Interaction Traces for Reasoning on Experiences.

Abstract: This paper addresses Trace-Based Reasoning (TBR) by using Case-Based Reasoning (CBR) as a descriptive framework. TBR is a reasoning paradigm in which inferences are made on specific objects called traces. Traces are sequential records of events observed and stored during an interactive process. We report two contributions. First, we propose a review of the current researches related to TBR. Then, we compare CBR and TBR. From this comparison, we show that the exploitation of traces instead of cases as knowledge sources raises very specific challenges. More precisely, new methods for defining similarity measures and for performing adaptation of traces are required. These new methods have to take into account the sequential properties of traces. In the discussion, we emphasis the benefits of using traces as a knowledge container in a reasoning process and we pinpoint promising applications of TBR.

TStore: A Trace-Base Management System using Finite-State Transducer Approach for Trace Transformation

Abstract: This paper presents TStore, a Trace-Base Management System (TBMS) handling storage, transformation and exploitation of traces collected by external applications. As the experiments reported in this paper demonstrate, TStore brings a solution to performances and storage issues usually encountered by TBMS. To exploit traces, transformations are used. TStore provides predefined transformation functions as well as a customized transformation based on Finite State Transducers (FST), which are also presented in this paper.

Similarity Measures to Compare Episodes in Modeled Traces

Abstract: This paper reports on a similarity measure to compare episodes in modeled traces. A modeled trace is a structured record of observations captured from users’ interactions with a computer system. An episode is a sub-part of the modeled trace, describing a particular task performed by the user. Our method relies on the definition of a similarity measure for comparing elements of episodes, combined with the implementation of the Smith-Waterman Algorithm for comparison of episodes. This algorithm is both accurate in terms of temporal sequencing and tolerant to noise generally found in the traces that we deal with. Our evaluations show that our approach offers quite satisfactory comparison quality and response time. We illustrate its use in the context of an application for video sequences recommendation.

Building a trace-based system for real-time strategy game traces

Abstract: We describe the conception of a visualization and transformation tool for traces of the real-time strategy (RTS) computer game StarCraft. The development of our tool StarTrace is driven by the domain the traces originate from as well as the observable elements those traces contain. We elaborate on those influences, which also include both the structure of the existing game traces and the requirement to use these traces to improve the performance of an machine learning (ML) agent that attempts to learn to play parts of the game.

Collecting interaction traces in distributed semantic wikis

Abstract: In the Kolflow project, our general objective is to develop an assistance engine suitable for distributed applications. In order to provide contextualized and relevant assistance, we feed the assistance engine with interaction traces. Interaction traces record events occurring while users are interacting with applications. These traces become containers of valuable knowledge to providing assistance. Collecting interaction traces is a challenging issue that has been thoroughly studied in the context of local applications. In contrast, few approaches focus on collecting interaction traces in distributed applications. Yet, when applications are distributed, collecting interaction traces is even more challenging because new difficulties arise, such as data synchronization and multi-synchronous collaboration. In this paper, we propose a model and a tool for collecting traces in a distributed environment. The originality of the model is that it is tailored to fit distributed applications. We implemented the model in Collectra, a tool to collect interaction traces in distributed web applications. Collectra collects interaction traces and stores them in a dedicated trace-base management system. We report on the experiments we have conducted in order to evaluate performances of Collectra (both response time and memory space). Results of the experiments show that Collectra performs well and that it can be used to support the assistance tasks carried out by the assistance engine.

Inferring Actions and Observations from Interactions

Abstract: This study follows the Radial Interactionism (RI) cognitive modeling paradigm introduced previously by Georgeon and Aha (2013). An RI cognitive model uses sensorimotor interactions as primitives-—instead of observations and actions-—to represent Piagetian (1955) sensorimotor schemes. Constructivist epistemology suggests that sensorimotor schemes precede perception and knowledge of the external world. Accordingly, this paper presents a learning algorithm for an RI agent to construct observations, actions, and knowledge of rudimentary entities, from spatio- sequential regularities observed in the stream of sensorimotor interactions. Results show that the agent learns to categorize entities on the basis of the interactions that they afford, and appropriately enact sequences of interactions adapted to categories of entities. This model explains rudimentary goal construction by the fact that entities that afford desirable interactions become desirable destinations to reach.

Mesures de Similarité pour Comparer des Épisodes dans les Traces Modélisées

Abstract: Cet article rend compte d'une mesure de similarite pour comparer des episodes de traces modelisees. Une trace modelisee est un enregistrement structure d'observations capturees a partir des interactions avec les utilisateurs d'un systeme informatique. Un episode est une sous-partie de la trace modelisee, decrivant une tâche particuliere executee par l'utilisateur. Notre methode se base sur la definition de la mesure de similarite pour comparer des elements d'episodes. Combine avec l'implementation de l'algorithme de Smith-Waterman pour la comparaison des episodes. Cet algorithme est a la fois precis en termes de sequence temporelle et tolerante au bruit trouves generalement dans les traces que nous traitons. Nos evaluations montrent que notre approche offre une qualite tout a fait satisfaisante de comparaison et en temps de reponse. Nous illustrons son utilisation dans le cadre d'une application de recommandation de sequences de video.

Collaborative Knowledge Acquisition under Control of a Non-Regression Test System

Abstract: This paper introduces BeGoood, a generic system for man- aging non-regression tests on knowledge-bases. BeGoood is a system al- lowing to define test plans in order to monitor the evolution of knowledge- bases. Any system answering queries by providing results in the form of set of strings can be tested with BeGoood. BeGoood has been devel- oped following a REST architecture and is independent of any applica- tion domain. This paper describes the architecture of the system and gives a use case to illustrate how BeGoood is able to manage a collab- orative knowledge evolution in the framework of a case-based reasoning system.

Modéliser les traces d'interaction pour raisonne a partir de l'expérience tracée ?

Abstract: Dans cet article, nous nous interessons a la problematique de l’ingenierie des connaissances dans le contexte particulier du Raisonnement a Partir de l’Experience Tracee (RaPET). Le RaPET est un paradigme de raisonnement dans lequel les sources de connaissances principales sont les traces d’interactions modelisees. Les traces modelisees s’averent etre des representations particulierement adaptees a la fois pour collecter facilement des experiences et pour accompagner l’ingenierie des connaissances a partir de ces experiences. Les traces possedent les bonnes proprietes que l’on attend d’un outil de representation des connaissances : souples, riches, expressives et calculables. Afin de mettre en lumiere les problematiques d’ingenierie des connaissances soulevees par l’utilisation des traces comme sources de connaissances, nous proposons une rapide ´etude du RaPET en utilisant le cadre descriptif bien connu du raisonnement a partir de cas (RaPC). Nous discutons ensuite des defis souleves et montrons les benefices attendus a developper et a utiliser ces representations.

Cooking with computers, a winning recipe!

Abstract: In our daily lives, cooking becomes more and more important. In many countries, you just have to switch television on to see the proliferation of cooking shows of all kinds: live cooking classes, cooking at home challenges, cooking contests (for everyone or for professionals), short recipes demos, etc. This witnesses two emergent phenomena: cooking gains a growing interest as a leisure, and people gain awareness of the role of food in our daily lives as well as the many reasons why should ”eat well”. In academic research, this trend is also observed, and many funding are related to well-being, health, nutrition, and other issues related to food. In many research fields, cooking is present: smart cooking appliances, cooking robots, vision techniques to recognize ingredients, and so on. In the areas of Web and artificial intelligence, the challenges are also numerous. How to represent recipes? How to adapt recipes to meet the needs of users? How to take into account the social aspects of cooking? Can we create social recipes books? In this presentation we will provide an overview of the challenges that cooking raises for computer science and information technology. We start by introducing the Taaable project, a website that makes use of case-based reasoning to adapt recipes. From that presentation, we draw a lessons learned while working on this project. Then we expand the scope of our study to other projects and initiatives combining computer and cooking. We adopt different points of view in that study, depending on the projects and their goals. We focus on the many issues addressed by these initiatives, such as representation of cooking recipes, suggestions, recommendations, dietary practices, personalization, social side, creativity, etc. Finally, we will review the prospects for research in this field and discuss current initiatives.