山本 岳洋

We investigate how explicit search roles assigned to group members affect their search performance and behavior in collaborative information seeking (CIS). Although several roles have been proposed in CIS, how these roles affect the search performances and behaviors of the members has not yet been explored. We focus on the existing Gatherer and Surveyor roles and analyze their effects on search performances and query formulation behaviors. The goal of our study is to understand the relationships between the roles and search behaviors and get insights into developing algorithms such as query suggestions or document rankings adaptive to the roles and behaviors. We conducted a user study with 20 participants in 10 pairs, where each pair of Gatherer and Surveyor were asked to perform a recall-oriented collaborative search task. We first analyzed the search performance of the two roles in terms of recall and diversity. We also analyzed how their queries were affected by their preceding queries or webpages that were visited through a questionnaire and log analysis. Finally, we discussed what algorithms would be required to support role-based CIS.

We address the problem of entity identification on a microblog with special attention to indirect reference cases in which entities are not referred to by their names. Most studies on identifying entities referred to by their full/partial name or abbreviation, while there are many indirectly mentioned entities in microblogs, which are difficult to identify in short text such as microblogs. We therefore tackled indirect reference cases by developing features that are particularly important for certain types of indirect references and modeling dependency among referred entities by a Conditional Random Field (CRF) model. In addition, we model non-sequential order dependency while keeping the inference tractable by dynamically building dependency among entities. The experimental results suggest that our features were effective for indirect references, and our CRF model with adaptive dependency was robust even when there were multiple mentions in a microblog and achieved the same high performance as that with the fully connected CRF model.

本稿では，検索タスク実行時のユーザの検索行動と，終了時のタスクに対する満足度との関係性を調査する．従来の情報検索に関する研究では，適合ページを多く提示することがユーザの満足度の向上につながるという前提の下で，検索結果のランキング手法が考案されてきた．しかし，それぞれの適合ページで記述されているタスクの答えが異なる場合は，かえってユーザの不満を引き起こす可能性がある．また，情報検索に対する専門性やタスクに関する事前知識といった属性の有無についても，ユーザの検索行動や満足度の評価基準に影響を与えることが予想される．そこで我々は，事実発見型タスクの検索ログに対して被験者が発見した答えを抽出することでデータセットを作成し，これらの2種類のユーザ属性が両者の関係性に与える影響を分析した．その結果，（1）情報検索の専門知識を持つユーザについては，発見された答えの一貫性と満足度との間に負の相関関係が存在する可能性がある，（2）情報検索の専門知識を持つユーザは，答えの発見以後も長い時間をかけてタスクに取り組む，および（3）情報検索の専門知識を持たないユーザは，タスク開始から一定時間が経過した後も，特定の答えに絞り込んだ検索を行わない，という傾向が見られた．In this paper, we investigate the relationship between user behavior observed in search tasks and satisfaction perceived by users in the tasks. Various kinds of methods for ranking search results have been proposed by existing work on information retrieval under the assumption that providing many relevant pages would lead to user satisfaction. As for search tasks where inconsistent answers are found, however, users may feel dissatisfied about the information obtained from the relevant results. As well as the type of search tasks, user attributes such as expertise in information retrieval and prior knowledge on the task could affect search behavior of users and their satisfaction perception. To analyze the effect of the two attributes on the relationship, we extracted answers from each page in search logs of fact-finding tasks. As a result of analysis of this dataset, we found the different tendencies in accordance with the presence or absence of these attributes: (1) finding inconsistent answers may cause dissatisfaction of search experts, (2) search experts still continue to search after finding some answer candidates, and (3) users without search expertise try to search for any answers even in the closing stage of search sessions.

This study investigated query formulations by users with Cognitive Search Intents (CSI), which are needs for the cognitive characteristics of documents to be retrieved, e.g. comprehensibility, subjectivity, and concreteness. We proposed an example-based method of specifying search intents to observe unbiased query formulations. Our user study revealed that about half our subjects did not input any keywords representing CSIs, even though they were conscious of given CSIs.

This study investigated query formulations by users with Cognitive Search Intents (CSIs), which are users' needs for the cognitive characteristics of documents to be retrieved, e.g. comprehensibility, subjectivity, and concreteness. Our four main contributions are summarized as follows: (i) we proposed an example-based method of specifying search intents to observe query formulations by users without biasing them by presenting a verbalized task description (ii) we conducted a questionnaire-based user study and found that about half our subjects did not input any keywords representing CSIs, even though they were conscious of CSIs (iii) our user study also revealed that over 50% of subjects occasionally had experiences with searches with CSIs, while our evaluations demonstrated that the performance of a current Web search engine was much lower when we not only considered users' topical search intents but also CSIs and (iv) we demonstrated that a machine-learning-based query expansion could improve the performances for some types of CSIs. Our findings suggest users over-adapt to current Web search engines, and create opportunities to estimate CSIs with non-verbal user input. Copyright 2014 ACM.

"Image search" on the basis of social tags is now a popular tool provided by image-sharing services. Some images are annotated with similar tags, while others are annotated with dissimilar ones. In this study, a concept called "tag-diversity," which represents how diverse tags are annotated to an image, is proposed, and two methods to estimate it are proposed. We conducted the experiment to investigate how the two proposed methods accurately compute tag-diversity. The results of the experiment show that both methods outperformed the baseline method, which calculates tag-diversity on the basis of the number of annotated tags. We also show some images with low and high tag-diversity, and discuss how tag-diversity can improve the current image search.

This paper proposes a query auto-completion interface that displays additional information on query suggestions and their search results along with the original query suggestions. Users of conventional query auto-completion interfaces cannot determine whether they can obtain relevant information from the search results of suggested queries until they actually issue them and check their search results. To address this problem, we propose a query auto-completion interface that incorporates useful information, such as words in the search results and images of the suggested queries, into the original query suggestions.

本稿では，動画共有ウェブサイトにおいて日々アップロードされる膨大な楽曲動画について，ユーザがその動画に対して興味があるかどうかを短時間で判断する手段として，15秒のサムネイル動画を自動生成する手法を提案する．ここでは，楽曲動画作成者と楽曲動画視聴者に注目し，楽曲動画のサビ検出技術と，視聴者の盛り上がり検出技術を使うことにより，サムネイル動画を自動生成する仕組みを実現する．また，評価実験により，組合せ手法の有効性とその特徴を明らかにする．The number of uploading video clips to video sharing Web sites has been explosively increasing. Then, it is not easy for users to find their preferable video clip. In this paper, we propose several methods to generate a fifteen-second thumbnail video clip from an original video clip based on analysis of viewer's responses and analysis of audio features. We clear the advantage and characteristics of the combination method with analysis of viewer's responses and analysis of audio features based on the evaluation test.

本稿では，印象に基づく楽曲検索実現のために，動画共有サイト上に投稿された楽曲動画を，可愛らしい，切ない，元気がでるといった印象に分類する手法を提案する．楽曲動画の印象分類のため，ユーザの投稿した時刻同期コメントに着目し，単語の品詞，文字の繰返し構造，楽曲のサビ区間の3つを利用する．実験では1,314本の楽曲動画を7印象クラスに分類し，提案手法がF値のマクロ平均で0.659を達成しベースライン手法よりも高い精度を得た．また，楽曲の歌詞や音響特徴量を用いた分類手法とも比較し，提案手法の有効性を示した．This paper proposes a method to classify music video clips, which are uploaded to the video sharing service, into the mood categories such as "cute," "sorrow" and "cheerful." The method leverages viewers' time-synchronized comments posted to video clips to classify the video clips into moods. It extracts features from the comments in the terms of (1) parts-of-speech, (2) lengthened words and (3) chorus parts of the music. Our experimental results showed that out method achieved the best classification performance (Macro F-measure of 0.659) compared with some baselines. In addition, our method outperformed the conventional approaches that utilize lyrics and audio features of musics.

Web上の情報量の増加にともない，Web検索エンジンを利用するユーザの意図は多様化している．本稿では，こうした多様な検索意図を，ユーザの視線情報を利用することで，検索時にリアルタイムに推定する手法を提案する．我々は，Webページ中でユーザが実際に注目している対象に着目し，「注目度の高い単語ほど，ユーザの検索意図に適合している」という仮定を置くことで，検索意図のリアルタイム推定に取り組む．本稿では，この仮定に基づき4種類の検索意図推定手法を提案し，ユーザ実験によって有効な推定手法の評価および考察を行った．さらに実験結果の分析から，検索意図推定における視線情報の有用性についても確認することができた．Search intents of Web search engine users become more diversified along with the rapid growth of information on the Web. This paper addresses the problem of estimating such diversified intents of Web search users from their search behaviors in real-time. In estimating searcher intents, we focus on their eye movements on the browsed Web pages, and assume that "terms that draw a high degree of searcher's attention are probably relevant to his/her search intent". Based on this assumption, we proposed four types of search intent estimation methods, and evaluated the estimation accuracy in each method through the experiment. In addition, analysis of experimental result reveals that by using eye movements data we can estimate the unique search intent of each user even if they perform the same search tasks.

旅行に関する情報をWeb検索エンジンを通じて収集する場合，観光地や宿泊先などさまざまな観点から検索を行う必要がある．このように，いくつかのサブタスクから構成される検索タスクでは，1つの検索クエリのみですべての情報を得ることは難しく，クエリの修正をともなう検索が反復的に行われることによって，タスクが実行されることが多い．本稿では，こうした検索クエリの修正を，事前に予測する手法を提案する．提案手法は，現在の検索行動を特徴量とする分類器を構成することで，次の検索におけるクエリ修正タイプの予測を行う．実際の検索行動のログデータを用いた評価実験の結果，提案手法は5種類の検索クエリ修正タイプを約41%の精度で分類可能なことが明らかになった．また，分類器構成に用いる特徴量を変化させることで，修正タイプの予想に有用な特徴量の検証も行った．さらに今後の展望として，次の検索において実際に入力されるクエリの予想可能性についても考察する．In search tasks composed of multiple sub-tasks (e.g., trip planning task), it is difficult for searchers to obtain information that satisfies their information needs completely with a single search query. In these tasks, they usually have to search iteratively by reformulating the query. This paper addresses the problem of predicting searchers' query reformulations beforehand. The proposed method tries to predict which category the following reformulation belongs to by constructing a classifier from search behavior data. As a result of evaluation based on the real search log data, we found that our method can predict query reformulation types with about 41% accuracy. We also analyze which and to what extent the user's behavior data is useful for predicting query reformulations. In addition to that, we discuss the predictability of the query itself issued in the following the search as the next step of this work.

The NTCIR INTENT task comprises two subtasks: Subtopic Mining, where systems are required to return a ranked list of subtopic strings for each given query and Document Ranking, where systems are required to return a diversified web search result for each given query. This paper summarises the novel features of the Second INTENT task at NTCIR-10 and its main findings, and poses some questions for future diversified search evaluation. Copyright © 2013 ACM.

The One Click Access Task (1CLICK) of NTCIR requires systems to return a concise multi-document summary of web pages in response to a query which is assumed to have been submitted in a mobile context. Systems are evaluated based on information units (or iUnits), and are required to present important pieces of information first and to minimise the amount of text the user has to read. Using the official Japanese results of the second round of the 1CLICK task from NTCIR-10, we discuss our task setting and evaluation framework. Our analyses show that: (1) Simple baseline methods that leverage search engine snippets or Wikipedia are effective for "lookup" type queries but not necessarily for other query types (2) There is still a substantial gap between manual and automatic runs and (3) Our evaluation metrics are relatively robust to the incompleteness of iUnits. Copyright © 2013 ACM.

This paper proposes a method to discover how a user's search intent changes using his/her behavior during a Web search. A Web search user has a particular search intent and formulates search queries according to that intent. It is, however, a difficult task for the user to formulate a optimal query, a single query able to find documents which completely satisfy his/her information need, by himself. After issuing the initial query, the user usually examines the search results, and modifies his/her initial query. The subsequent queries may be a query of Specialization type, Parallel Move type and so on. By recording these subsequent queries and the corresponding user behavior (including eye-gazing behavior), the present work tries to find the relationship between the user's query reformulation and user's behavior. The proposed method constructs a SVM classifier from the behavior log data obtained from the search and browsing processes. Our experimental results show that the proposed method can classify the next query reformulation into five categories using only the current search behavior data with about 41 % accuracy, greater than the baseline methods. We also analyze which and to what extent the user's behavior data is useful for predicting query reformulations. Copyright 2013 ACM.

This short paper proposes a method to classify music video clips uploaded to a video sharing service into music mood categories such as "cheerful," "wistful," and "aggressive." The method leverages viewer comments posted to the music video clips for the music mood classification. It extracts specific features from the comments: (1) adjectives in comments, (2) lengthened words in comments, and (3) comments in chorus sections. Our experimental results classifying 695 video clips into six mood categories showed that our method outperformed the baseline in terms of macro and micro averaged F-measures. In addition, our method outperformed the existing approaches that utilize lyrics and audio signals of songs. Copyright © 2013 ACM.

Building test collections based on nuggets is useful evaluating systems that return documents, answers, or summaries. However, nugget construction requires a lot of manual work and is not feasible for large query sets. Towards an efficient and scalable nugget-based evaluation, we study the applicability of semi-automatic nugget extraction in the context of the ongoing NTCIR One Click Access (1CLICK) task. We compare manually-extracted and semi-automatically- extracted Japanese nuggets to demonstrate the coverage and efficiency of the semi-automatic nugget extraction. Our findings suggest that the manual nugget extraction can be replaced with a direct adaptation of the English semi-automatic nugget extraction system, especially for queries for which the user desires broad answers from free-form text. Copyright © 2013 ACM.

This paper tackles the problem of mining subgoals of a given search goal from data. For example, when a searcher wants to travel to London, she may need to accomplish several subtasks such as "book flights," "book a hotel," "find good restaurants" and "decide which sightseeing spots to visit." As another example, if a searcher wants to lose weight, there may exist several alternative solutions such as "do physical exercise," "take diet pills," and "control calorie intake." In this paper, we refer to such subtasks or solutions as subgoals, and propose to utilize sponsored search data for finding subgoals of a given query by means of query clustering. Advertisements (ads) reflect advertisers' tremendous efforts in trying to match a given query with implicit user needs. Moreover, ads are usually associated with a particular action or transaction. We therefore hypothesized that they are useful for subgoal mining. To our knowledge, our work is the first to use sponsored search data for this purpose. Our experimental results show that sponsored search data is a good resource for obtaining related queries and for identifying subgoals via query clustering. In particular, our method that combines ad impressions from sponsored search data and query co-occurrences from session data outperforms a state-of-the-art query clustering method that relies on document clicks rather than ad impressions in terms of purity, NMI, Rand Index, F1-measure and subgoal recall. © 2012 ACM.

In this paper, we propose a two-stage system using user's eye movements to accommodate the increasing demands to obtain information from the Web in an efficient way. In the first stage the system estimates a user's search intent as a set of weighted terms extracted based on the user's eye movements while browsing Web pages. Then in the second stage, the system shows relevant information to the user by using the estimated intent for re-ranking search results, suggesting intent-based queries, and emphasizing relevant parts of Web pages. The system aims to help users to efficiently obtain what they need by repeating these steps throughout the information seeking process. We proposed four types of search intent estimation methods (MLT, nMLT, DLT and nDLT) considering the relationship among intents, term frequencies and eye movements. As a result of an experiment designed for evaluating the accuracy of each method with a prototype system, we confirmed that the nMLT method works best. In addition, by analyzing the extracted intent terms for eight subjects in the experiment, we found that the system could estimate the unique search intent of each user even if they performed the same search tasks.

A diversified search result for an underspecified query generally contains web pages in which there are answers that are relevant to different aspects of the query. In order to help the user locate such relevant answers, we propose a simple extension to the standard Search Engine Result Page (SERP) interface, called AspecTiles. In addition to presenting a ranked list of URLs with their titles and snippets, AspecTiles visualizes the relevance degree of a document to each aspect by means of colored squares ("tiles"). To compare AspecTiles with the standard SERP interface in terms of usefulness, we conducted a user study involving 30 search tasks designed based on the TREC web diversity task topics as well as 32 participants. Our results show that AspecTiles has some advantages in terms of search performance, user behavior, and user satisfaction. First, AspecTiles enables the user to gather relevant information significantly more efficiently than the standard SERP interface for tasks where the user considers several different aspects of the query to be important at the same time (multi-aspect tasks). Second, AspecTiles affects the user's information seeking behavior: with this interface, we observed significantly fewer query reformulations, shorter queries and deeper examinations of ranked lists in multi-aspect tasks. Third, participants of our user study found AspecTiles significantly more useful for finding relevant information and easy to use than the standard SERP interface. These results suggest that simple interfaces like AspecTiles can enhance the search performance and search experience of the user when their queries are underspecified. © 2012 ACM.

本研究では，比較文のマイニングによってエンティティ間の順序関係を明らかにし，ある観点で見た場合に2つのものの間にあてはまるエンティティ（補間エンティティ）およびその系列の発見をする手法に取り組む．本稿では特に，"この店よりはおいしくて，あの店よりはリーズナブルな店を見つけたい"，"2つの本の間にあてはまるような難しさを持つ本を見つけたい"，といった，主観的評価における補間エンティティを発見する手法を提案する．提案手法では，検索エンジンを用いて比較文を発見，集約し，（評価対象，比較対象，評価，極性）の組を抽出し，それに基づきグラフを生成する．そして，得られたグラフより補間エンティティおよびその系列を発見する．本稿ではグラフから補間エンティティおよびその系列を発見する手法として3つの手法を提案し，評価を行った．We propose a method of detecting intermediate entities or sequences between two examples by detecting order of entities based on comparative sentences. We focus on finding intermediate entities based on subjective evaluations. For example, "I want to find a restaurant that is better than this restaurant but is cheaper than that one." and "I want to find a book that is intermediate level between two books." The main idea of our proposed method is collecting comparative sentences with entities. First, it collects comparative sentences with particular expressions and extracts comparative relations from them and generates a directed graph of their relations. Finally, it finds an optimal path from the graph. We proposed three methods to rank nodes and paths and evaluated their effectiveness by conducting the experiments.

映像の編集や検索を行う際に，ある人物がより視聴者の注目を集めている映像を検索したいということはよくある．また，現在視聴している映像の関連映像として，登場人物の活躍パターンが似ている映像を推薦してほしいということもよくある．しかし，映像に登場する人物がどの程度視聴者の注目を集めているかは，映像のテキストデータや画像解析だけでは判断できないことが多く，視聴者の反応に基づいて注目されている人物を推定する必要がある．そこで本稿では視聴者の反応として映像の再生時刻に沿って付与されたコメントを用いて，映像に登場する人物が視聴者の注目を集めているシーンの推定と各シーンにおける各登場人物の活躍の度合いの推定を行う手法の提案をする．ただし，本稿ではニコニコ動画に投稿された映像の中で，コメントが一定数以上付与された映像に登場する，視聴者に名前が広く知れわたっている人物を対象とする．When a user edits or searches a video clip, he/she often hopes to extract a scene or search video clips in which a character is active. Moreover, he/she also often hopes to be recommended a video clip that has a similar activity pattern of characters as a relevant video clip that he/she is watching. However, it is difficult in many cases to judge the active characters from only text data or image analysis. We estimate the degree of attention based on viewers' reaction. In this paper we use comments posted to a video clip as the viewers' reaction. We propose a method to estimate the spotlighted scenes of each character in a video clip and the degree of it. We especially target video clips that is uploaded to Nico Nico Douga and a character whose name is known widely.

我々は膨大な検索結果をさまざまな観点から閲覧し，効率良く多くの方法を閲覧するための仕組みとして，検索結果全体の俯瞰的インタフェース（タームクラウド）と語ベース適合フィードバックに基づく検索結果の再ランキングシステムを提案してきた．提案してきた俯瞰的インタフェースは有効なものであったが，提示する単語は単純な語の出現頻度に基づき選ばれていたため，ユーザの再ランキング行為を促進するには不十分であった．そこで本稿では，よりユーザの興味を引く単語を提示するため，"有名な観光地"や"美味しい和菓子屋"といった観点に着目し，そうした観点をQ&Aコンテンツから抽出する手法を提案する．ユーザ実験の結果，提案手法により得られた観点は，従来のクエリ推薦や頻度に基づく単語抽出手法に基づく手法に比べてユーザの興味を引く単語を多く提示可能であることが分かった．We have previously proposed a system that reranks Web search results based on users' term-based feedback. The system enables users to explore search results from diverse viewpoints. We have also proposed an interface called TermCloud which are generated from frequent terms appear in Web search results. However, these terms are not enough to support users reranking operations. In this paper, we propose a method to extract aspects like "famous spots" or "delicious restaurant" by using Q&A contents in order to suggest users more interesting terms. The results from the user experiments revealed that our method could provide more interesting terms than several baseline methods including conventional query suggestions.

This paper proposes a system called "RerankEverything", which enables users to rerank search results in any search service, such as a Web search engine, an e-commerce site, a hotel reservation site, and so on. This system helps users explore diverse search results. In conventional search services, interactions between users and systems are quite limited and complicated. By using RerankEverything, users can interactively explore search results in accordance with their interests by reranking search results from various viewpoints. Experimental results show that our system potentially help users search more proactively. When using our system, users were more likely to click search results that were initially low ranked. Users also browsed through more diverse search results by reranking search results after giving various types of feedback with our system. © 2011 ACM.

In this paper, we propose a method for helping users explore information via Web searches by using a question and answer (Q&amp A) corpus archived in a community Q&amp A site. When users do not have clear information needs and have little knowledge about the task domain, it is difficult for them to create queries that adequately reflect their information needs. We focused on terms like "famous temples," "historical townscapes," and "delicious sweets," which we call "adjective facets", and developed a method of extracting these facets from question and answer archives at a community Q&amp A site. We evaluated the effectiveness of our adjective facets by comparing them with several baselines. © 2011 ACM.

本稿では，さまざまなランキング結果をユーザのインタラクションに応じて自由に再ランキングできるシステムRerankEverythingを提案する．既存の検索エンジンやウェブサービスでは，サービス側が用意した機能を利用することでしか，ランキング結果を変更することができない．RerankEverythingでは，ユーザはランキング結果に対して「この単語を含む結果を上位・下位に再ランキングしたい」であるとか，「この尺度でランキング結果を並べ替えたい」といった意図を簡単なインタラクションを介して伝えることができ，自らの興味や嗜好に合わせてランキング結果を再ランキングすることができる．また，本システムを利用することで，ユーザはウェブ検索結果だけでなく，動画・商品・ホテル・ニュースなどさまざまなランキング結果を再ランキングできるようになる．本稿ではシステムを（1）直接操作による再ランキングインタフェース，（2）タームクラウドによるランキング結果閲覧支援，（3）パーサ作成インタフェースと属性値の動的抽出，という3つの観点から設計し，ユーザ実験によりシステムの有用性を確認した．In this paper we propose a system called RerankEverything, which enables users to rerank search results in any search service. In conventional search services, interactions between users and systems are quite limited and complicated. In addition, search functions and interactions to refine search results differ depending on the services. By using RerankEverything, users can interactively explore search results in accordance with their interests by reranking search results from various viewpoints. The reranking interaction in our system is simple and unified independently of the various search services. To realize our system, we design and implement the system from three aspects: (1) a reranking interface with direct interaction, (2) a TermCloud to encourage users to browse and rerank search results from various viewpoints, and (3) an interactive wrapper generation interface and a dynamic attribute extraction method. We evaluated the usefulness of our system by conducting the user studies.

This paper proposes a system called RerankEverything, which enables users to rerank search results in any search service, such as a Web search engine, an e-commerce site, a hotel reservation site and so on. In conventional search services, interactions between users and services are quite limited and complicated. In addition, search functions and interactions to refine search results differ depending on the services. By using RerankEverything, users can interactively explore search results in accordance with their interests by reranking search results from various viewpoints. © 2010 Copyright is held by the author/owner(s).

Recently, users can find various kinds of information in the Web. When a user browses information, he/she sometimes wants to browse more desirable information by adding/deleting few more. However, there is no service to browse such desirable information from current information. We focused on such users' browse/search intention. Here, each information consists of some elements such as ingredients, persons, and places. We call the information "collective Web object". In this work, we propose a method to enable users to browse from current collective Web object to desirable collective Web object by adding one element into it or deleting one element from it. In addition, we introduce the concept of structural stability of collective Web object based on constructing elements and apply our method to recipe search. We implemented a prototype system and performed experiments to evaluate the usefulness and the applicability of our method.

We propose a system for finding intermediate entities from two examples by using web search engine indices. For example, a user wants to find recipients of the Nobel Peace Prize in the thirty years between Mother Teresa in 1979 and Barack Obama in 2009. In this example, the answer is, for example, Kofi Atta Annan. In this situation, the user wants to find something intermediate between two entities. We first describe the problem of finding entities between two examples. We then propose a system for extracting intermediate entities between two inputs by using a Web search engine indices. The system focuses on the positions of terms in Web pages and then extracts candidate terms that are likely to appear between the two inputs. Then, our system ranks candidate terms based on term frequencies and positions. Finally, we conducted experiments to show the usefulness of our system. © 2010 ACM.

We previously proposed a reranking system for Web searches based on user interaction. The system encouraged users to interact with terms in search results or with frequent terms displayed in the tagcloud visualization style. Over 20,000 interaction logs of users were analyzed, and the results showed that more than 70% of users had interacted with the terms in the tagcloud area. Therefore, this visualization style is thought to have great potential in supporting users in their Web searches. This visualization style is referred to as TermCloud in this paper. We describe how TermCloud can increase the effectiveness of users' Web searches, and we propose a technique to generate a more useful TermCloud than the frequency-based TermCloud. Then, we show the usefulness of our method based on the experimental test.

Search engines return a huge number of Web search results, and the user usually checks merely the top 5 or 10 results. However, the user sometimes Must collect information exhaustively Such as collecting all the publications which a certain person had written, or gathering a lot of useful information which assists the user to buy. In this case, the user must repeatedly check search results that are clearly irrelevant. We believe that people Would use a search system which provides the reranking or classifying functions by the user's interaction. We have already proposed a reranking system based oil the user's edit-and-propagate operations. In this paper, we introduce the drag-and-drop operation into our system to support the user's exhaustive search.

We propose a method for finding an intermediate entity between two examples on the Web. For example, a user wants to find events that occurred between the Battle of Red Cliffs and the death of Cao Cao. In this situation, the user wants to find something intermediate between two events, processes, or objects. We first describe the problem of finding an entity between two examples. We then propose a method for extracting an intermediate entity between two inputs using a Web search engine. The method focuses on the positions of words in Web pages and then extracts words that are likely to appear between the two inputs. Finally, we show the usefulness of our method based on experiments. Copyright 2009 ACM.

本研究では，ウェブ検索においてユーザが編集操作を用いることで検索結果を動的にリランキングする手法を提案する．ユーザの検索意図には様々な種類が存在し，検索エンジンが返す結果は必ずしもユーザの意図を反映したものではない．このような場合，ユーザは検索結果を下位まで1 つずつチェックするか，新しいクエリで再検索を行う必要がある．本研究の目的は，ユーザとシステムが編集操作を通して対話を行うことを可能とすることによって，ユーザの検索意図を反映することである.本稿では，検索結果に対するユーザの削除操作や強調操作によって，検索結果をリランキングする手法を提案した．また，システムを実装し，システムの有用性を検証した．This paper proposes a method of a dynamically reranking Web search results according to the user's editing operations while browsing the search results. Because users' Web search intentions are diverse, search engines cannnot always return search results that satisfy the user's search intention adequately. Hence, the user must check search results sequentially, or re-search using a new query. Our goal is to reflect the users' search intentions by editing search results through user interaction and propagating the intention of the editing operations. In this paper, we propose a method of reranking web search results depends on users' deletion operations and emphasis operations. We describe the implementation of our system and show experimental results.

This paper proposes novel interaction techniques for parallel search tasks. The system displays multiple search results, returned by a search engine, side-by-side for each search query. The system enables the user to rerank search results using a reranking algorithm based on vertical and horizontal propagation of his/her intention. A method of recommending operations for specific keywords is also proposed, supporting operations such as a shift to a parallel search with an alternate term, upgrading or downgrading results in terms of a specific viewpoint, and so on. Applications for the proposed system are also discussed.

This paper proposes a novel technique of browsing Web pages called "Edit-and-Propagate" operation based browsing, where edit operation is regarded as the third operation for interacting with the WWW after conventional clicking and scrolling towards realizing the Editable Web Browser. Our method enables users to delete/emphasize any portion of a browsed Web page at any time and modifies the page by propagating the edit operation. For example, the user can easily delete almost any uninteresting portion of a Web page merely by deleting an example of an unwanted portion. While browsing a Web search result page, the user can rerank search results by deleting an unwanted term or by emphasizing an important term. In this paper, we describe the concept of "Edit-and-Propagate" based browsing, and the implementation of our prototypes. Then we describe the results of our evaluation, which demonstrate the usefulness of our system. Copyright 2007 ACM.