Hideyuki Fujita

A ubiquitous photo mapping considering users' lines of sight

This paper proposes a new framework for mapping and retrieving photographs, maps and cyberspaces to each other. Our target photographs are enhanced with spatial metadata such as geographic coordinates where they were taken and directions where they focused on. We assume photographs having such spatial metadata become popular. In a common framework, such photographs are mapped to their viewpoints by using their location information generated by GPS. We think this framework has less function for practical spatial queries. For example, even if a photograph is mapped to a certain point on a map, a scene of the point may not be shown in the photograph. It may show a different direction than user wants to see. A problem is that though locations or objects shown in photographs are important for users, viewpoints are not positions of them but positions of cameras from which photographs were taken. We therefore map each photograph to a vector from its viewpoint to its gazing point, and named the vector as photo vector. A prototype system based on our framework provides functions such as handling advanced spatial queries for retrieving photographs, visualizing how many photographs show each location, and mapping text labels having URLs and geographic coordinates to the appropriate positions on photographs. In this framework, photographs, maps and cyberspaces are mapped to each other.

A mobile phone-based exploratory citizen sensing environment

Coping with ill-structured problems in a city involves continuous, opportunistic, and multi-perspective processes, which existing pervasive technologies for citizen participation cannot easily support. Based on two preliminary case studies, we propose Scene Memo, a mobile phone-based exploratory citizen-sensing environment that uses dynamically shared tags to provide social cues and scaffold participants.

Geo-tagged Twitter collection and visualization system

Mobile social media generate valuable data for analyzing human behavior and events in the real world. In this study, we developed a distributed system for collecting geo-tagged data from Twitter. The proposed system can collect several times as much data as commonly used methods. We also developed a spatio-temporal visualization tool for displaying the collected data. We conducted a data-collection and visualization experiment in central Tokyo and showed that the collected data reflected many events in the real world.

Place-tagged Podcasts with Synchronized Maps on Mobile Media Players

PodWalks are a new kind of Podcasts where virtual narrators guide users through real-world spaces. PodWalks can increase a users appreciation of real spaces by connecting them to audio tours stored on portable media players that can be listened to any time and anywhere without using mobile telecommunication services. One disadvantage of PodWalks is that it can be difficult for listeners to synchronize their locations with the location that the narrator is describing and, as a result, listeners become disoriented. To solve this problem we propose a new framework—maPodWalks—as an extension to PodWalks which provides maps that are synchronized to the narrations on mobile media players. By forming linkages between virtual and real-world experiences, maPodWalks help users develop mental maps of places of interest. In this paper, we discuss the characteristics of maPodWalks and present a prototype maPodWalk Maker that allows users to easily create geocoded audio tours. The results of system tests comparing the performance of PodWalks and maPodWalks are reported as well.

Context Weaver

Mobile crowdsourcing allows people to collect data using a large pool of participants. In this paper, we focus on mobile crowdsourcing for citizens to solve local issues in context. We argue that such crowdsourcing environments need to support exploration, a continuous, opportunistic, and multi-perspective process that existing crowd sensing systems cannot easily support. We have developed a system called Context Weaver, which connects participants using networked mobile devices in order to support collaborative exploration, and conducted field trials to understand the effect of networking participants in the crowdsourced data-collection activities that encompass planning, execution, and analysis phases. We discuss a methodology for exploratory mobile crowdsourcing by citizens based on the provision of mutual awareness and rapid feedback in context. The proposed methodology can provide a basis for a model of networked mobile crowdsourcing which can exploit not only the man-power but also the creativity of citizens to gather relevant data.

From crowding detection to community fieldwork: supporting sensing work in context

We describe our experiences with the prototype crowd sensing environments for supporting crowding detection and community fieldwork, and discuss the need to support sensing work in context. Sensing work is inseparable from the shifting observation modes in the overall inquiry process.

A User Study of a Map-Based Slideshow Editor

Abstract Our research goal is to facilitate the sharing of stories with digital maps. Today, many kinds of “my map” applications allow users to create maps that display personal collections of places visited, along with photographs, videos, and texts. The authors of such maps can organize and publish their collections of places by creating maps, and viewers can browse these collections freely by panning and zooming the maps or searching by location. It is difficult, however, for authors to represent their experiences as a sequence of events, mainly because it is up to the viewer to decide which map locations will be viewed, and in what order. Therefore, one future requirement for this type of map application is to facilitate map-based storytelling. The objective of this study is to identify and discuss the characteristics of map-based stories and the effectiveness of maps in editing them. To help users communicate map-based stories in a more narrative fashion, we have developed software for mapping photo ...

Skip Search Approach for Mining Probabilistic Frequent Itemsets from Uncertain Data

Due to wider applications of data mining, data uncertainty came to be considered. In this paper, we study mining probabilistic frequent itemsets from uncertain data under the Possible World Semantics. For each tuple has existential probability in probabilistic data, the support of an itemset is a probability mass function (pmf). In this paper, we propose skip search approach to reduce evaluating support pmf for redundant itemsets. Our skip search approach starts evaluating support pmf from the average length of candidate itemsets. When an evaluated itemset is not probabilistic frequent, all its superset of itemsets are deleted from candidate itemsets and its subset of itemset is selected as a candidate itemset to evaluate next. When an evaluated itemset is probabilistic frequent, its superset of itemset is selected as a candidate itemset to evaluate next. Furthermore, our approach evaluates the support pmf by difference calculus using evaluated itemsets. Thus, our approach can reduce the number of candidate itemsets to evaluate their support pmf and the cost of evaluating support pmf. Finally, we show the effectiveness of our approach through experiments.

Pairwise Expansion: A New Topdown Search for mCK Queries Problem over Spatial Web

This paper focuses on the problem of m-Closest Keywords (mCK) queries over spatial web objects. An \(mCK\ query\) is to find the optimal set of objects (object-set) in the sense that they are the spatially-closest records and satisfy m user-given keywords. We propose a new approach called Pairwise Expansion to find an exact solution of mCK queries based on topdown search of an on-the-fly quad-tree. This approach first enumerates object-pairs in a topdown way, then picks up each ‘closer’ object-pair and expands it into candidate object-sets. Experimental results show that this approach is more efficient than existing topdown search strategies and applicable for real spatial web data.

A new algorithm for m-closest keywords query over spatial Web with grid partitioning

In this paper, we focus on the issue of the m-closest keywords (mCK) query over spatial data in the Web. The mCK query is a problem to find the optimal set of records in the sense that they are the spatially-closest records that satisfy m user-given keywords. The mCK query was proposed by Zhang et al[1]. They assumed a specialized R*-tree to store all records and proposed an Apriori-based enumeration of MBR-combinations. However, this assumption of the prepared R*-tree is not always applicable; Twitter or Flickr provides only records having position information without any prepared data-partitioning. Many services like Google Maps only provide grid partitioning at most. Thus, in this paper, we do not expect any prepared data-partitioning, but assume that we create a grid partitioning from necessary data only when an mCK query is given. Under this assumption, we propose a new search-strategy termed Diameter Candidate Check (DCC), and show that DCC can efficiently find a better set of grid-cells at an earlier stage of search, thereby reducing search space greatly.