Recently, e-learning systems for self-learning with various types of retrieval functions have been developed. This paper describes a method for keyword extraction using retrieval information stored from many students through the retrieval functions. Firstly, we show that teachers tend to consider technical terms as important, while students unfamiliar with the technical terms tend to retrieve the terms, therefore there is a clear correlation between keywords extracted by the teachers and the retrieval words by the students. Secondly, we propse a method (...) utilizing retrieval information from the students for keyword extraction, and show that the mehod can achieve quite better performance than a method extracting keywords using only lecture information. (shrink)

In this paper, we focus on a problem in designing for home robots. There are some design methods for conventional artifacts such as home electric appliances. However, not all methods can be applied effectively for robots. Especially, we deal with the situation in which a robot asks a user to help it by expressing its internal state. We then propose a novel design method ``motion overlapping '' by which a robot can perform human-like behavior to express its internal state. We (...) consider that human-like behavior of a robot causes a user to understand its internal state ``mind'' easily. A small sweeping robot which performs ``back-and-forth'' motion is designed based on MO. In experiments, we compare the expressing by MO with sounding by a buzzer and lighting by a LED as conventional nonverbal expressing. We investigate effects on users' action of cooperating with the sweeping robot. We find that the expressing by MO causes most of the users to help the robot. The differences in those expressing methods are statistically significant. The results show that our proposed method is effective as one of design method for home robots. (shrink)

We investigate induction from the viewpoint of nonmonotonic reasoning. Induction we consider in this paper is descriptive induction. Hypotheses from descriptive induction have the weak property that they only describe rules with respect to the observations and do not realize an inductive leap. In this paper, we define a new form of descriptive induction with circumscription and the idea of explanation and show two procedures for computing it. The new descriptive induction is called circumscriptive induction. By deciding the roles of (...) predicates in circumscription, we can intentionally minimize models of a given inductive problem. By adopting the idea of explanation, we can distinguish between background knowledge and observations. Additionally, we consider the relationship between the way of choosing the roles of predicates in computing circumscription and the property of hypotheses obtained by circumscriptive induction. It is shown that hypotheses from circumscriptive induction reflect a difference between background knowledge and observations and do not realize an inductive leap. We also investigate revision of hypotheses which is as important as generation of hypotheses. In a process of hypothesis revision, a difference between previous induction and circumscriptive induction is clearly characterised. (shrink)

Physical skills and language skills are both fundamental intelligent abilities of human being. In this paper, we focus our attention to such sophisticated physical skills as playing sports and playing instruments and introduce research activities aiming at elucidating and verbalizing them. This research area has been launched recently. We introduce approaches from physical modeling, measurements and data analysis, cognitive science and human interface. We also discuss such issues as skill acquisition and its support systems. Furthermore, we consider a fundamental issue (...) of individual differences occurring in every application of skill elucidation. Finally we introduce several attempts of skill elucidation in the fields of dancing, manufacturing, playing string instruments, sports science and medical care. (shrink)

This paper addresses the challenging problem of automating the human's intelligent ability to evaluate output from machine translation systems, which are subsystems of Speech-to-Speech MT systems. Conventional automatic MT evaluation methods include BLEU, which MT researchers have frequently used. BLEU is unsuitable for SSMT evaluation for two reasons. First, BLEU assesses errors lightly at the beginning or ending of translations and heavily in the middle, although the assessments should be independent from the positions. Second, BLEU lacks tolerance in accepting colloquial (...) sentences with small errors, although such errors do not prevent us from continuing conversation. In this paper, the authors report a new evaluation method called RED that automatically grades each MT output by using a decision tree. The DT is learned from training examples that are encoded by using multiple edit distances and their grades. The multiple edit distances are normal edit dista nce defined by insertion, deletion, and replacement, as well as extensions of ED. The use of multiple edit distances allows more tolerance than either ED or BLEU. Each evaluated MT output is assigned a grade by using the DT. RED and BLEU were compared for the task of evaluating SSMT systems, which have various performances, on a spoken language corpus, ATR's Basic Travel Expression Corpus. Experimental results showed that RED significantly outperformed BLEU. (shrink)

Reinforcement Learning is a promising technique for creating agents that can be applied to real world problems. The most important features of RL are trial-and-error search and delayed reward. Thus, agents randomly act in the early learning stage. However, such random actions are impractical for real world problems. This paper presents a novel model of RL agents. A feature of our learning agent model is to integrate the Analytic Hierarchy Process into the standard RL agent model, which consists of three (...) modules: state recognition, learning, and action selecting modules. In our model, the AHP module is designed with {\\it primary knowledge} that humans intrinsically have in a process until a goal state is attained. This integration aims at increasing promising actions instead of completely random actions in the standard RL algorithms. Profit Sharing is adopted as a RL method for our model, since PS is known to be useful even in multi-agent environments. To evaluate our approach in a multi-agent environment, we test a PS RL method with our agent model on a pursuit problem in a grid world. Computational results show that our approach outperforms the standard PS in terms of learning speed in the earlier stages of learning. We also show that the learning performance of our approach is superior at least competitive to that of the standard one in the final stages of learning. (shrink)

This paper presents a mathematical model for decision making processes where the knowledge for the decision is constructed automatically from subjective information on the Internet. This mathematical model enables us to know the required degree of accuracy of knowledge acquisition for constructing decision support systems using two technologies: automated knowledge acquisition from information on the Internet and automated reasoning about the acquired knowledge. The model consists of three elements: knowledge source, which is a set of subjective information on the Internet, (...) knowledge acquisition, which acquires knowledge base within a computer from the knowledge source, and decision rule, which chooses a set of alternatives by using the knowledge base. One of the important features of this model is that the model contains not only decision making processes but also knowledge acquisition processes. This feature enables to analyze the decision processes with the sufficiency of knowledge sources and the accuracy of knowledge acquisition methods. Based on the model, decision processes by which the knowledge source and the knowledge base lead to the same choices are given and the required degree of accuracy of knowledge acquisition is quantified as required accuracy value. In order to show the way to utilize the value for designing the decision support systems, the value is calculated by using some examples of knowledge sources and decision rules. This paper also describes the computational complexity of the required accuracy value calculation and shows a computation principle for reducing the complexity to the polynomial order of the size of knowledge sources. (shrink)

In the process of searching for suspected person, there is a case where an eyewitness looks through some photographs of criminal's face, or mug shots, to identify witnessed suspect's face or a searcher examines some mug shots in order to find some photographs that seem similar to a montage. However the eyewitness or the searcher has a heavy load to look through many photographs. It is widely known that this load causes a marked decline in human's ability to identify photographs. (...) Therefore there is a need for a facial image retrieval system that can help humans to identify the facial images. An interactive facial image retrieval system combined the retrieval using linguistic keywords and the simulated breeding was proposed in order to reduce the user's load. The system performs the retrieval by using the impressive adjective first, and then retrieves the facial images by using simulated breeding that can treat the user's subjectivity. However, this method still requires the user to do heavy work. In order to improve the system, we add a retrieval method using Euclidean distance to the system for retrieving the facial images that are displayed for the user's first selection. In our experiment, firstly displayed facial images in the proposed system are similar to the target, which leads to the reduction of user's load. (shrink)

The authors aim at constructing an agent which learns appropriate actions in a Multi-Agent environment with and without social dilemmas. For this aim, the agent must have nonrationality that makes it give up its own profit when it should do that. Since there are many studies on rational learning that brings more and more profit, it is desirable to utilize them for constructing the agent. Therefore, we use a reward-handling manner that makes internal evaluation from the agent's rewards, and then (...) the agent learns actions by a rational learning method with the internal evaluation. If the agent has only a fixed manner, however, it does not act well in the environment with and without dilemmas. Thus, the authors equip the agent with several reward-handling manners and criteria for selecting an effective one for the environmental situation. In the case of humans, what generates the internal evaluation is usually called emotion. Hence, this study also aims at throwing light on emotional activities of humans from a constructive view. In this paper, we divide a Multi-Agent environment into three situations and construct an agent having the reward-handling manners and the criteria. We observe that the agent acts well in all the three Multi-Agent situations composed of homogeneous agents. (shrink)

In this paper, we discuss a method to dynamically determine the generality of the target concept in a class hierarchy, when learning default rules, i.e., rules including exceptions with Inductive Logic Programming. The ILP system for default rules has to learn both the target concept and its opposite, if it is based on a three valued setting, in which we clearly discriminate among the three values: what is true, what is false, and what is unknown. Thus in order to learn (...) rules which holds as generally as possible in a class hierarchy implicitly existing in given examples, we should give a higher priority to the concept which is more general, or covers more examples than does the other in the hierarchy. For this purpose, our method first finds out the general rule from a set of candidate rules independently of the concept it defines. Then the body of the rule can be viewed as the description defining the most general class in the hierarchy. Therefore, according to the ratio of positive examples it covers, we can determine which of the concepts, the target one or its opposite, is more general, and dynamically change the head of the rule to the negative literal if the latter concept is more general. In this paper, we formalize this method as a new ILP system, GREX, and discuss it with some examples. (shrink)

There have been several previous studies on measuring the semantic similarity between words whose concepts are represented as points in a multi-dimensional vector space acquired from text data such as electronic dictionaries or text corpora. A central problem in these studies is how to select orthonormal basis vectors for the space which represents attributes of the words. We propose a method of building the space by combining two representative methods, one using singular value decomposition and the other using the contents (...) of a thesaurus. The proposed method was evaluated both for the purposes of similar word retrieval and for document retrieval. The evaluations showed that the proposed combination is more effective than either of the original methods alone for both of these tasks. (shrink)

Knowledge discovery in databases has been studied intensively recent years. In KDD, inductive classifier learning methods which were developed in statistics and machine learning have been used to extract classification rules from databases. Although in KDD we have to deal with large databases in many cases, many of the previous classifier learning methods are not suitable for large databases. They were designed under assumption that any data in databases is accessible on demand and they usually need to access a datum (...) several times in a process of learning. So, they require a huge memory space or a large I/O cost to access storage devices. In this paper, we propose a classifier learning method, we call CIDRE, in which data summaries are constructed and classifiers are learned from the summaries. This learning method is realized by using a clustering method, we call MCF-tree, which is an extension of CF-tree proposed by Zhang et al. In the method, we can specify the size of memory space occupied by data summaries, and databases are swept only once to construct the summaries. In addition, new instances can be inserted into the summaries incrementally. Thus, the method possesses important properties which are desirable to deal with large databases. We also show empirical results, which indicate that our method performs very well in comparison to C4.5 and naive Bayes, and the extension from CF-tree to MCF-tree is indispensable to achieve high classification accuracy. (shrink)

The purpose of this study is development of a supporting system for teacher's design of lesson plan. Especially design of lesson plan which relates to the new subject "Information Study" is supported. In this study, we developed a system which generates teaching and learning activity sequences by interlinking lesson's activities corresponding to the various conditions according to the user's input. Because user's input is multiple information, there will be caused contradiction which the system should solve. This multiobjective optimization problem is (...) resolved by Distributed Genetic Algorithms, in which some fitness functions are defined with reference models on lesson, thinking and teaching style. From results of various experiments, effectivity and validity of the proposed methods and reference models were verified; on the other hand, some future works on reference models and evaluation functions were also pointed out. (shrink)

The goal of our research is to develop a software environment in which users can select appropriate Knowledge Discovery systems from various candidates, and compare each obtained result with others. This paper reports a prototype of such an environment, named EVLD. The key idea in constructing EVLD is introducing three-level architecture into it. We adopt the logic of discovery proposed by Plotkin for the middle level, and put every Knowledge Discovery system on the lowest level. By designing the transformation of (...) the two levels, users of EVLD obtain outputs of every discovery system in the form of logic programs. They can compare the outputs by comparing the logic programs. Moreover the logic programs are stored in a rule base, and can be used as a background theory for another trial of discovery systems. Though the logic of discovery was invented thirty years ago with philosophical consideration of discovery, it includes Knowledge Discovery from relational databases in first-order logic. Therefore constructing EVLD is connecting recent Knowledge Discovery systems with Plotkin's philosophical consideration. (shrink)

Many stochastic search algorithms have recently been developed to make more remarkable progress than systematic search algorithms because stochastic algorithms sometimes solve large-scale constraint satisfaction problems in a practical time. However, such stochastic algorithms have the drawback of getting stuck in local optima which are not acceptable as final solutions. We analyze an iterative improvement algorithm from the viewpoint of constraint structures causing local optima. Using the graph-coloring problem with three colors, an archetype problem to evaluate constraint satisfaction algorithms, we (...) study the local graph structures around which so many conflicts thrash. We propose a key constraint structure, an LM pair, which may induce a local optimum, and clarify the mechanism of how conflicting colors for an LM pair obstruct the stepwise refinement of hill-climbing. Experimental results show that LM pairs are strongly correlated with the search efficiency of the stochastic search algorithm. (shrink)

This paper presents conceptual clustering algorithms using regression analysis. The basic idea is that given data can be classified to the class “existing” and so conceptual clustering is transformed to classification. The algorithms consist of transforming given data to the data with a class, obtaining a function by regression analysis, approximating the function by a Boolean function, and generating a concept hierarchy from the Boolean function. Regression analysis includes linear regression analysis and nonlinear regression analysis by neural networks. The algorithms (...) can perform the multiple classification and generate simple clusters. The algorithms using linear regression analysis and neural networks have been applied to real data. Results show that the algorithm using neural networks works well. (shrink)

Cost-based abduction, which is a technique for identifying the best explanation for a given observation based on the assumption of a set of hypothesis, is a useful knowledge processing framework for practical problems such as diagnosis, design and planning. However, the speed of reasoning of this approach is often slow. To overcome this problem, Kato et al. previously presented a more efficient cost-based abduction system, that utilized the A * search technique, however, the time and space complexities in this technique (...) are exponential, so the identification of the optimal solution is difficult in practical applications. In this paper, we present three new systems in which the user can define the computational complexity, for identification of a near-optimal solution. First, we introduce two search control techniques; a real-time A * search approach in which the user can define the look-ahead depth or space, and the multi-agent real-time A * approach in which the user can define the number of real-time A * agents used in the search. We describe the implementation of three cost-based abduction reasoning systems for predicate logic knowledge bases and test the proposed systems using a diagnostic logic circuit problem. The results show that proposed systems can identify a near-optimal solution according to the predefined polynomial order of complexity, including the selection of either linear or exponential computational complexity. It is also shown that inference time and success rate are dependent on the user-defined parameters, that the three proposed systems exhibit similar performance characteristics, and that they all offer significant speed advantages over the previously described technique. (shrink)

Inductive Logic Programming employs expressive representation languages, i.e. Prolog Programs, so that ILP can handle structural data which other traditional inductive learner can not or hardly do. In this reason, ILP has been regarded as one of the most important technologies in the area of Data Mining and Knowledge Discovery in Databases recently. However, ILP usually needs enormous computational time to obtain the results from a huge amount of data appearing in such problems as Data Mining. To cope with this (...) problem and to make ILP more practical, we need more efficient algorithms. Since learning by ILP can be regarded as the search problem, we need to consider the reduction of the number of hypotheses to be evaluated and the efficient evaluation algorithm in order to make ILP more efficient. One of the most effective ways in reducing the hypothesis space to be searched is to compute a Most Specific Hypothesis from one positive example by Inverse Entailment. Since MSH bounds the hypothesis space, it is important to decide which example should be used to generate MSH. In this paper, in order to reduce the computational time, we propose the following three algorithms for ILP systems which employ coverset algorithm, inverse entailment and top-down search strategy: selection of search space, incremental search within one class, integration of hypothesis spaces among the different classes. The main common feature of these three algorithms is that, instead of single MSH, plural MSHs are considered in the search process. Experiments were conducted to assess the effectiveness of the proposed algorithms. The results show that the proposed algorithms are useful for reducing the number of candidate hypotheses to be evaluated as well as the total computational time for induction. (shrink)

Miten ajattelu oikeuksista pohjautuu antiikin ja varhaisen keskiajan filosofiaan ja yhteis­kunnalliseen kehitykseen? Mitä tarkoittaa, että jollakulla on oikeus johonkin, ja kenellä oikeuksia voi olla? Entä mihin universaalit ihmisoikeudet perustuvat? Julkisuudessa ja politiikassa puhutaan yhä useammin oikeuksista. Turvapaikanhakijoiden kohtelusta keskusteltaessa esiin nostetaan etenkin ihmisoikeudet. Huoli ilmaston­muutoksesta on kiihdyttänyt pohdintaa eläinten ja luonnonympäristön oikeuksista. Kiivasta väittelyä synnyttää myös sanan­vapauden ja vihapuheen rajoittamisen suhde. Mitä oikeudet ovat? -teoksessa oikeuden käsitettä ja kehitystä pohtivat niin oikeus­tieteilijät kuin filosofitkin. Se on ensimmäinen suomenkielinen kirja, jossa näiden (...) kahden tieteenalan edustajat keskustelevat oikeuksista saman pöydän ääressä. Miten ajattelu oikeuksista pohjautuu antiikin ja varhaisen keskiajan filosofiaan ja yhteis­kunnalliseen kehitykseen? Mitä tarkoittaa, että jollakulla on oikeus johonkin, ja kenellä oikeuksia voi olla? Entä mihin universaalit ihmisoikeudet perustuvat? Teoksessa pureudutaan muun muassa näihin kysymyksiin niin juridisesta kuin filosofisestakin näkö­kulmasta. Monet luvuista muodostavat parin, jotka argumentoivat toisiaan vastaan. Avoimella ja väittelyyn kannustavalla otteellaan teos kutsuu lukijankin mukaan keskusteluun oikeuksista. (shrink)

The outcome of this is explored, in detail, in relation to the teaching of literature, creative writing and drama.