Search

Develop a solution approach to resolve the hierarchical rough set problem.Explore the most specific decision attribute level by level in the level-search procedure.Apply the proposed approach to the case in the transportation industry to select green fleets.This corporation reduces pollution emissions over the long-term by choosing the green vehicles in a desired ratio. Rough set theory (RST) has been the subject of much study and numerous applications in many areas. However, most previous studies on rough sets have focused on finding rules where the decision attribute has a flat, rather than hierarchical structure. In practical applications, attributes are often organized hierarchically to represent general/specific meanings. This paper (1) determines the optimal decision attribute in a hierarchical level-search procedure, level by level, (2) merges the two stages, generating reducts and inducting decision rules, into a one-shot solution that reduces the need for memory space and the computational complexity and (3) uses a revised strength index to identify meaningful reducts and to improve their accuracy. The selection of a green fleet is used to validate the superiority of the proposed approach and its potential benefits to a decision-making process for transportation industry.

Framing is a phenomenon that is studied and debated widely in sociology and political science. It refers to the manner in which audiences interpret information and justify their claims or activities. The subconscious influence of framing might lead to opinion changes and social movements. However, multi-frame classification on microblogging data has not yet been investigated. In this study, we aim to classify a large number of posts into frames. We describe in detail the implementation of a new algorithm for multi-frame classification tasks called Frame Dispatcher, which aims to classify microblogging data into frames. In our experiments, we extracted over 15,000 posts from approximately 200 Facebook fan pages concerning an anti-curriculum student movement. The experimental results show that Frame Dispatcher can classify microblogging data into frames efficiently and effectively.

To solve the problem of automatically drawing landmark areas for nice vista points, in this paper, photos metadata of popular urban landmarks or natural landscapes are collected and extracted from Flickr and Google Map. The Landmark Area Recognition System (LARS) is proposed to efficiently recognize astonishing regions of landmarks for exploring nice visiting and photographic POIs (Point of Interests) of these landmarks. Based on crowdsourcing photos and tags obtained from social networks, LARS implements LBSE (Location-Based Search Engine) for searching near objects efficiently. Next, DBCGM (Density-Based Clustering with Greedy Method) is proposed to cluster the landmark photos into regions. Based on crowdsourcing verifications on the photo-landmark relevance, the data set for experiments were collected for experimental evaluations. The result shows that DBCGM outperforms other density-based clustering methods. Finally, LARS employs the Concave Hull algorithm to draw the landmark area on Google Map as the demonstration of LARS applications.

“Energy” has been one element of the development of human civilization, also a power for national industry, construction and economic development. The green energy has become the cornerstone in sustainable development to secure such energy supply but may accommodate opinions from controversial perspectives when this subject is discussed. This study develops an interactive big data system, which aims at aggregating data from Facebook, PTT, news, and provides an interactive interface for energy domain experts. The “interaction” characterizes the seamless integration between users and the system to construct the controversial issue sets of energy, which could be identified and established autonomously in this study. The approach using tags of the link in two controversial issues can help end-users effectively query on demand. The energy relevant issues can be fully aware and provided to the decision makers from the positive and negative viewpoints.

Patent strategy is the overriding mechanism that helps direct investment, resource allocation, expectations, and policy development within an organization. Much studies of patent, for example, patent classification, patent analysis, patent management, patent strategy planning have been made. Due to the high cost of devoting to the research and development for a new patent application, it is essential for a company to develop the patent portfolio based on analyzing related information for fitting with cost constraint and maximizing the benefit. However, only few research attempts to develop new patents with the consideration of resource allocation, for example, optimizing budget utilization. In addition, the previous studies did not derive significant technologies and induct rules for resource allocation through patent analysis. In some cases, the patent analysis may process qualitative information that is difficult to analyze by standard statistical techniques. The rough set approach, which is suitable for processing qualitative information, is required to induct decision rules to derive critical technologies of patents. In this paper, a systematic approach to analyze existing patent information based on rough set theory with the consideration of resource allocation is developed. A case study is presented to demonstrate the contribution of the proposed approach which assists on decision-making in patent reform or invention with constraint resource.

Interoperability among multi-entities (companies) with heterogeneous knowledge sources becomes a research focus in the field of Supply chain management (SCM). Specifically, sharing knowledge among multiple entities in a supply chain is crucial. However, only a few studies have addressed the problem of interoperability and knowledge sharing in supply chains. Current technologies, such as EDI, RosettaNet or the current Web, are useful for sharing data/information, rather than knowledge. This paper proposes a solution for sharing knowledge with the semantic web. The solution involves (i) a semi-structured knowledge model to represent knowledge in not only an explicit and sharable, but also a meaningful format, (ii) an agent-based annotation process to resolve issues associated with the heterogeneity of knowledge documents, and (iii) an articulation mechanism to improve the effectiveness of interoperability between two heterogeneous ontologies. Based on the proposed solution, entities in a supply chain can represent, seek, and share knowledge effectively.

We study the problem of mining the informative structure of a news Web site that consists of thousands of hyperlinked documents. We define the informative structure of a news Web site as a set of index pages (or referred to as TOC, i.e., table of contents, pages) and a set of article pages linked by these TOC pages. Based on the Hyperlink Induced Topics Search (HITS) algorithm, we propose an entropy-based analysis (LAMIS) mechanism for analyzing the entropy of anchor texts and links to eliminate the redundancy of the hyperlinked structure so that the complex structure of a Web site can be distilled. However, to increase the value and the accessibility of pages, most of the content sites tend to publish their pages with intrasite redundant information, such as navigation panels, advertisements, copy announcements, etc. To further eliminate such redundancy, we propose another mechanism, called InfoDiscoverer, which applies the distilled structure to identify sets of article pages. InfoDiscoverer also employs the entropy information to analyze the information measures of article sets and to extract informative content blocks from these sets. Our result is useful for search engines, information agents, and crawlers to index, extract, and navigate significant information from a Web site. Experiments on several real news Web sites show that the precision and the recall of our approaches are much superior to those obtained by conventional methods in mining the informative structures of news Web sites. On the average, the augmented LAMIS leads to prominent performance improvement and increases the precision by a factor ranging from 122 to 257 percent when the desired recall falls between 0.5 and 1. In comparison with manual heuristics, the precision and the recall of InfoDiscoverer are greater than 0.956.

This paper presents an intelligent Internet information system, Automatic Classifier for the Internet Resource Discovery (ACIRD), which uses machine learning techniques to organize and retrieve Internet documents. ACIRD consists of a knowledge acquisition process, document classifier, and two-phase search engine. The knowledge acquisition process of ACIRD automatically learns classification knowledge from classified Internet documents. The document classifier applies learned classification knowledge to classify newly collected Internet documents into one or more classes. Experimental results indicate that ACIRD performs as well or better than human experts in both knowledge acquisition and document classification. By using the learned classification knowledge and the given class lattice, the ACIRD two-phase search engine responds to user queries with hierarchically structured navigable results (instead of a conventional flat ranked document list), which greatly aids users in locating information from numerous, diversified Internet documents.

We propose a novel method that integrates dictionary, heuristics and data mining approaches to efficiently and effectively recognize exact protein names from the literature. According to the protein name dictionary and heuristic rules published in related studies, core tokens of protein names can be efficiently detected. However, exact boundaries of protein names are hard to be identified. By regarding tokens of a protein name as items within a transaction, we apply mining associations to discover significant sequential patterns (SSPs) from the protein name dictionary. Based on SSPs, protein name parts are extended from core tokens to left and right boundaries for correctly recognizing the protein name. Based on Yapex101 corpus, Protein Name Recognition System (PNRS) achieves the F-score (74.49%) better than existing systems and papers.

It is very convenient to carry smartphones and take photos on the trip. With Apps installed in smartphones, users can easily share photos to friends on social networks. Huge amount of interesting photos are therefore accumulated through smartphones without facilities to manage these photos. We propose an integrated system that provides a mobile App for photographing on the trip, a desktop App for synchronizing photos with the web platform for sharing and organizing photos. While photographing, the web service recommends significant tags to photos and gets manual tags, the App transparently accumulates context-information for travel photos. Then, the desktop App facilitates users to collect and choose interesting photos for sharing and storing on the web platform. Finally, the system applies search engine and web mining techniques to extract textual sentences from pages that contain relevant information about photos for writing travel blogs easily and efficiently.

The data-driven characteristic of the Version Space rule-learning method works efficiently in memory even if the training set is enormous. However, the concept hierarchy of each attribute used to generalize/specialize the hypothesis of a specific/general (S/G) set is processed sequentially and instance by instance, which degrades its performance. As for ID3, the decision tree is generated from the order of attributes according to their entropies to reduce the number of attributes in some of the tree paths. Unlike Version Space, ID3 generates an extremely complex decision tree when the training set is enormous. Therefore, we propose a method called AGE (A_RCH+OG_L+ASE_, where ARCH="Automatic geneRation of Concept Hierarchies", OGL="Optimal Generalization Level", and ASE="Attribute Selection by Entropy"), taking advantages of Version Space and ID3 to learn rules from object-oriented databases (OODBs) with the least number of learning features according to the entropy. By simulations, we found the performance of our learning algorithm is better than both Version Space and ID3. Furthermore, AGE's time complexity and space complexity are both linear with the number of training instances.

As the rapid development of smart phones and applications, users are familiar to take photos with smart phones on the trip and share photos to friends on social networks. Those photos are usually tagged and shared with trivial location information, such as attractions derived from mobile locations contributed from Facebook users. However, these attractions are not useful for enhancing the photo semantics. In this paper, we propose the Tag Recommendation System (TRS) that automatically mine significant tags from Flickr's photos so that these tags can be recommended to improve the semantics of photos according to location information. Experiments based on the user experience show that TRS can achieve about 85% satisfaction rates.

International audience; Automatic emotion sensing in textual data is crucial for the development of intelligent interfaces in many interactive computer applications. This paper describes a high-precision, knowledgebase-independent approach for automatic emotion sensing for the subjects of events embedded within sentences. The proposed approach is based on the probability distribution of common mutual actions between the subject and the object of an event. We have incorporated web-based text mining and semantic role labeling techniques, together with a number of reference entity pairs and hand-crafted emotion generation rules to realize an event emotion detection system. The evaluation outcome reveals a satisfactory result with about 85% accuracy for detecting the positive, negative and neutral emotions.

The explosive growth of the World Wide Web continues to revolutionize information editing, publishing and accessing patterns. Within the Web infrastructure, individuals can easily edit and publish documents that contain hyperlinks to other documents published by the same or other Web sites. As a result, the Web contains information on almost any subject available anywhere to anyone at anytime. However, this explosive information growth has made the task of finding information like trying to find a needle in a haystack. Although directory services (like Yahoo!1) and search engines (like Google2) facilitate information searches, many users still have difficulty locating useful information. Browsing directories is time consuming as there are a seemingly infinite number of possible topics. For example, Open Directory (currently the largest directory database) contains over 460,000 categorics3. Users must click and click and click to find a target directory and browse documents. Furthermore, the construction of directories is labor-intensive and the directory service cannot keep up with Web growth. Finding documents using search engines is frustrating as search results usually contain thousands of links. Although some search engines like Google apply hyperlink analysis to provide better ranking, it is still of ten ineffective.

In this paper, we study the problem of mining the informative structure of a news Web site which consists of thousands of hyperlinked documents. We define the informative structure of a news Web site as a set of index pages (or referred to as TOC, i.e., table of contents, pages) and a set of article pages linked by TOC pages through informative links. It is noted that the Hyperlink Induced Topics Search (HITS) algorithm has been employed to provide a solution to analyzing authorities and hubs of pages. However, most of the content sites tend to contain some extra hyperlinks, such as navigation panels, advertisements and banners, so as to increase the add-on values of their Web pages. Therefore, due to the structure induced by these extra hyperlinks, HITS is found to be insufficient to provide a good precision in solving the problem. To remedy this, we develop an algorithm to utilize entropy-based Link Analysis on Mining Web Informative Structures. This algorithm is referred to as LAMIS. The key idea of LAMIS is to utilize information entropy for representing the knowledge that corresponds to the amount of information in a link or a page in the link analysis. Experiments on several real news Web sites show that the precision and the recall of LAMIS are much superior to those obtained by heuristic methods and conventional ink analysis methods.

In this paper, we present our design and implementation of ASIS multimedia digital library. Besides bibliographic information, our service includes an archive of digital audio and video streams, CD titles, and WWW contents. This system not only aims at end customer supports, e.g., searching, streaming, and browsing operations, but it also facilitates admission control, content production, distribution processes, and license control, etc. We address several design issues in designing a multimedia digital library, e.g., the need for accounting support and system management, for integrating third party system components and varied kind of multimedia, and for a computer aided tool to automatically process the backend production. Our solutions to these problems are presented.