Discussions and Limitations

There are some limitations in this work. First, to capture trust information in the real world, we quantify it by asking users to assign the trust values. The invasive requirements toward users thus may cause some disfavor and the trustworthy issues (i.e. some misleading or skewed situations of recommender system). Obviously, the phenomenon that over than half of objects is isolated will debase the value of SIP score. This may causes the recommendation score lays particular stress on the other two scores (i.e. TR and SS), which distort the recommendation scores and denotation of recommendation mechanism. Second, our trust models are constructed on an agent-to-agent level which cannot reflect trustworthiness in an object-to-object level. That is, with regard to objects of certain agent, we treat each object as the same trust level and each of them has the same trust value relative to the requester. In the future work, we will design a more comprehensive trust model to tackle with this issue to induce a complete and robust recommendation mechanism.

As to SS score, we design an interface for the requester to select some of his/her posts to compare content similarity with others. Unlike search engine, some brief keywords would induce numbers of results which make users hard to digest and unable to find what they really want. More index terms would be helpful for users to accurately locate the needed information [27]. In this work, article selecting process (i.e. select the posts to list into comparison target) would indeed increase the efficiency and accuracy of calculation of semantic similarity. As for processing procedures of Chinese words, each step could be refined and advanced for more accuracy calculation of SS scores.

In recommendation strategies, four existing recommendation (i.e. Random, Comment, Citation, Hotness) approaches applied by Wretch is used here as benchmark approaches.

From the experimental evaluation results, the neural model (ANN+All) and the linear model (All) outperform the others. Still, we may wonder that if we take the traditional collaborative-based filtering, content-based filtering, or even other recommendation techniques into the comparison approaches, is the proposed model still has a better performance than the others? We should extend the recommendation strategies for further comparison for a more convincible evaluation in future work.

Moreover, we can observe that the MAPE performance of the model seems insignificant under the limitation of the insufficient training data, even though the recommendation prediction model still outperformed the other approaches, including the synthetical approach without BPNN training.

6.3 Future Works

Recommendation is an interesting topic in blog applications. Depending on their preferences and interests, the synthetical blog recommendation system will help bloggers to find out not only interesting blogs and blog posts but also trustworthy and socially homoeo-bloggers. In future works, there are still several issues in blog application and social networking:

First, finding the influential bloggers for marketing is attention demanding. Finding marketing influential bloggers for marketing will not only allow us to better understand the interesting activities happening in a social network, but also present unique opportunities for industry, sales, and advertisements. With the advent of online social network, viral marketing/word-of-mouth is increasingly being recognized as a crucial issue in social influence and marketing domains. Especially on blogs, it provides a finest platform both for

advertisers to market a new product/service and for customers to locate the purchasing suggestions and comments. In conclusion, finding influential blog sites in the blogosphere is an important research problem, which investigates how these blog sites influence the external world and within the blogosphere [8] In future works, we will address a novel problem of finding influential bloggers for marketing on the blogosphere by proposing a preliminary MIV (Marketing Influential Value) model. We will induce two dimensions, network-based factors, and content-based factors, to identify the potential marketing-aided nodes to help marketers/advertisers in promoting their products/services with less efforts and costs.

Second, proposing a dynamic blog recommender system. There exists a tradeoff between a precise recommender system and computational efficiency. A precise recommendation must gather as more information as possible from bloggers; however, it will result in decreasing of the computation ability as well. We should develop an efficient approach and process, to make the recommendation more realistic and to update the relationships dynamically. As to the computational ability, recommendation may perform better by searching for a more scalable tools and technologies. Such as, cloud computing technique can handle the data processing and computation well under large amount of data. In conclusion, a scalable recommender system is needed in blog service of the real-world application.

Third, integrating the social relations in different social networking services is interesting.

In the era of web2.0, users may use many social webs to satisfy their own needs. Upload the images to an image sharing website, join a online community on a social networking site, write a review in a product information sharing site, or publish a diary on their own blogs. It is interesting if we collect all information in these sites and aggregate the social relations among them. We could find out the domain experts, influential nodes, or authorities by proposing a ranking mechanism. We believe that the idea is quite promising and proactive in applications of social-networked service.

References

[1] Adar, E., Zhang, L., Adamic, L., & Lukose, R. (2004). Implicit structure and the dynamics of blogspace. Workshop on the Weblogging Ecosystem: Aggregation, Analysis and Dynamics, WWW.

[2] ALi-Hasan, N., & Adamic, L. (2007). Expressing Social Relationships on the Blog through Links and comments, ICWSM.

[3] Berendt, B., & Navigli, R. (2006). Finding your way through blogspace: Using semantics for cross-domain blog analysis, In AAAI Symposium on Computational Approaches to Analyzing Weblogs.

[4] Brin, S., and Page, L. (1998). The Anatomy of a large-scale hypertextual web search engine, In proceedings of 7th international World Wide Web Conference.

[5] Fujimura, K., Inoue, T., & Sugisaki, M. (2005). The EigenRumer algorithm for ranking blogs, 2nd Annual Workshop on the Weblogging Ecosystem: Aggregation, Analysis and Dynamics, WWW.

[6] Golbeck, J., & Hendler, J. (2006). FilmTrust: Movie Recommendations using Trust in web-based Social Networks, IEEE Consumer Communications and Networking Conference.

[7] Golbeck, J. (2006). Trust and nuanced profile similarity in online social networks, Journal of Artificial Intelligence Research.

[8] Kathy E. Gill. How can we measure the influence of the blogosphere? In Proceedings of the WWW’04: workshop on the Weblogging Ecosystem: Aggregation, Analysis and

Dynamics, 2004.

[9] Keinberg, J. M. (1999). Authoritative sources in hyperlinked environment, Proceedings of the ninth annual ACM-SIAM symposium on Discrete algorithms, 46(5).

[10] Kolari, P., Finin, T., & Lyons, K. (2007). On the structure, properties and utility of internal corporate blogs, ICWSM.

[11] Kritikopoulos, A., Sideri, M., & Varlamis, I. (2006). BlogRank: ranking weblogs based on connectivity and similarity features, Proceedings of the 2nd international workshop on Advanced architectures and algorithms for internet delivery and applications, 198.

[12] Lee, H.-Y., Ahn, H., & Han, I. (2007). VCR: Virtual community recommender using the technology acceptance model and the user’s needs type, Expert Systems with Applications, 33(4), 984-995(12).

[13] Lin, Y.-R., Sundaram, H., Chi, Y., Tatemura, J., & Tseng, B. (2006). Discovery of blog communities based on mutual awareness, 3rd Annual Workshop on the Weblogging

Ecosystem.

[14] Ma, W.-Y., & Chen, K.-J. (2003). Introduction to CKIP Chinese word segmentation system for the first international Chinese word segmentation bakeoff, Proceedings of the Second SIGHAN Workshop on Chinese Language Processing, 168-171. Online demo system: http://ckipsvr.iis.sinica.edu.tw/

[15] Manning, C.D., and Schutze, H. (1999). Foundations of statistical natural language processing, MIT Press, Cambridge, MA.

[16] Massa, P., & Bhattacharjee, B. (2004). Using trust in recommender systems: an experimental analysis, In Proc. of 2nd Int. Conference on Trust Management.

[17] Matsuo, Y., & Yamamoto, H. (2007). Diffusion of Recommendation through a Trust Network, ICWSM.

[18] Negnevitsky, M., (2005). Artificial Intelligence- A Guide to Intelligent System, Second edition published, Person education, Ltd, pp.175~180.

[19] O’Donovan J, & Smyth, B. (2005). Trust in recommender systems, Proceedings of the 10th international conference on Intelligent user interfaces, 167-174.

[20] Papagelis, M., Plexousakis, D., & Kutsuras, T. (2005). Alleviating the Sparsity Problem of Collaborative Filtering Using Trust Inference, In Proceedings of iTrust, 224–239.

[21] Song, W., Phoha, V. V. (2005). Opinion filtered recommendation trust model in

peer-to-peer networks, Lecture Notes in Computer Science, 237-244.

[22] Tsai, T.-M., Shih, C.-C., & Chou, S.-C. T. (2006). Personalized Blog Recommendation:

Using the value, semantic, and Social Model, Innovations in Information Technology, 1-5.

[23] Weihua, S., Phoha, V. V, & Xu X. (2004). An Adaptive Recommendation Trust Model in Multiagent System, Proceedings of the IEEE/WIC/ACM International Conference on

Intelligent Agent Technology (IAT' 04), 9, 462-465.

[24] Wikipedia, Neural network: http://en.wikipedia.org/wiki/Neural_network [25] Wikipedia, Wretch: http://en.wikipedia.org/wiki/Wretch_%28website%29 [26] Wretch blog: http://www.wretch.cc/blog/

[27] Yang, K., Yu, N., Valerio, A., Zhang, H., & Ke, W. (2007). Fusion approach to finding opinions in blogosphere, ICWSM.

[28] Zou, H., Xia, G., Yang, F., & Yang, H. (2007). A neural network model based on the multi-stage optimization approach for short-term food price forecasting in China.

Appendix A

The User Interface of Trust Score Form

Appendix B

The User Interface of User Evaluation Form