Innovation Detection Based on User-Interest Ontology of Blog Community

Abstract

Abstract. Recently, the use of blogs has been a remarkable means to publish user interests. In order to find suitable information resources from a large amount of blog entries which are published every day, we need an information filtering technique to automatically transcribe user interests to a user profile in detail. In this paper, we first classify user blog entries into service domain ontologies and extract interest ontologies that express a user's interests semantically as a hierarchy of classes according to interest weight by a top-down approach. Next, with a bottom-up approach, users modify their interest ontologies to update their interests in more detail. Furthermore, we propose a similarity measurement between ontologies considering the interest weight assigned to each class and instance. Then, we detect innovative blog entries that include concepts that the user has not thought about in the past based on the analysis of approximated ontologies of a user's interests. We present experimental results that demonstrate the performance of our proposed methods using a large-scale blog entries and music domain ontologies.

An Information Diffusion Based Recommendation Framework for Micro-Blogging

“…Ontology is defined as a formal specification of a shared conceptualization consisting of entities, attributes, and relationships. Nakatsuji, Miyoshi, and Otsuka (2006) used Web Ontology Language to extract userinterest ontology from blog articles. Then they applied an ontology-based similarity measurement to cluster bloggers whose interests were alike.…”

Extracting Know-Who/Know-How Using Development Project-Related Taxonomies

“…In [19], [20], we proposed user interest extraction from users' blog entries following the taxonomy of content items that is created by content providers, and similarity measurement between those extracted user interests. The evaluation in [19], [20] shows that the proposed method can accurately extract user interests from blogs and measure the similarity of users.…”

“…The evaluation in [19], [20] shows that the proposed method can accurately extract user interests from blogs and measure the similarity of users. In this paper, we apply the above previously proposed method to knowledge extraction from the mail messages accumulated in mailing lists.…”

“…We can apply our previous method of classifying blog entries into taxonomies [19], [20], to classify e-mails into the concepts of the small taxonomies. However, we should consider that e-mails differ from blog entries in at least one characteristic; i.e., we discuss various topics in an e-mail thread exchanged over a period of time.…”

“…They then calculate the similarity between users based on the covariance of vector V of users. In our previous work [19], [20], we proposed a method of extracting user interests according to a taxonomy of content items by analyzing the descriptions of users in their blog entries, and measured the similarities between the interests of users. Combining the extracted user knowledge based on concepts with data-driven concepts yields more accurate similarity measurements.…”