Bangyong Liang

Recommendation over a Heterogeneous Social Network

With the Web content having been changed from homogeneity to heterogeneity, the recommendation becomes a more challenging issue. In this paper, we have investigated the recommendation problem on a general heterogeneous Web social network. We categorize the recommendation needs on it into two main scenarios: recommendation when a person is doing a search and recommendation when the person is browsing the information. We formalize the recommendation as a ranking problem over the heterogeneous network. Moreover, we propose using a random walk model to simultaneously ranking different types of objects and propose a pair-wise learning algorithm to learn the weight of each type of relationship in the model. Experimental results on two real-world data sets show that improvements can be obtained by comparing with the baseline methods.

iASA: learning to annotate the semantic web

With the advent of the Semantic Web, there is a great need to upgrade existing web content to semantic web content. This can be accomplished through semantic annotations. Unfortunately, manual annotation is tedious, time consuming and error-prone. In this paper, we propose a tool, called iASA, that learns to automatically annotate web documents according to an ontology. iASA is based on the combination of information extraction (specifically, the Similarity-based Rule Learner—SRL) and machine learning techniques. Using linguistic knowledge and optimal dynamic window size, SRL produces annotation rules of better quality than comparable semantic annotation systems. Similarity-based learning efficiently reduces the search space by avoiding pseudo rule generalization. In the annotation phase, iASA exploits ontology knowledge to refine the annotation it proposes. Moreover, our annotation algorithm exploits machine learning methods to correctly select instances and to predict missing instances. Finally, iASA provides an explanation component that explains the nature of the learner and annotator to the user. Explanations can greatly help users understand the rule induction and annotation process, so that they can focus on correcting rules and annotations quickly. Experimental results show that iASA can reach high accuracy quickly.

Risk Minimization Based Ontology Mapping

The key point to reach interoperability over distributed ontologies is the mediation between them, called ontology mapping. Absolutely manually specified mapping is tedious and time consumption. Additional, how to ensure the consistency and deal with error prone in manual process, further how to maintain the mapping with the evolution of ontologies are all beyond manual work. Therefore, it is indeed necessary to automatically discover the mapping between ontologies so that mergence and translation of different ontology-based annotations become possible. Existing (semi-)automatic processing system are restricted to limited information, which depress the performance especially when the taxonomy structures have little overlapping or the instances have few commons. In this paper, based on Bayesian decision theory, we propose an approach called RiMOM to automatically discover mapping between ontologies. RiMOM treats the entire mapping problem as a decision problem instead of similarity problem in previous work. It explicitly and formally gives a complete decision model for ontology mapping. Based on shallow NLP, this paper also introduces a method to deal with instances heterogeneity, which is a long-standing problem for information processing. Experiments on real world data show that RiMOM is promising.

Semantic similarity based ontology cache

This paper addresses the issue of ontology caching on semantic web. The Semantic Web is an extension of the current web in which information is given well-defined meaning, better enabling computers and people to work in cooperation. Ontology serves as the metadata for defining the information on semantic web. Ontology based semantic information retrieval (semantic retrieval) is becoming more and more important. Many research and industrial works have been made so far on semantic retrieval. Ontology based retrieval improves the performance of search engine and web mining. In semantic retrieval, a great number of accesses to ontologies usually lead the ontology servers to be very low efficient. To address this problem, it is indeed necessary to cache concepts and instances when ontology server is running. Existing caching methods from database community can be used in the ontology cache. However, they are not sufficient for dealing with the problem. In the task of caching in database, usually the most frequently accessed data are cached and the recently less frequently accessed data in the cache are removed from it. Different from that, in ontology base, data are organized as objects and relations between objects. User may request one object, and then request another object according to a relation of that object. He may also possibly request a similar object that has not any relations to the object. Ontology caching should consider more factors and is more difficult. In this paper, ontology caching is formalized as a problem of classification. In this way, ontology caching becomes independent from any specific semantic web application. An approach is proposed by using machine learning methods. When an object (e.g. concept or instance) is requested, we view its similar objects as candidates. A classification model is then used to predict whether each of these candidates should be cached or not. Features in classification models are defined. Experimental results indicate that the proposed methods can significantly outperform the baseline methods for ontology caching. The proposed method has been applied to a research project that is called SWARMS.

SWARMS: A New Tool for Domain Exploration in Semantic Web

This paper is concerned with ?domain exploration?. By domain exploration, we mean searching for data on a specific domain which has been modeled by ontology. We have found that exploration needs on a domain can be categorized into data search and navigation, through an analysis of survey results. In our approach, the domain data is organized as instances of domain concepts. The instances are indexed and data search is conducted by index-based search which is also known the keyword based search. In data navigation, users can select an instance in the result of the index based search to display in visualization view. In visualization view, we use graph to visualize the instance and its context. By instance?s context, we mean that other instances which has relations between the selected one. Users can click a node in the graph to navigate to its visualization view. In visualization view, we provide constraint-based search, users can input properties? values to search for instances that has the input value in the certain properties. Other than index-based search, we provide association search for any two instances. By association of two instances, we mean that the possible direct relations and indirect relations between the two instances. In this paper, we will describe SWARMS? architecture, component technologies and its performance in FOAF domain, because FOAF(Friend-Of-A-Friend) [1] domain?s data size is largest among the three domains. The FOAF domain contains more than 90,000 persons and more than 100,000 publications of the people in the domain.

Knowledge Sharing by Grid Technology

Nowadays, the distributed computing mode is more and more popular. A system regularly needs to distribute its data in different places. The knowledge base system is also among the systems that work on dispersed data. The web provides a ubiquitous medium for seamlessly integrating distributed applications, formats and contents, making them well suited for enterprise knowledge management. In this article, we discuss a framework for knowledge sharing by web service and grid technology. We will also mention how other knowledge bases can be integrated into this framework and share their knowledge. Finally we give a prototype system based in this framework and discuss the future work.