(mouse over papers to see full abstracts)
Adapting numerical representations of lung contours using Case-Based Reasoning and Artificial Neural Networks
In case of a radiological emergency situation involving accidental human exposure, a dosimetry evaluation must be established as soon as possible. In most cases, this evaluation is based on numerical representations and models of subjects. Unfortunately, personalised and realistic human representations are often unavailable for the exposed subjects. However, accuracy of treatment depends on the similarity of the phantom to the subject. The EquiVox platform (Research of Equivalent Voxel phantom) developed in this study uses Case-Based Reasoning principles to retrieve and adapt, from among a set of existing phantoms, the one to represent the subject. This paper introduces the EquiVox platform and Artificial Neural Networks developed to interpolate the subject’s 3D lung contours. The results obtained for the choice and construction of the contours are presented and discussed.
A Case-Based Approach to Cross Domain Sentiment Classification
This paper considers the task of sentiment classification of subjective text across many domains, in particular on scenarios where no in-domain data is available. Motivated by the more general applicability of such methods, we propose an extensible approach to sentiment classification that leverages sentiment lexicons and out-of-domain data to build a case-based system where solutions to past cases are reused to predict the sentiment of new documents from an unknown domain. In our approach the case representation uses a set of features based on document statistics, while the case solution stores sentiment lexicons employed on past predictions allowing for later retrieval and reuse on similar documents. The case-based nature of our approach also allows for future improvements since new lexicons and classification methods can be added to the case base as they become available. On a cross domain experiment our method has shown robust results when compared to a baseline single-lexicon classifier where the lexicon has to be pre-selected for the domain in question.
Case-Based Project Scheduling
This paper presents a new approach for solving the Resource-Constrained Project Scheduling Problem using Case-Based Reasoning in a constructive way. Given a project to be scheduled our method retrieves similar projects scheduled in the past, selects the most similar project, and reuses as much as possible from the old solution to build a schedule for the project at hand. The result of this process is a partial schedule that is later extended and revised to produce a complete and valid schedule by a modified version of the Serial Schedule Generation Scheme. We present experimental results showing that our approach works well under reasonable assumptions. Finally, we describe several ways to modify our algorithm in the future so as to obtain even better results.
Opponent Type Adaptation for Case-Based Strategies in Adversarial Games
We describe an approach for producing exploitive and adaptive case-based strategies in adversarial games. We describe how adaptation can be applied to a precomputed, static case-based strategy in order to allow the strategy to rapidly respond to changes in an opponent's playing style. The exploitive strategies produced by this approach tend to hover around a precomputed solid strategy and adaptation is applied
directly to the precomputed strategy once enough information has been gathered to classify the current opponent type. The use of a precomputed, seed strategy avoids performance degradation that can take place when little is known about an opponent. This allows our approach an advantage over other exploitive strategies whose playing decisions rely on large, individual opponent models constructed from scratch. We evaluate the approach within the experimental domain of two-player Limit Texas Hold'em poker.
A Case-Based Solution to the Cold-Start Problem in Group Recommenders
We extend a group recommender system with a case base of previous group recommendation events. We show that this offers a potential solution to the cold-start problem. Suppose a group recommendation is sought but one of the group members is a new user who has few item ratings. We can copy ratings into this user's profile from the profile of the most similar user in the most similar group from the case base. In other words, we copy ratings from a user who played a similar role in some previous group event. We show that copying in this way, i.e. conditioned on groups, is superior to copying nothing and also superior to copying ratings from the most similar user known to the system.
Exploiting Extended Search Sessions for Recommending Search Experiences in the Social Web
HeyStaks is a case-based social search system that allows users to create and share case bases of search experiences (called staks) and uses these staks as the basis for result recommendations at search time. These recommendations are added to conventional results from Google and Bing so that searchers can benefit from more focused results from people they trust on topics that matter to them. An important point of friction in HeyStaks is the need for searchers to select their search context (that is, their active stak) at search time. In this paper we extend previous work that attempts to eliminate this friction by automatically recommending an active stak based on the searchers context (query terms, Google results, etc.) and demonstrate significant improvements in stak recommendation accuracy.
GENA: A Case-Based Approach to the Generation of Audio-visual Narratives
This paper presents GENA, a case-based reasoning system capable of generating audio-visual narratives by drawing from previously annotated content. Broadcast networks spend large amount of resources in covering many events and many different types of audiences. However, it is not reasonable for them to cover smaller events or audiences, for which the cost would be greater than the potential benefits. For that reason, it is interesting to design systems that could automatically generate summaries, or personalized news shows for these smaller events or audiences. GENA was designed in collaboration with Televisio de Catalunya (the public Catalan broadcaster) precisely to address this problem. This paper describes GENA, and the techniques that were designed to address the complexities of the problem of generating audio-visual narrative. We also present an experimental evaluation in the domain of sports.
Diverse Plan Generation by Adapting Episodic Knowledge and by Using Complete Models to Generate Solutions from Scratch: A Comparative Study
Plan diversity has been explored in case-based planning, which relies on episodic knowledge, and in first-principles planning, which relies on complete planning domain models. Comparisons of these two approaches to planning have so far not addressed their relative merits with regard to obtaining diverse plans. We contrast two systems for generating diverse plans (a case-based planner and a first-principles planner), identifying their relative strengths and weaknesses, and perform a comparative experimental evaluation of them on a real-time strategy game domain.
A Case-based Approach to Mutual Adaptation of Taxonomic Ontologies
We present a general framework for addressing the problem of semantic intelligibility among artificial agents based on concepts integral to the case-based reasoning research program. For this purpose, we define case-based semiotics (\cbs) (based on the well known notion of the semiotic triangle) as the model that defines semantic intelligibility. We show how traditional CBR notions like transformational adaptation can be used in the problem of two agent achieving mutual intelligibility over a collection of concepts (defined in \cbs).
Feature Weighting and Confidence based Prediction for Case Based Reasoning Systems
The quality of the cases maintained in a casebase has a direct
influence on the quality of the proposed solutions. The presence
of cases that do not conform to the similarity hypothesis decreases the
alignment of the casebase and often degrades the performance of a CBR
system. It is therefore important to find out the suitability of each case
for the application of CBR and associate a solution with a certain degree
of confidence. Feature weighting is another important aspect that
determines the success of a system, as the presence of irrelevant and redundant
attributes also result in incorrect solutions. We explore these
problems in conjunction with a real-world CBR application called InfoChrom.
It is used to predict the values of several soil nutrients based
on features extracted from a chromatogram image of a soil sample. We
propose novel feature weighting techniques based on alignment, as well
as a new alignment and confidence measure as potential solutions. The
hypotheses are evaluated on UCI datasets and the casebase of Infochrom
and show promising results.
Natural Language Generation through Case-based Text Modification
Natural Language Generation (NLG) is one of the longstanding problems in Artificial Intelligence. In this paper, we focus on a subproblem in NLG, namely "surface realization through text modification": given a source sentence and a desired change, produce a grammatically correct and semantically coherent sentence that implements the desired change. Text modification has many applications within text generation like interactive narrative systems, where stories tailored to specific users are generated by adapting or instantiating a pre-authored story. We present a case-based approach where cases correspond to pairs of sentences implementing specific modifications. We describe our retrieval, adaptation and revise procedures. The main contribution of this paper is an approach to perform case-adaptation in textual domains.
Learning and Reusing Goal-Specific Policies for Goal-Driven Autonomy
In certain adversarial environments, reinforcement learning (RL) techniques require a prohibitively large number of episodes to learn a high-performing strategy for action selection. For example, Q-learning is particularly slow to learn a policy to win complex strategy games. We propose GRL, the first GDA system capable of learning and reusing goal-specific policies. GRL is a case-based goal-driven autonomy (GDA) agent embedded in the RL cycle. GRL acquires and reuses cases that capture episodic knowledge about an agent’s (1) expectations, (2) goals to pursue when these expectations are not met, and (3) actions for achieving these goals in given states. Our hypothesis is that, unlike RL, GRL can rapidly fine-tune strategies by exploiting the episodic knowledge captured in its cases. We report performance gains versus a state-of-the-art GDA agent and an RL agent for challenging tasks in two real-time video game domains.
Developing Case-Based Reasoning Applications using myCBR 3
This paper presents the Open Source tool myCBR which has been re-implemented as standalone application with a designated application programming interface that can be used as plug-in for various applications. We will introduce how knowledge according to Richter's knowledge containers can be modeled and how myCBR has been successfully applied within various applications. Especially we introduce novel features of myCBR that support knowledge engineers developing more comprehensive applications making use of existing knowledge such as Linked Data or User Generated Content. The applications presented in this paper present the high variety how CBR can be applied for web-based and mobile technologies as well as configuration, diagnostic or decision support tasks.
Case-Based Aggregation of Preferences for Group Recommenders
We extend a group recommender system with a case base of previous group recommendation events. We show that this offers a new way of aggregating the predicted ratings of the group members. Using user-user similarity, we align individuals from the active group with individuals from the groups in the cases. Then, using item-item similarity, we transfer the preferences of the groups in the cases over to the group that is seeking a recommendation. The advantage of a case-based approach to preference aggregation is that it does not require us to commit to a model of social behaviour and to find a way to express that model in a set of formulae. Rather, aggregation of predicted ratings will be a lazy and local generalization (in the spirit of CBR) of the behaviours captured by the neighbouring cases in the case base.
A competitive measure to assess the similarity between two time series
Time series are ubiquitous, and a measure to assess their similarity is a core part of many systems, including case-based reasoning systems. Although several proposals have been made, still the more robust and reliable time series similarity measures are the classical ones, introduced long time ago. In this paper we propose a new approach to time series similarity based on the costs of iteratively jumping (or moving) between the sample values of two time series. We show that this approach can be very competitive when compared against the aforementioned classical measures. In fact, extensive experiments show that it can be statistically significantly superior for a number of data sources. Since the approach is also computationally simple, we foresee its application as an alternative off-the-shelf tool to be used in many case-based reasoning systems dealing with time series.
Adapting Spatial and Temporal Cases
Qualitative algebras form a family of languages mainly used to represent knowledge depending on space or time. This paper proposes an approach to adapt cases represented in such an algebra. A spatial example in agronomy and a temporal example in cooking are given. The idea behind this adaptation approach is to apply a substitution and then repair potential inconsistencies, thanks to belief revision on qualitative algebras.
Retrieval and Clustering for Business Process Monitoring: Results and Improvements
Business process monitoring is a set of activities for organizing process instance logs and for highlighting non-compliances and adaptations with respect to the default process schema. Such activities typically serve as the starting point for a-posteriori log analyses.
In recent years, we have implemented a tool for supporting business process monitoring, which allows to retrieve traces of process execution similar to the current one. Moreover, it supports an automatic organization of the trace database content through the application of clustering techniques.
Retrieval and clustering rely on a distance definition able to take into account temporal information in traces.
In this paper, we report on such a tool, and present the newest experimental results.
Moreover, we introduce our recent research directions, that aim at improving the tool performances, usability and visibility with respect to the scientific community.
Specifically, we propose a methodology for avoiding exhaustive search in the trace database, by identifying promising regions of the search space, in order to reduce computation time.
Moreover, we describe how our work is being incorporated as a plug-in in ProM, an open source framework for process mining and process analysis.
Eager User Classification for Accessibility-Based CCBR Question Selection
The CCBR question selection is often guided by the discriminativeness of questions, to minimize dialog length for retrieval. However, there is no guarantee that users will always be able or willing to answer the most discriminative questions. Consequently, user characteristics may play an important role in determining the value of particular questions. This paper presents a method for customizing CCBR question selection to reflect the types of questions the user is likely to answer. Based on background knowledge about response probabilities for different questions by different user groups, the approach classifies new users, based on the questions they choose to answer, and applies the resulting classifications to predict their likelihood of answering particular questions in the future. The paper discusses Accessibility Influenced Attribute Selection (AIAS), which balances response likelihood against information gain to select questions likely to provide useful information to the CCBR system. It presents a refinement of AIAS, AIAS+, which also considers the potential of questions to discriminate between candidate user groups, with the aim of speeding user classification to enable better selection of following questions. Experiments with simulated users show improvement over a baseline method, and experiments in simulated domains illuminate the domain characteristics under which the method is expected to be effective.
Adaptation in a CBR-based Solver Portfolio for the Satisfiability Problem
The satisfiability problem was amongst the very first problems proven to be NP-Complete, and arises in many real world domains such as hardware verification, planning, scheduling, configuration, and telecommunications. Recently, there has been growing interest in using portfolios of solvers for this problem. In this paper we present a cased-based reasoning approach to SAT solving. A key challenge is the adaptation phase, which we focus on in some depth. We present a variety of adaptation approaches, some heuristic, and one that computes an optimal Kemeny ranking over solvers in our portfolio. Our evaluation over a large case-base of problem instances from artificial, hand-crafted, and industrial domains, shows the power of a CBR approach, and the importance of the adaptation scheme used.