Shun Hattori

Mining the web for hyponymy relations based on property inheritance

Concept hierarchies, such as hyponymy and meronymy relations, are very important for various natural language processing systems. Many researchers have tackled how to mine very large corpora of documents such as the Web for them not manually but automatically. However, their methods are mostly based on lexico-syntactic patterns as not necessary but sufficient conditions of concept hierarchies, so they can achieve high precision but low recall when using stricter patterns or they can achieve high recall but low precision when using looser patterns. In this paper, property inheritance from a concept to its hyponyms is assumed to be necessary and sufficient conditions of hyponymy relations to achieve high recall and not low precision, and we propose a method to acquire hyponymy relations from the Web based on property inheritance.

Mining the web for appearance description

This paper presents a method to extract appearance descriptions for a given set of objects. Conversion between an object name and its appearance descriptions is useful for various applications, such as searching for an unknown object, memory recall support, and car/walk navigation. The method is based on text mining applied to Web search results. Using a manually constructed dictionary of visual modifiers, our system obtains a set of pairs of a visual modifier and a component/class for a given name of object, which best describe its appearance. The experimental results have demonstrated the effectiveness of our method in discovering appearance descriptions of various types of objects.

Automatic generation of multimedia tour guide from local blogs

It has recently become a common practice for people to post their sightseeing experiences on weblogs (blogs). Their blog entries often contain valuable information for potential tourists, who can learn about various aspects not found on the official websites of sightseeing spots. Bloggers provide images, videos and texts regarding the places they visited. This implies that popular travel routes could be extracted according to the information available in blogs. In this paper, we describe a system that extracts typical visitors travel routes based on blog entries and that presents multimedia content relevant to those routes. Typical travel routes are extracted by using a sequential pattern mining method. We also introduce a new user interface for presenting multimedia content along the route in a proactive manner. The system works as an automatically generated tour guide accessible from a PC or a mobile device.

Secure Spaces and Spatio-Temporal Weblog Sensors with Temporal Shift and Propagation

This paper defines three kinds of Weblog Sensors to mine the Web, especially CGM such as Weblog documents for spatio-temporal data about a target phenomenon in the physical world, and tries to validate the potential and reliability of these Weblog Sensors’ spatio-temporal data by measuring the correlation with weather statistics of Japan Meteorological Agency as real-world data.

Object-oriented Semantic and Sensory Knowledge Extraction from the Web

Automatic knowledge extraction from such a very large document corpus as the Web is one of the hottest research topics in the domain of Artificial Intelligence and Database technologies. This chapter introduces my object-oriented and the existing methods to extract semantic (e.g., hyponymy and meronymy) and sensory (e.g., visual and aural) knowledge from the Web, and compares them by showing several experimental results. My object-oriented semantic knowledge extraction is based on property inheritance(s) and property aggregation, and repeatedly improves the extracted results of both hyponymy and meronymy relations. Meanwhile, my object-oriented sensory knowledge extraction is improved by utilizing the extracted hyponymy and meronymy relations. Finally, this chapter introduces my Sense-based Object-name Search (SOS) to enable users to identify the concrete name of a target object which they do not know only by inputting its hyponym (class-name) and some sensory descriptions, as an application system to utilize the Web-extracted semantic and sensory knowledge.

Ability-based expression control for secure spaces

In public spaces, there are a number of different contents such as visitors, physical information resources, and virtual information resources via their embedded output devices. Therefore, we might unexpectedly enter the public spaces that have our unauthorized contents and/or unwanted characteristics, i.e., they are not always secure and safe. To solve this problem, my previous work has introduced the concept of “Secure Spaces”, physical environments in which any visitor is protected from being pushed her unwanted information resources on and also any information resource is always protected from being accessed by its unauthorized visitors, and the model and architecture for space entry control and information access control based on their dynamically changing contents. Aiming to build more flexible Secure Spaces, this paper proposes an extended model for not only spatial entry control but also Ability-Based Expression Control according to how preferentially a virtual information resource should be outputted in a Secure Space where shared by visitors with perceptibility and understandability of the virtual information resources content and expression.

Peculiar image search by Web-extracted appearance descriptions

We have become able to get enough approvable images of a target object just by submitting its object-name to a conventional keyword-based Web image search engine. However, because the search results rarely include its uncommon images, we can often get only its common images and cannot easily get exhaustive knowledge about its appearance (look and feel). As next steps of image searches in the Web, it is very important to discriminate between “Typical Images” and “Peculiar Images” in the approvable images, and moreover, to collect many different kinds of peculiar images as exhaustively as possible. This paper proposes a novel method to precisely retrieve peculiar images of a target object by its typical/peculiar appearance descriptions (e.g., color-names) extracted from the Web and/or its typical/peculiar image features (e.g., color-features) converted from them, as a solution to the 1st next step of image retrievals in the Web.

Query Modification Based on Real-World Contexts for Mobile and Ubiquitous Computing Environments

With the growing amount of information on the WWW and the improvement of mobile computing environments, mobile Web search engines will increase significance or more in the future. Because mobile devices have the restriction of output performance and we have little time for browsing information slowly while moving or doing some activities in the real world, it is necessary to refine the retrieval results in mobile computing environments better than in fixed ones. However, since a mobile user’s query is often shorter and more ambiguous than a fixed user’s query which is not enough to guess his/her information demand accurately, too many results might be retrieved by commonly used Web search engines. This paper proposes two novel methods for query modification based on real-world contexts of a mobile user, such as his/her geographic location and the objects surrounding him/her, aiming to enhance location-awareness, and moreover, context-awareness, to the existing location-free information retrieval systems.

Object-name search by visual appearance and spatio-temporal descriptions

When a user searches the Web for information about a target object by submitting a keyword-based query to such a conventional Web search engine as Google, the precision and recall of the search results depend a great deal on whether or not s/he has known exactly the concrete name of the target object. However, the user does not always know the concrete name of any target object that s/he has encountered in the real world and wanted information about. In this paper, we propose an application system of Object-Name Search that helps her/him to identify the concrete name of the target object by such ambiguous features as its class-name, visual appearance and spatio-temporal information. When the user inputs a class-name, visual appearance and/or real-world context descriptions, our system returns not only concrete object-names ranked by her/his specification but also their typical images, visual appearance and spatio-temporal descriptions. And then the user can also modify her/his original specification repeatedly by using their typical features as a useful reference.

ReCQ: real-world context-aware querying

This paper proposes a method of context-aware querying in mobile/ubiquitous Web searches, which provides mobile users with four capabilities: (1) context-aware keyphrase inference to help them input a keyphrase as a part of their keyword-based query, (2) context-aware subtopic tree generation to help them specify their information demand on one subtopic, (3) discovery of comparable keyphrases to their original query to help them make better decisions, and (4) meta vertical search focused on one subtopic to make the retrieval results more precise.