12.01.2015 Views

Download - Academy Publisher

Download - Academy Publisher

Download - Academy Publisher

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

ontology, and the domain dictionary. After that, the<br />

credibility of the extracted competitive intelligence is<br />

evaluated based on a defined social-network-based<br />

credibility model. When users submit queries about<br />

competitive intelligence through the user interface, the<br />

query processing model will retrieve appropriate results<br />

from the competitive intelligence database.<br />

The main function of the competitive intelligence<br />

extraction module is to extract domain-constrained<br />

competitive intelligence from Web pages and further to<br />

deliver them to the credibility evaluation module. We use<br />

an entity-based approach in this module to extract<br />

competitive intelligence, which will be discussed in the<br />

next section.<br />

The credibility evaluation module adopts the socialnetwork-based<br />

method to evaluate competitive<br />

intelligence credibility. The credibility of competitive<br />

intelligence is influenced by a lot of factors. These factors<br />

are classified into two types in our paper, which are the<br />

inner-site factors and inter-site factors. Then we use<br />

different algorithms to evaluate the competitive<br />

intelligence credibility according each type of factors.<br />

Finally we will integrate the both results and make a<br />

comprehensive evaluation on the competitive intelligence<br />

credibility.<br />

The user interface supports keyword-based queries on<br />

competitive intelligence. Users are allowed to input topics,<br />

time, or locations as query conditions.<br />

The query processing module aims at returning<br />

competitive intelligence related with given topics or other<br />

conditions. Competitive intelligence workers can further<br />

process the returned results and produce integrated<br />

competitive intelligence. This module contains two<br />

procedures. The first one is a database retrieval procedure,<br />

and the second is clustered visualization of the results.<br />

The system provides several ways of clustered<br />

visualization, including time-based clustering, locationbased<br />

clustering, and topic-based clustering.<br />

IV. A SPATIOTEMPORAL-EVOLUTION-BASED APPROACH<br />

TO CREDIBILITY EVALUATION<br />

Most Web pages contain time and location information.<br />

That information is useful to evaluate the credibility of<br />

Web-based competitive intelligence. Our approach<br />

consists of two ideas. The first idea is to construct a<br />

timeline for Web-based competitive intelligence, and<br />

further to develop algorithms to determine the credibility<br />

of Web-based competitive intelligence. The second idea<br />

is to construct a location distribution of related Web<br />

pages, which contribute the extraction of Web-based<br />

competitive intelligence. And then we use the location<br />

distribution to measure the credibility of Web-based<br />

competitive intelligence.<br />

Fig.3 and Fig.4 shows the two ideas. In Fig.3, we see<br />

that the number of Web pages related to a certain<br />

competitive intelligence element increased sharply in July,<br />

2009, based on which we can infer that this element is<br />

very possible to be true from July, 2009. While in Fig.4,<br />

we find that in July, 2009, the IP locations of related Web<br />

pages are mostly in China. Suppose that the competitive<br />

intelligence is about China, we can draw the conclusion<br />

that in July, 2009, the competitive intelligence gained the<br />

most attention in China, so it is very possible that the<br />

competitive intelligence is credible.<br />

Figure 2. The timeline evaluation of Web-based competitive<br />

intelligence<br />

Figure 3. The location distribution of Web-based competitive<br />

intelligence<br />

V. CONCLUSIONS AND FUTURE WORK<br />

In this paper we present a framework of extracting and<br />

evaluating Web-based competitive intelligence. Our<br />

future work will concentrate on the design and<br />

implementation of the algorithms mentioned in the paper<br />

and aim at implementing a prototype and conducting<br />

experimental study.<br />

VI. ACKNOWLEDGEMENT<br />

This work is supported by the Open Projects Program<br />

of National Laboratory of Pattern Recognition, the Key<br />

Laboratory of Advanced Information Science and<br />

Network Technology of Beijing, and the National Science<br />

Foundation of China under the grant no. 60776801 and<br />

70803001.<br />

REFERENCES<br />

[1] Akamine, S., Kawahara, D., Kato, Y., et al., (2009)<br />

WISDOM: A Web Information Credibility Analysis<br />

System, Proc. of the ACL-IJCNLP 2009 Software<br />

Demonstrations, Suntec, Singapore, pp. 1-4<br />

[2] Alfarez, A., Hailes S., (1999) Relying On Trust to Find<br />

Reliable Information, Proc. of International Symposium<br />

on Database, Web and Cooperative Systems (DWA-COS)<br />

[3] Deng, D., Luo, L., (2007) An Exploratory Discuss of New<br />

Ways for Competitive Intelligence on WEB 2.0, In W.<br />

Wang (ed.), Proc. of IFIP International Conference on e-<br />

195

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!