Yuki Arase

Mining people's trips from large scale geo-tagged photos

Photo sharing is one of the most popular Web services. Photo sharing sites provide functions to add tags and geo-tags to photos to make photo organization easy. Considering that people take photos to record something that attracts them, geo-tagged photos are a rich data source that reflects peoples memorable events associated with locations. In this paper, we focus on geo-tagged photos and propose a method to detect peoples frequent trip patterns, i.e., typical sequences of visited cities and durations of stay as well as descriptive tags that characterize the trip patterns. Our method first segments photo collections into trips and categorizes them based on their trip themes, such as visiting landmarks or communing with nature. Our method mines frequent trip patterns for each trip theme category. We crawled 5.7 million geo-tagged photos and performed photo trip pattern mining. The experimental result shows that our method outperforms other baseline methods and can correctly segment photo collections into photo trips with an accuracy of 78%. For trip categorization, our method can categorize about 80% of trips using tags and titles of photos and visited cities as features. Finally, we illustrate interesting examples of trip patterns detected from our dataset and show an application with which users can search frequent trip patterns by querying a destination, visit duration, and trip theme on the trip.

A game based approach to assign geographical relevance to web images

Geographical context is very important for images. Millions of images on the Web have been already assigned latitude and longitude information. Due to the rapid proliferation of such images with geographical context, it is still difficult to effectively search and browse them, since we do not have ways to decide their relevance. In this paper, we focus on the geographical relevance of images, which is defined as to what extent the main objects in an image match landmarks at the location where the image was taken. Recently, researchers have proposed to use game based approaches to label large scale data such as Web images. However, previous works have not shown the quality of collected game logs in detail and how the logs can improve existing applications. To answer these questions, we design and implement a Web-based and multi-player game to collect human knowledge while people are enjoying the game. Then we thoroughly analyze the game logs obtained during a three week study with 147 participants and propose methods to determine the image geographical relevance. In addition, we conduct an experiment to compare our methods with a commercial search engine. Experimental results show that our methods dramatically improve image search relevance. Furthermore, we show that we can derive geographically relevant objects and their salient portion in images, which is valuable for a number of applications such as image location recognition.

A web browsing system for cellular-phone users based on adaptive presentation

Cellular phones are widely used to access the Web. However, most available Web pages are designed for desktop PCs, and it is inconvenient to browse these large Web pages on a cellular phone with a small screen and poor interfaces. Users who browse a Web page on a cellular phone have to scroll through the whole page to find the desired content, and must then search and scroll within that content in detail to get useful information. This paper describes the design and implementation of a novel Web browsing system for cellular phones. This system includes a Web page overview to reduce scrolling operations when finding objective content within the page. Furthermore, it adaptively presents content according to its characteristics to reduce burdensome operations when searching within content.

User activity understanding from mobile phone sensors

Context acquisition is an important technology for ubiquitous computing. An ideal approach would be easy to deploy and non-intrusive to peoples life. Mobile phones equipped with advanced sensors are preferable platform owing to their user-friendliness and freedom from extra costs to deploy. In this study, we propose to use a mobile phone to detect user contexts. We formally define the concept of context and then describe applications that leverage peoples long-term activity, which can be inferred from their contexts.

Mobile search assistance from HCI aspect

Mobile internet access has become an important technology to our life. According to this trend, search engines have been providing search services for mobile phones. Since a mobile phone has a limited interface, such as a small screen and keypad, even simple interactions are burdensome for users. Therefore, to provide comfortable web browsing experience, efforts on HCI aspect are also essential. In this paper, we describe our studies that tackle this problem by HCI-based approach. We propose systems that reduce the number of operations on inputting a query to a search engine, enable clipping and saving a web content, and assist cooperative web search conducted by multiple mobile users.

A web browsing system based on adaptive presentation of web contents for cellular phones

Cellular phones have already been widely used to access the Web. However, most existing Web pages are designed for desktop PCs, and thus, it is inconvenient to browse these large Web pages on a cellular phone with a small screen and poor interfaces. Users who browse a Web page on a cellular phone have to scroll the whole page to find an objective content, and then, have to scroll within the content in detail to get useful information. In this paper, we propose a novel browsing system to break off these burdensome operations by adaptively presenting Web contents according to their characteristics.

Exploiting Twitter for Spiking Query Classification

We propose a method for classifying queries whose frequency spikes in a search engine into their topical categories such as celebrities and sports. Unlike previous methods using Web search results and query logs that take a certain period of time to follow spiking queries, we exploit Twitter to timely classify spiking queries by focusing on its massive amount of super-fresh content. The proposed method leverages unique information in Twitter—not only tweets but also users and hashtags. We integrate such heterogeneous information in a graph and classify queries using a graph-based semi-supervised classification method. We design an experiment to replicate a situation when queries spike. The results indicate that the proposed method functions effectively and also demonstrate that accuracy improves by combining the heterogeneous information in Twitter.

Bayesian nonparametric modeling of user activities

Human activity modeling is becoming more and more important in ubiquitous computing as it builds a foundation for higher-level applications in areas such as e-health and activity recommendation systems. Many existing works in this area focus on recognizing a pre-defined set of activities using some devices in the supervised learning setting, however, it is hard to define activities and label sensor data, especially for a new environment. In this note we aim to recognize activities in an unsupervised way - segment activity sensor reading sequence and group the segments into meaningful categories by leveraging Sticky HDP-HMM. We have conducted experiments on a sensor dataset collected in an office area using a smartphone and the result shows that our method frees annotation process and renders good activity recognition result.

Content comparison functions for mobile co-located collaborative web search

Due to the recent popularization of mobile devices, such as mobile phones and PDAs, most people have their own mobile devices in Japan. In this situation, a user can search the Web not only by himself, but together with his friends and families, which aims to find information that meets requirements from all of them. In such collaborative search, users first search the Web separately and collect contents of interest, and then, share the contents with others by showing a screen with each other. However, because mobile devices generally have a small screen, this process is burdensome for users. In this paper, to solve this problem, we propose an interface to support users engaged in collaborative search to share and compare their collected contents. The proposed interface provides functionalities to effectively share contents collected by all members and to add reviews in order to make content search and comparison easier.

Wikipedia Page View Reflects Web Search Trend

The frequency of a web search keyword generally reflects the degree of public interest in a particular subject matter. Search logs are therefore useful resources for trend analysis. However, access to search logs is typically restricted to search engine providers. In this paper, we investigate whether search frequency can be estimated from a different resource such as Wikipedia page views of open data. We found frequently searched keywords to have remarkably high correlations with Wikipedia page views. This suggests that Wikipedia page views can be an effective tool for determining popular global web search trends.