Showing papers by "Katsumi Tanaka published in 2020"

Search Support Tools

Abstract: This chapter presents in-depth reviews of search tools for supporting information search. With the brief introduction of cutting-edge search support tools, we describe the key ideas behind the tools and implications for design. We also discuss the limitations of conventional search interfaces to explore directions for future research on search support tools.

Context-Guided Learning to Rank Entities

Abstract: We propose a method for learning entity orders, for example, safety, popularity, and livability orders of countries. We train linear functions by using samples of ordered entities as training data, and attributes of entities as features. An example of such functions is f(Entity) \(= +0.5\) (Police budget) \(-0.8\) (Crime rate), for ordering countries in terms of safety. As the size of training data is typically small in this task, we propose a machine learning method referred to as context-guided learning (CGL) to overcome the over-fitting problem. Exploiting a large amount of contexts regarding relations between the labeling criteria (e.g. safety) and attributes, CGL guides learning in the correct direction by estimating a roughly appropriate weight for each attribute by the contexts. This idea was implemented by a regularization approach similar to support vector machines. Experiments were conducted with 158 kinds of orders in three datasets. The experimental results showed high effectiveness of the contextual guidance over existing ranking methods.

Searching for Microblogs Referring to Events by Deep Dynamic Query Strategies

Abstract: We address the problem of searching for microblogs referring to events, which are difficult to find because microblogs may refer to events without using event’s contents and a searcher may not use suitable queries for a search engine. We therefore propose a dynamic search process based on MDP that takes query strategies optimized for the current search state. As key components of the dynamic search process, we propose an RNN-based model for predicting long-term returns of a search process, and a DNN-based model that tries to match between the representations of microblogs and those of events for identifying relevant microblogs. Experimental results suggest that the dynamic search process could effectively search for microblogs, especially for implicitly referred events. Moreover, we show high applicability of our proposed approach to unseen events for which any relevant microblogs were not available in the training phase.