Tomoya Noro

News-Topic oriented hashtag recommendation in twitter based on characteristic co-occurrence word detection

Hashtags, which started to be widely used since 2007, are always utilized to mark keywords in tweets to categorize messages and form conversation for topics in Twitter. However, it is hard for users to use hashtags for sharing their opinions/interests/comments for their interesting topics. In this paper, we present a new approach for recommending news-topic oriented hashtags to help Twitter users easily join the conversation about news topics in Twitter. We first detect topic-specific informative words co-occurring with a given target word, which we call characteristic co-occurrence words, from news articles to form a vector for representing the news topic. Then by creating a hashtag vector based on tweets with the same hashtag, we calculate the similarity between these two vectors and recommend hashtags of high similarity scores with the news topic. Experimental results show that our approach could recommend hashtags which are highly relevant to the news topics, helping users share their tweets with others in Twitter.

Partial information extraction approach to lightweight integration on the web

We present partial information extraction approach to light-weight integration on the Web. Our approach allows us to extract dynamic contents created by scripts as well as static HTML contents. Our approach has three application areas: automatic generation of Web services from Web applications, automatic integration of Web applications with Web services on desktop computers, and automatic integration of mobile phone applications with Web applications and Web services on mobile phones.

Personal News RSS Feeds Generation Using Existing News Feeds

Nowadays more and more news sites publish news stories using news RSS feeds for easier access and subscription on the Web. Generally, news stories are grouped by several categories and each category corresponds to one news RSS feed. However there are no uniform standards for categorization. Each news site has its own way of categorization for grouping news stories. These dissimilar categorization can not always satisfy every individual user, and generally the provided categories are not detailed enough for personal using. In this paper, we proposed a method for users to create customizable personal news RSS feeds using existing ones. We implemented a news directory system(NDS) which can retrieve news stories by RSS feeds and classify them. Using this system, we can recategorize news stories from original RSS feeds, or subdivide one RSS feed to a more detailed level. With the classification information for each news article, we offer customizable personal news RSS feeds to subscribers.

Effectiveness of Incorporating Follow Relation into Searching for Twitter Users to Follow

Twitter is one of the most popular microblogging services that facilitate real-time information collection, provision, and sharing. Following influential Twitter users is one way to get valuable information related to a topic of interest efficiently. Recently many researches on this issue have been done and, in general, it is said that the follow relation is not useful for measuring user influence. In this paper, we study effectiveness of incorporating not only the tweet activity (retweet and mention) but also the follow relation into searching for good Twitter users to follow for getting information on a topic of interest. We present a method for finding Twitter users based on both the follow relation and the tweet activity, and show the follow relation could improve the performance as compared with methods based on only the tweet activity.

A state propagation method for consistency checking of web service function invocations in web applications

When we realize Web applications using already existing Web service functions, we often have a situation in which preconditions of Web service function invocations may not be satisfied or necessary data may not be yet provided to produce output Web pages. Traditional approaches to consistency checking of Web service function invocations are use of linear temporal logic formulae and finite-state model based checking. These approaches may not be for every programmers. In this paper we propose an easy-to-use method for consistency checking of Web service function invocations. Our approach uses preconditions and postconditions of Web service function invocations and computation of weakest execution states.