Knowledge Science, Engineering and Management

In this paper we present an analysis of DEMETER (Detection of Electromagnetic Emissions Transmitted from Earthquake Regions) satellite data by using the wavelet-based data mining techniques. The analyzed results reveal that the possible anomalous variations exist around the earthquakes. The methods studied in this work include wavelet transformations and spatial/temporal continuity analysis of wavelet maxima. These methods have been used to analyze the singularities of seismic precursors in DEMETER satellite data, which are associated with the two earthquakes of Wenchuan and Pure recently occurred in China.

With the emerging microbloging and social networking platforms, interests become more and more important for user-driven Web applications. Nevertheless, there is no specific logical system that can be directly used to describe human interests and relevant inference rules. In this paper, we introduce the interest logic. We give a formal language to describe the proposed interest logic, then we discuss its semantics and axiomatization. Following the proposed interest logic, we discuss some interesting characteristics of human interests. With the discussion of factors that are related to human interests, we propose some possible extensions of interests logic. Finally, we give several applications of interest logic on the Web platforms to illustrate its potentials and effectiveness.

Customer targeting, which aims to identify and profile the households that are most likely to purchase a particular product or service, is one of the key problems in database marketing. In this paper, we propose an ensemble learning approach to address this problem. Our main idea is to construct different learning hypothesis by random sampling and feature selection. The advantage of the proposed approach for customers targeting is two-folded. First, the uncertainty and instability of single learning method is decreased. Second, the impact of class imbalance on learning bias is reduced. In the empirical study, logistic regression is employed as the basic learning method. The experimental result on a real-world dataset shows that our approach could achieve promising targeting accuracy with time parsimony.

Inconsistency has been considered as one of the main classes of defects in software requirements specification. Various logic-based techniques have been proposed to manage inconsistencies in requirements engineering. However, identifying an appropriate proposal for resolving inconsistencies in software requirements is still a challenging problem. In this paper, we propose a logic-based approach to generating appropriate proposals for handling inconsistency in software requirements. Informally speaking, given an inconsistent requirements specification, we identify which requirements should be given priority to be changed for resolving the inconsistency in that specification, by balancing the blame of each requirement for the inconsistency against its value for that requirements specification. We follow the viewpoint that minimal inconsistent subsets of a set of formulas are the purest forms of inconsistencies in that set. According to this viewpoint, a potential proposal for resolving inconsistencies can be described by a possible combination of some requirements to be changed that can eliminate minimal inconsistent subsets. Then we propose a method of evaluating the degree of disputability of each requirement involved in the inconsistency in a requirements specification. Finally, we provide an algorithm of generating appropriate proposals for resolving the inconsistency in a requirements specification based on the degree of disputability of requirements.

This paper is concerned with a fundamental issue in knowledge representation and reasoning that, surprisingly, has received little attention so far. The point is that inserting some logically weaker (but, in a sense, more precise) information within a logic-based representation is not a straightforward process if the extra information must prevail. Indeed, it does neither prevail by itself nor disable the already existing logically stronger (but less precise) information that subsumes it. A general framework for solving this problem is introduced and instantiated to the task of making some rules prevail over more general ones.

Nowadays, recommender systems are becoming increasingly important because they can filter noisy information and predict users’ preferences. As a result, recommender system has become one of the key technologies for the emerging personalized information services. To these services, when making recommendations, the items’ qualities, items’ correlation, and users’ preferences are all important factors to consider. However, traditional memory-based recommender systems, including the widely used user-oriented and item-oriented collaborative filtering methods, can not take all these information into account. Meanwhile, the model-based methods are often too complex to implement. To that end, in this paper we propose a Gaussian process based recommendation model, which can aggregate all of above factors into a unified system to make more appropriate and accurate recommendations. This model has a solid statistical foundation and is easy to implement. Furthermore, it has few tunable parameters, therefore it is very suitable for a baseline algorithm. The experimental results on the MovieLens data set demonstrate the effectiveness of our method, and it outperforms several state-of-the-art algorithms.

Ontology is a conceptual model, which is used on data exchange between heterogeneous data sources in semantic web, and liked by many more people. Because of the shortage of the uniform standards for constructing ontology, it brings in lots of problems of ontology heterogeneity. Ontology mapping aims at these problems, and semantic similarity between ontologies is the key part of ontology mapping. In this paper we propose a hybrid approach for measuring semantic similarity between ontologies based on WordNet, denoted by WNOntoSim. WordNet is used to calculate semantic similarity between ontologies in elemental level. We compute semantic similarity between ontologies in structural level by constructing contexts of node where the structure of ontology is encoded, and combine these scores to obtain a comprehensive semantic similarity between ontologies. Experimental results on test dataset of competition on ontology matching provided by 3

rd

ISWC show WNOntoSim gives a better performance and improves the Average F-Measure, comparing against some state of the art related methods. Especially, it displays more competitive in general ontology.

In this paper, we propose a goal-oriented recipe recommendation system that utilizes information about nutrition on the Internet. Our system enables users without knowledge about nutrition to search easily for recipes with natural language to improve specific health conditions. The natural language includes ’I want to cure my acne’ and ’I want to recover from my fatigue’. To do that, we created a co-occurrence database that listed the co-occurrence of 45 common nutrients with nouns such as cold, acne, bone etc. Then we created a recipe database by collecting 800,000 recipes from www.cookpad.com the system and analyzed each recipe to calculate the amount of a nutrient in a dish. We compared the results of our system to the results we obtained by calculating the nutrient information manually. Evaluation was done on 1000 dishes. We measured the effectiveness of the system using F-Measure and the average F-measure was 0.64 respectively.

Recent research has shown the usefulness of social tags as a data source to feed resource classification. Little is known about the effect of settings on folksonomies created on social tagging systems. In this work, we consider the settings of social tagging systems to further understand tag distributions in folksonomies. We analyze in depth the tag distributions on three large-scale social tagging datasets, and analyze the effect on a resource classification task. To this end, we study the appropriateness of applying weighting schemes based on the well-known TF-IDF for resource classification. We show the great importance of settings as to altering tag distributions. Among those settings, tag suggestions produce very different folksonomies, which condition the success of the employed weighting schemes. Our findings and analyses are relevant for researchers studying tag-based resource classification, user behavior in social networks, the structure of folksonomies and tag distributions, as well as for developers of social tagging systems in search of an appropriate setting.

Similarity measure is a basic task in time series data mining and attracts much attention in the last decade. This paper considers time series similarity measure from an information theoretic perspective. Based on the function of degree of disagreement (FDOD), a new time series similarity measure method is proposed. The empirical result indicates that the method of this paper can solve the unequal time series and has less time complexity. Meanwhile, it also can measure the similarity between multivariate time series.

To deal with the tri-relation of user-resource-tag in folksonomies and the data sparsity problem in personalized recommendation, we propose a user taste diffusion model based on the tripartite hypergraph to recommend resources for users. Through the defined tri-relation model and diffusion probability matrix, the user’s taste is diffused from itself to other users, resources and tags. When diffusion stops, the candidate resources can be identified then be ranked according to the taste values. As a result the top resources that have not been collected by the given user are selected as the final recommendations. Benefiting from the introduction of iterative diffusion mechanism, the recommendation results not only cover the resources collected by the given user’s direct neighbors but also cover the ones which are collected by his/her extended neighbors. Experimental results show that our method performs better in terms of precision and recall than other recommendation methods.

To realize the vision of intelligent agents on the web, agents need to be capable of understanding people’s behavior. Such an understanding would enable them to better predict and support human activities on the web. If agents had access to knowledge about human goals, they could, for instance, recognize people’s goals from their actions or reason about people’s goals. In this work, we study to what extent it is feasible to automatically construct concept hierarchies of domain-specific human goals. This process consists of the following two steps: (1) extracting human goal instances from a search query log and (2) inferring hierarchical structures by applying clustering techniques. To compare resulting concept hierarchies, we manually construct a golden standard and calculate taxonomic overlaps. In our experiments, we achieve taxonomic overlaps of up to ~51% for the health domain and up to ~60% for individual health subdomains. In an illustration scenario, we provide a prototypical implementation to automatically complement goal concept hierarchies by means-ends relations, i.e. relating goals to actions which potentially contribute to their accomplishment.

Our findings are particularly relevant for knowledge engineers interested in (i) acquiring knowledge about human goals as well as (ii) automating the process of constructing goal concept hierarchies.

Ontology change log data is a valuable source of information which reflects the changes in the domain, the user requirements, flaws in the initial design or the need to incorporate additional information. Ontology change logs can provide operational as well as analytical support in the ontology evolution process. In this paper, we present a novel approach to deal with change representation and knowledge discovery from ontology change logs. We look into different knowledge gathering aspects to capture every single facet of ontology change. The ontology changes are formalised using a graph-based approach. The knowledge-based change log facilitates detection of similarities within different time series, discovering implicit dependencies between ontological entities and reuse of knowledge. We analyse an ontology change log graph in order to identify frequent changes that occur in ontologies over time. We identify different types of change sequences based on their order and completeness. Analysis of change logs also assists in extracting new change patterns and rules which cannot be found by simply querying or processing ontology change logs.

It is well known that environment plays an important part in requirements engineering(RE). However, at present the time-continuous properties of the environment are not considered yet in RE. This paper proposes to model the time-continuous environment by constructing a software environment ontology. Based on this ontology, a small example is given for deriving software requirements specification using problem frames approach under different situations. The results show that the software behaviors can be more accurately determined with respect to time-continuous environment by using time as a measurement.

Intrusion detection is one of the most essential factors for security infrastructures in network environments, and it is widely used in detecting, identifying and tracking the intruders. Traditionally, the approach taken to find attacks is to inspect the contents of every packet. An alternative approach is to detect network applications based on flow statistics characteristics using machine learning. We propose online Internet intrusion detection based on flow statistical characteristics in this paper. Experiment results illustrate this method has high detection accuracy using Seeded-Kmeans clustering algorithm. It is noticeable that the statistics of the first 12 packets could detect online flow with high accuracy.

After introducing some approaches to knowledge science as well as the School of Knowledge Science at Japan Advanced Institute of Science and Technology, this paper introduces a new book entitled

Knowledge Science–Modeling the Knowledge Creation Process

(Nakamori ed., 2011) to suggest what knowledge science should be. The authors of this book are experienced researchers in knowledge science with the background of systems science, and core members of the

International Society for Knowledge and Systems Sciences

. This paper also introduces a theory under development for knowledge synthesis or construction, which is a methodology to collect and synthesize a variety of knowledge to solve contemporary, complex real-life problems.

With the rapid development of online Knowledge Sharing Communities (KSCs), the problem of finding experts becomes increasingly important for knowledge propagation and putting crowd wisdom to work. A recent development trend of KSCs is to allow users to add text tags for annotating their posts, which are more accurate than traditional category information. However, how to leverage these user-generated tags for finding experts is still under-explored. To this end, in this paper, we develop a novel approach for finding experts in tag based KSCs by leveraging tag context and the semantic relationship between tags. Specifically, the extracted prior knowledge and user profiles are first used for enriching the query tags to infer tag context, which represents the user’s latent information needs. Then, a topic model based approach is applied for capturing the semantic relationship between tags and then taking advantage of them for ranking user authority. We evaluate the proposed framework for expert finding on a large-scale real-world data set collected from a tag based Chinese commercial Q&A web site. Experimental results clearly show that the proposed method outperforms several baseline methods with a significant margin.

Similarity based clustering, which is to find the extrinsic clusters in data by taking as input a collection of real-valued similarities between data points, has been playing an important role in data analysis and engineering. Lots of work had been done in this field. However, data clustering is an rather challenge problem as there is no labeled data available. We observe that an ideal similarity matrix should be close to an adjacency matrix up to a scale. Based on this idea, we develop a scaled adjacency matrix (SAM) clustering algorithm that could find an optimal adjacency matrix in some sense for a given similarity matrix. Based on the learnt adjacency matrix, clustering could be performed straightforwardly. Upon three assumptions on the similarity matrix, we prove that the performance of SAM is robust. Experimental results also show that SAM is effective.

Requirements elicitation is one of the hardest and most critical parts of software development. Ontology technology provides a good way to support this work. In the literature, there many methods proposed about how to make use of ontology in requirement elicitation. However most of such methods assume that there already exists some domain ontology for reuse. Different from this works, our approach breaks this assumption and claims that we should merge the ontology acquiring process and the requirement elicitation process together. In order to do that, this paper proposes an upper level ontology for process centered problem domain to help the analysts to represent the requirement, and defines an engineerable process to guide the ontology acquiring integrated requirement elicitation process. In the end, we make a case study in the taxation domain to illustrate the effectiveness of our approach.

Effective and efficient knowledge dissemination and reasoning in distributed, real-time, and embedded (DRE) systems remains a hard problem due to the need for tight time constraints on evaluation of rules and scalability in dissemination of knowledge events. Limitations in satisfying the tight timing properties stem from the fact that most knowledge reasoning engines continue to be developed in managed languages like Java and Lisp, which incur performance overhead in their interpreters due to wasted precious clock cycles on managed features like garbage collection and indirection. Limitations in scalable dissemination stem from the presence of ontologies and blocking network communications involving connected reasoning agents. This paper addresses the existing problems with timeliness and scalability in knowledge reasoning and dissemination by presenting a C++-based knowledge reasoning solution that operates over a distributed and anonymous publish/subscribe transport mechanism provided by the OMG’s Data Distribution Service (DDS). Experimental results evaluating the performance of the C++-based reasoning solution illustrate microsecond-level evaluation latencies, while the use of the DDS publish/subscribe transport illustrates significant scalability in dissemination of knowledge events while also tolerating joining and leaving of system entities.

Belief function theory is an appropriate framework to model different forms of knowledge including probabilistic knowledge. One simple and efficient way to reason under uncertainty, is the use of compact graphical models, namely directed acyclic graphs. Therefore naturally, a question crosses the mind: If we deal with Bayesian belief knowledge does the network collapse into a Bayesian network? This paper attempts to answer this question by analyzing different forms of belief function networks defined with conditional beliefs defined either with a unique conditional distribution for all parents or a single conditional distribution for each single parent. We also propose a new method for the deconditionalization process to compute the joint distribution.

In the knowledge-based economy intangible production (de-materialised production) plays the main role. Intangible production is an advanced manufacturing process performed on the information level, where input materials, semi products and final products are in a digital form. The production network consists of nodes, each of which performs processing of information and knowledge through collaboration with other nodes. The paper focuses on the kind of intangible production where the production process utilizes different types of knowledge and competence is the final product. The educational organizations and distance learning are a good example of this type of the intangible production. In the paper a model of intangible production network for competence development in the context of educational organization is discussed. The proposed approach can be used to develop and manage knowledge-base systems on the level of ontology.

Since the last decade the wide spread language for expressing ontologies relies on Description Logics (DLs). However, most of the versions syntactically anchor their modeling primitives on classical logic and require additional theories (i.e., first-order logic, ...) for simultaneously supporting (i) the introduction of constant values (e.g., for individuals) (ii) the limitation of expressiveness for decidability and (iii) the introduction of variables for reasoning with rules. In this paper we show that the introduction of a type theoretical formalism that relies both on a constructive logic and on a typed lambda calculus is able to go beyond these aspects in a single theory. In particular we will show that a number of logical choices (constructive logic, predicative universes for data types, impredicative universe for logic, ...) about the theory will lead to an highly expressive theory which allows for the production of conceptually clean and semantically unambiguous ontologies.

A critical challenge in quantitative financial risk analysis is the effective computation of volatility and correlation. However, the dynamic nature of financial data environments create the challenges for robust correlation computing, particularly when the number of financial instruments and the volume of transactions grow dramatically. To this end, in this paper, we present an organized study of rank correlation computing for financial risk analysis in dynamic environments. Specifically, we focus on Kendall’s

τ

, which is widely recognized as a robust correlation measure for evaluating financial risk. Kendall’s

τ

is not widely used in practice partially because its computation complexity is

O

(

n

2

), making it difficult to frequently recompute in dynamic environments. After carefully studying the computational properties of Kendall’s

τ

, we reveal that Kendall’s

τ

is very computation-friendly for incremental computing of correlations, since the relativity of existing observations will not change as new observations come in. Based on this finding, we develop a

τ

Grow algorithm for dynamically computing Kendall’s

τ

. Also, even for one-time static Kendall’s

τ

computation, we observe that the Kendall’s

τ

correlations on smaller time pieces can provide concise summaries of how Kendall’s

τ

evolves over the whole period. Finally, the effectiveness and the efficiency of the proposed methods have been demonstrated through the experiments on real-world financial data.

The article proposes the decision support method for selection of project team members basing on their knowledge, experience and collaboration skills. According to project management best practices presented in the article, successful selection of project team members requires formal model of knowledge as well as accurate measures of knowledge worker performance in completing assigned project tasks. The method is driven by the model of competence that meets these requirements by providing quantitative analysis of knowledge and skills as well as employing methods of mathematical programming and elements of fuzzy sets theory. The author of the paper proposes the multiple criteria decision approach for project team building based on the quantitative method for competence analysis of project team members.

Police agencies have been collecting an increasing amount of information to better understand patterns in criminal activity. Recently there is a new trend in using the data collected to predict where and when crime will occur. Crime prediction is greatly beneficial because if it is done accurately, police practitioner would be able to allocate resources to the geographic areas most at risk for criminal activity and ultimately make communities safer. In this paper, we discuss a new four-order tensor representation for crime data. The tensor encodes the longitude, latitude, time, and other relevant incidents. Using the tensor data structure, we propose the Empirical Discriminative Tensor Analysis (EDTA) algorithm to obtain sufficient discriminative information while minimizing empirical risk simultaneously. We examine the algorithm on the crime data collected in one Northeastern city. EDTA demonstrates promising results compared to other existing methods in real world scenarios.

Agent-oriented analysis and design is a prosperous approach to model and build software systems. On the other hand, semantic web services are software components that have been emerging to enable dynamic service discovery, composition, invocation, and provide services, which can be considered as the main function of an agent. Semantic web service composition is a feature that improves the flexibility of the system. In this paper we propose a multi-agent system for web service composition. We modeled it by MaSE (Multi-agent System Engineering) methodology, which is a top-down approach. Also an implementation of our system is illustrated for semantic search engines. The case study shows that our system is feasible and effective for composition of semantic web services in a distributed network.

As the Web is becoming the largest knowledge repository which contains various entities and their relations, the task of related entity retrieval excites interest in the field of information retrieval. This challenging task is introduced in TREC 2009 Entity Track. In this task, given an entity and the type of the target entity, as well as the nature of their relation described in free text, a retrieval system is required to return a ranked list of related entities that are of the target type. It means that entity ranking goes beyond entity relevance and integrates the judgment of relation into the evaluation of the retrieved entities. In this paper, we propose a probability model using relation pattern to address the task of related entity retrieval. This model takes into account both relevance and relation between entities. We focus on using relation patterns to measure the level of relation matching between entities, and then to estimate the probability of occurrence of relation between two entities. In addition, we represent entity by its context language model and measure the relevance between two entities by a language model approach. Experimental results on TREC Entity Track dataset show that our proposed model significantly improves retrieval performances over baseline. The comparison with other approaches also reveals the effectiveness of our model.

This paper introduces our research work in fuzzy target-oriented decision analysis and its application to kansei-based evaluation of traditional crafts. After a brief introduction into fuzzy target-oriented decision analysis, we formulate a general target-oriented approach to multi-attribute evaluation problem for personalized recommendation. The central idea of this approach is to first interpret a particular user’s request as a target (or benchmark) at which the user would be only interested in candidates meeting this target, and then use a combination of target-oriented decision analysis and aggregation operators for defining an evaluation function that quantifies how well a candidate meets the user’s target. As for illustration, we will introduce a target-based evaluation method for multi-feature ranking of traditional craft products using

kansei

data and preferences specified by consumers, where product items are assessed according to the so-called

kansei

features, and kansei data are treated as categorical data.

For imbalanced data sets, examples of minority class are sparsely distributed in sample space compared with the overwhelming amount of majority class. This presents a great challenge for learning from the minority class. Enlightened by SMOTE, a new over-sampling method, Random-SMOTE, which generates examples randomly in the sample space of minority class is proposed. According to the experiments on real data sets, Random-SMOTE is more effective compared with other random sampling approaches.

In this paper we present a formalisation for a previously elaborated model-based approach for representing knowledge intensive processes in the domain of nuclear knowledge management. It is shown how a Nuclear Fuel Cycle Process Model can be formalised and represented visually by using the concepts of pattern theory. The formalisation is then applied to the configuration of acquisition paths and the analysis of the evolution of the models. Thereby a basis for the application of formal queries and advanced visual analyses can be established.

Weighting connections between different layers within a lattice structure is an important issue in the process of modeling activity recognition within smart environments. Weights not only play an important role in propagating the relational strengths between layers in the structure, they can be capable of aggregating uncertainty derived from sensors along with the sensor context into the overall process of activity recognition. In this paper we present two weight factor algorithms and experimental evaluation. According to the experimental results, the proposed weight factor methods have a better performance of reasoning the complex and simple activity than other methods.

The counterpart theory which was introduced by David Lewis is an alternative semantics to the possible worlds semantics for quantified modal logic. Lewis interprets modal claims by using a translation from quantified modal logic into the counterpart theory. Due to the flexibility of semantics of the counterpart theory, Lewis’s translation may translate an unsatisfiable formula to a satisfiable one. In this paper, two properties are defined to describe Lewis’s translation, called the faithfulness and fullness. The former implies a translation which preserves the satisfiability of formulas, whereas the latter implies the preservation of the unsatisfiability. We show that Lewis’s translation is faithful but not full. To make Lewis’s translation full, two auxiliary axioms are added to restrain the counterpart relation such that every possible object has exactly one counterpart in every possible world. Under the circumstances, we show that Lewis’s translation is a faithful and full translation.

A reference frame was selected from a video sequence, and corners were found by Harris operator. Corners were tracked by Lucas-Kanade optical flow which was based on Gaussian Pyramid, like this those corners were traced in the current frame. Then corresponding corner set between the reference frame and the current frame were obtained. An affine transformation matrix can be solved by those two corner sets above. Stabilized video sequence was got by affine transformation of current frame.

There are series of modes between actors as they transfer knowledge with each other. Different modes will totally lead to quite different efficiency and results, which will further influence on organizational performances and innovations. This paper generalizes nine kinds of modes of knowledge transfer between actors, and then classifies the organization based on the modes, at last simulates the different modes of knowledge transfer on small world networks according to the setting rules. Through the simulation experiment we compute the average knowledge store and knowledge variance, the result shows that different modes of knowledge transfer will affect efficiency of knowledge transfer. When there are entirely two-way solid line mode in the organizations, knowledge transfer will be the fastest. Organizations of greater density are efficient, in which different modes have less influence to knowledge transfer. And while there are more one-way lines, the variance will be bigger.

Guimerà and his colleagues proposed an interesting modelto study the evolution of collaboration networks, in which the creative teams are the basic building blocks of the collaboration network and the network grows by repetitively assimilating new teams. We argue that one limitation of this GUSA model is that the intrinsic mutual influence of the collaboration network and the collective production and diffusion of knowledge in the network is largely neglected. Based on this argumentation, we in this paper propose an abstract meta-model that extends and generalizes the GUSA model in order to study the evolutionary dynamics of collaboration networks with the team assembly mechanism. By integrating the mechanism of team-wide knowledge production and diffusion, the proposed meta-model provides a unified framework to simultaneously study knowledge dynamics and structural evolution of the network. In tune with the proposed meta-model, an agent-based modeling framework is briefly discussed.

Missing data imputation is a key issue of data pre-processing in data mining field. Though there are many methods for missing value imputation, almost each of these imputation methods has its limitation and is designed for either numeric attributes or categorical attributes. This paper presents IMIC, a new missing value Imputation method for Mixed numeric and categorical attributes based on Incomplete data hierarchical clustering after the introduction of a new concept Incomplete Set Mixed Feature Vector (ISMFV). The effect of the new method is valuated through the comparison experiment using 3 real data sets from UCI.

Traditional classification methods in machine learning assume that training data and testing data should share the same feature space and have the same data distribution. In real world applications, however, this assumption often does not hold. If there are very few labeled instances in the target domain for training, it is time-consuming to label them manually. In this case, a source domain which has semantic relationships with the target domain but has the different feature space or distribution can be used to assist the classification. In this paper, we propose a new method using rules to help the domain adaptation, which can well represent the knowledge relationships between source domain and target domain. In this algorithm we first discover term-term rules according to the term relationships in target domain to build the knowledge bridge, then we reconstruct the source domain using these rules and get a better classifier to improve the cross-domain classification performance. We conduct several cross-domain data sets and demonstrate that the proposed method is easy to understand and it has a better performance compared to state-of-art transfer algorithms.

Recently, the literature proposes many approaches to the interoperability of Web data based on the used of formal models. We consider Adaptive Web Applications, in which a relevant requirement is the ability to capture and manipulate context information. Since context data are often heterogeneous, their translations from one representation into another is an important issue that needs to be addressed to facilitate their integration. In this paper we describe a general framework supporting the representation of a large variety of context information and the translation between different formats. Translations are specified as compositions of elementary steps, defined by means of special operations. With these operations, we show how it is possible to process automatically a translation on the models that are provided as source and target.

In an emergency response system (ERS), resource matchmaking schema (RMS) consists of emergency impact assessment, resource utility classification and mapping between impact and utility. These tasks serve the resource allocation and must be done before launching rescue activities. However, related concepts and relations have little explication. Consequently, an adequate knowledge structure is necessary. In this paper, a domain ontology for RMS (RMS_Ontology) is formally defined, on which impact-utility mapping is implemented. Based on it, we illustrate how RMS works in the ERS application of highway network under severe weather conditions. Finally, a prototype system has been developed to facilitate knowledge management and to improve the performance of emergency disposal procedures.

Error-correcting output codes (ECOC) is an effective method to perform multi-classification via decomposing a multi-classification problem into many binary classification tasks, and then integrating the outputs of the subtasks into a whole decision. The researches on applying ECOC to multi-classification mainly focus on how to improve the correcting ability of output codes and how to enhance the classification effectiveness of ECOC. This paper addresses a simple but interesting and significant case of ECOC, the shortest ECOC, to perform fast multi-classification at the cost of sacrificing a very small classification precision. The strategy of balancing the positive and negative examples for each binary classifier of ECOC and the method of finding the optimal permutation of all original classes are further given. Preliminary experimental results show, the shortest ECOC uses fewest binary classifiers but can still obtain comparable or close classification precisions with several traditional encoding methods of ECOC.