Faculty Profiles - YOSHINO Takashi

Published Papers - YOSHINO Takashi

Division display21 - 40 of about 88／ All the affair displays >>

Multilingual Parallel Texts Creation Method Using Accurate Texts Created by Monolingual Users and Perforated Texts

福島拓, 吉野孝（Part： Last author )

情報処理学会論文誌 60 ( 1 ) 166 - 173 2019.01 [Refereed]

Parallel text is often used in fields where accurate multilingual communication support is required. It is a multilingual corpus that can perform accurate multilingual conversion. The volume of parallel texts available has been inadequate because they were all created manually. In this paper, we propose and evaluate a multilingual parallel texts creation method using a perforated text whereby words in the texts can be interchanged. This method uses accurate texts created in each language by monolingual users and multilingual pairs of perforated texts. The major contributions of this study are as followings. (1) We proposed and developed a multilingual parallel texts creation method using perforated texts. (2) The proposed method can create accurate parallel texts by using perforated texts and filling words being guaranteed accuracy.

Evaluation of a Cooperative Patient Information System for Home Care Providers

山本理絵, 吉野孝, 西端めぐみ, 中井國雄, 柳本将喜, 入江真行

情報処理学会論文誌 59 ( 5 ) 1351 - 1362 2018.05 [Refereed]

It is important to cooperate and share information among multi-professional healthcare providers involved in home medical care. However, many of the regional medical cooperation systems built in Japan focus on "one-way" information collaboration. Therefore, we explore two research subjects: "sharing of information in both directions" and "simplification of input forms." We developed a patient information sharing system for home healthcare cooperation focusing on communication support. Moreover, we analyzed the usage of the system based on the system logs. The contributions of this study are as follows. (1) By using a separate system specializing in communication support, multi-professional healthcare providers are able to share information both ways for cooperative medical care. (2) In the system, a patient's medical record and home nursing record, and a video of a patient's home environment can be shared online. This study shows that multi-professional healthcare providers can share information with comparative ease using the proposed system.

Proposal of an Integrated Support System for Disaster-preparedness Map Making Through Town-walk Type of Gathering Information

榎田宗丈, 福島拓, 吉野孝, 杉本賢二, 江種伸之（Part： Corresponding author )

情報処理学会論文誌 59 ( 3 ) 992 - 1004 2018.03 [Refereed]

Throughout Japan, disaster-preparedness maps have been made by the method of town-walk. This method improves disaster-preparedness and allows for a greater understanding of the area. The making of disaster-preparedness maps increases the disaster awareness of the participants. Further, the involvement improves participants' ability of self-help and increases cooperation in local communities. Currently, there is a system to support the making of conventional disaster-preparedness maps, but no support system for maps made using the town-walk method. Therefore, we have proposed and developed an integrated support system for the making of disaster-preparedness maps by gathering information from the town-walk. We conducted a comparison experiment between the proposed system and the conventional system (using paper maps). This experiment demonstrated that a consistent support function would be effective in developing disaster-preparedness maps using the town-walk method. We also confirmed that the proposed system influences the improvement of disaster-preparedness awareness, the understanding of the area, and the conventional system.

流言に関する気づきを提供するシステムの構築

柿本大輔, 宮部真衣, 荒牧英治, 吉野孝

ヒューマンインタフェース学会論文誌 20 ( 1 ) 1 - 11 2018.02 [Refereed]

Proposal of Parallel Texts Creation Method Using Multiple Machine-translation by Monolingual on Crowdsourcing Workers

山本里美, 福島拓, 吉野孝（Part： Last author,　Corresponding author )

情報処理学会論文誌 59 ( 1 ) 76 - 85 2018.01 [Refereed]

Our former study showed that crowdsourcing workers, even they are monolingual speaker, can make correct parallel texts to request tasks. There is a difference in the translation accuracy of the machine translation according to the sentences of the translation target. It is difficult to create bilingual sentences for many sentences. In this paper, we propose a method to use a sentence adapted to machine translation and to use multiple machine translation engines for the conventional method. A sentence adapted to machine translation is a sentence that is almost the same as the original, but the appropriate machine translation is done. We aim to improve the accuracy of machine translation sentences in the sentences of a machine translation with a low accuracy. The contribution of this paper is as follows. We proposed a method to use a sentence adapted to machine translation and to use multiple machine translation engines in the conventional method for the conventional method. We also proposed a combination of evaluation by the person and the evaluation by the computer as a method of evaluation of a sentence adapted to machine translation. We have shown that the proposed method is capable of making the appropriate machine translation sentences in the original 39 of 40 sentences (97%) with of 25 points higher than the conventional method.

Correctness Evaluation Method for Multilingual Parallel Texts Using Ratio of Working Times of Workers via Crowdsourcing

福島拓, 吉野孝（Part： Last author )

情報処理学会論文誌 59 ( 1 ) 69 - 75 2018.01 [Refereed]

Generally, multilingual support systems for applications requiring high accuracy use multilingual texts pairs called "parallel texts." Accuracy evaluation is important for providing parallel texts. However, because there are huge quantities of parallel texts, it is difficult to thoroughly collect accuracy evaluations. In this paper, we propose a correctness evaluation method for parallel text-candidates using crowdsourcing. This method ranks workers via crowdsourcing based on their ratio of work time. This idea aims to preferentially select workers who evaluate accurately. The major contribution of this study is a correctness evaluation method of parallel text candidates, having a higher accuracy than that of existing methods. Furthermore, the proposed method ranks workers based on their ratio of working time.

Extraction Method of Impressions about Tourist Spots from Tweets without Tourist Spot Names

渡邉小百合, 吉野孝（Part： Last author,　Corresponding author )

情報処理学会論文誌 59 ( 1 ) 43 - 51 2018.01 [Refereed]

New needs for tourist spots may arise because the appearance of new forms of travel in Japan, in which fans visit the locations of dramas and animated programs, and the number of foreign travelers to Japan has increased in recent years. Therefore, tourist spot operators want to know the needs of tourists and any problems with the tourist spot so that they can be solved. However, insufficient tourist information was provided by Twitter. In this study, we propose an extraction method of impressions about tourist spots to apply to tweets that do not contain tourist spot names. The experiments provided the following four results: (1) Tweets containing feature words contain tourist information. (2) Tweets without tourist spot names tweeted before and after tweets containing tourist spot names contain tourist information. (3) Replies to tweets containing tourist spot names contain tourist information. (4) Tweets with images that do not contain tourist spot names contain information regarding the food and entertainment available at tourist spots.

Construction of Information-credibility Verification-behavior Facilitation System for Preventing False Rumors Spreading

柿本大輔, 宮部真衣, 荒牧英治, 吉野孝

ヒューマンインタフェース学会論文誌(CD-ROM) 20 ( 1-4 ) 2018

Construction of Information-credibility Verication-behavior Facilitation System for Preventing False Rumors Spreading

Kakimoto Daisuke, Miyabe Mai, Aramaki Eiji, Yoshino Takashi

The Transactions of Human Interface Society ( Human Interface Society ) 20 ( 1 ) 1 - 12 2018 [Refereed]

As the widespread of microblogs, we are now capable of obtaining and conveying various information. Microblogs can send out information easily and in real-time, on the other hand, they may widely spread uncertain and false rumors. One of the factors that false rumors are spread is the lack of confirmation about the credibility of the information before retweeting. In this study, we have developed a system to make aware of existence of rumor in web browsing. We found that the proposed system can make aware of existence of rumor and does not disturb web browsing. This system can provide the awareness about rumors to users in web browsing, and can prevent spreading uncertain rumors.

DOI

The Expression of 3-D Movement which Can Give a Feeling of Intrusion in Partial Overlay-type Video Chat System

濱上宏樹, 吉野孝（Part： Last author,　Corresponding author )

情報処理学会論文誌 58 ( 1 ) 189 - 196 2017.01 [Refereed]

近年，Skypeなどの無料でビデオチャットを行えるツールの普及により，容易にビデオチャットを行える環境が整ってきた．そのため，今後は会議などのフォーマルな場面での利用だけでなく，ビデオチャットは，家族や友人とのインフォーマルなコミュニケーションでの利用が増えてくると考えられる．しかし，従来のビデオチャットには，対面環境に比べて対話相手との心理的な距離感を感じてしまうという問題点がある．本研究では，この問題点を改善するため，部分重畳型ビデオチャットシステム「ドアコムZ」の開発を行った．ドアコムZは，現実に存在する枠を介して遠隔地間を仮想的につなぎ，相手の空間とつながっているような表示を行うビデオチャットシステムである．深度情報を用いることで，遠隔空間における三次元的な移動を表現した．遠隔空間内における三次元的な移動の実現により，「遠隔空間において相手の背後に隠れる」「遠隔空間内を指差す」という表現が可能になった．遠隔空間における三次元的な移動の表現の有用性を示すために，侵入感の検証および被侵入感の検証を行った．検証実験より，三次元的な移動の表現は，二次元的な移動の表現に比べ，侵入感を向上させることが分かった．また，両手法ともに被侵入感が得られ，表現による大きな違いはないことが分かった．表現方法に関係なく，自身の空間にある物を指差される映像は，被侵入感を与えることが分かった．In recent years, the spread of tools that free video chat, such as Skype, has facilitate the use of video chat. We think that a video chat can be used in informal communication. However, the conventional video chat has a problem that you feel a sense of distance to a dialogue partner compared with the face-to-face environment. In order to improve the problem, we have developed a partial overlay-type video chat system "DOACOM-Z." DOACOM-Z can connect between remote places virtually through the frame which exists in an actual space. The system can express three-dimensional movement in the remote space using a depth sensor at both places. The system can give both expressions "hiding in the remote space" and "pointing in the remote space." The usefulness of the expression of three-dimensional movement in the remote space, we carried out a verification experiment of a feeling of intrusion and a verification experiment of a feeling of being intruded. From the result of the verification experiments, we found that the expression of three-dimensional movement can improve the feeling of intrusion compared with the expression of the two dimensional movement. In addition, a feeling of intrusion was obtained with both methods. We found that there is no significant differences between expressions. Regardless of the method, the image pointing object in its own space can provide a feeling of being intruded.

Making of Disaster-prevention Maps by Local Residents Using the AkariMap Evacuation Support System

吉野孝, 濵村朱里, 福島拓, 江種伸之（Part： Lead author,　Corresponding author )

情報処理学会論文誌 58 ( 1 ) 215 - 224 2017.01 [Refereed]

多くの市町村がハザードマップや防災マップを発行しているが，保管率や利用率は低い．そのため，電子媒体を利用したハザードマップや防災マップを提供および作成を支援するシステムが多く存在する．また防災マップの作成の過程は，地域の人同士の交流や連携を支援するために利用できるといわれている．そこで，災害発生前から利用可能なオフライン対応型災害時避難支援システム「あかりマップ」を利用し，地域の防災マップを作成する実験を行い，情報通信技術を用いた防災マップ作成における利用可能性や問題点などを考察した．本研究の貢献は以下の3点にまとめられる．(1)地域住民によって情報通信技術を用いて防災マップを作る際に，密集した場所への情報の登録方法や情報の提示方法の工夫が必要である．(2)ゲーミフィケーションを用いて単に情報の登録数を競わせるだけでなく，情報の正確性の確保への貢献も可視化する仕組みが必要である．(3)地域住民による情報通信技術を用いた防災マップの作成のために，情報通信技術に慣れた学生と一緒に防災マップの作成を実施する手法が有効だった．Many municipalities publish hazard and disaster-prevention maps, but the keeping rate and the utilization rate are low. Thus, various electronic media can create and distribute hazard and disaster-prevention maps. The process of making disaster maps is claimed to support exchange and cooperation between people in an area. We have developed an evacuation support system called AkariMap that can be used offline in the aftermath of a disaster. We use AkariMap to make disaster-prevention maps of an area, and we discuss its applicability and the problems of using information and communications technology (ICT). The contributions of this study are as follows. 1) To have residents make a disaster-prevention map by using ICT, we need a device that can register and present information about a congested area. 2) We created a method of visualizing the accuracy of the information, and of encouraging the registration of information via a gamified competition. 3) We showed that the method of making a disaster-prevention map was as effective when employed by residents using ICT as it was when used by a student accustomed to ICT.

日常利用可能なオフライン対応型災害時避難支援システム “あかりマップ” の実環境における利用可能性

濵村朱里, 福島拓, 吉野孝, 江種伸之（Part： Corresponding author )

情報処理学会論文誌 57 ( 1 ) 319 - 330 2016.01 [Refereed]

Proposal of a Pair of Dialogic Parallel Texts Creation Method Using Conversation Sentences on Crowdsourcing

山本里美, 福島拓, 吉野孝（Part： Last author,　Corresponding author )

IPSJ Journal ( Information Processing Society of Japan (IPSJ) ) 56 ( 3 ) 1080 - 1089 2015.03 [Refereed]

The opportunities for communication amongst people whose native languages differ are increasing by globalization. However, it is difficult to share exact information among many languages. A parallel text that combines example sentences and their accurate translation are used in a multilingual communication support. There are many required parallel texts. Moreover, it is not easy to collect a sufficient number of parallel texts. In this paper, we propose a pair of dialogic parallel text creation method of requesting correction for a machine translation sentence with low accuracy on crowdsourcing of a monolingual. The proposed method presents the conversation sentences with a machine translation sentence, in order that a worker may make the intention of example sentences easy to guess. The contribution of this paper is to clarify the effect of presentation of the conversation sentence in a pair of dialogic parallel text creation.

Evaluation of the Availability of an Evacuation Support System "AkariMap" before a Disaster for Use During Network Failure

濵村朱里, 福島拓, 吉野孝, 江種伸之（Part： Corresponding author )

IPSJ Journal ( Information Processing Society of Japan (IPSJ) ) 56 ( 1 ) 185 - 195 2015.01 [Refereed]

In the aftermath of the Great East Japan Earthquake, many research initiatives were launched in the fields of network and information technology in anticipation of other such major disasters in the future. However, a network cannot be accessed immediately after a disaster. Usually, few people can handle refuge-supporting information spontaneously at travel and business destinations. Furthermore, it is difficult to store refuge-supporting information pertaining to a wide area on a mobile phone. To address these critical issues, we have developed an evacuation support system named AkariMap, that can prompt the user to confirm information to be necessary after a disaster. AkariMap incorporates a notification function that notifies the user of refuge-supporting information at his/her present location by considering movement of the user. After the system notifies the user of refuge-supporting information around his/her location, the user can access AkariMap to obtain the information. Thereafter, the system stores the refuge-supporting information around the user in the mobile phone. The contributions of this study are as follows: (1) We have developed an evacuation support system with a notification function to support continuous use in before disaster. The notification function can prompt a user to confirm refuge-supporting information around the user. (2) Even if people use the notification function in the long term, the system triggers the user to use the system at the place where one is staying. Users can grasp easily refuge-supporting information in about 10 days.

Verification by Presentation of Contribution Information to Users of Motivation to Multilingual Parallel Text Registration Using the Notification Function on Mobile Phone

吉野孝, 西田祥子, 福島拓（Part： Lead author,　Corresponding author )

IPSJ Journal ( Information Processing Society of Japan (IPSJ) ) 56 ( 1 ) 118 - 125 2015.01 [Refereed]

Parallel text has been used for accurate multilingual communication and information sharing. The motivation of this research is the lack of parallel text. In recent times, the use of smartphones has increased, and the notification function has been widely employed in smartphones. Therefore, we have developed a registration system called Chocolan, with support for multilingual parallel text. Through experiment 1, we found that the notification function can motivate the users to register example sentences. Moreover, we found the following results from experiment 2: (1) Registration of example sentences is encouraged by each user's contribution information rather than each user's ranking information. (2) The contribution information on the contents of notification information may provide a key to consider a new example sentence.

Paired Dialogue Parallel Text Composition Model for Reliable and Unrestricted Multilingual Communication Support

福島拓, 吉野孝（Part： Last author )

IPSJ Journal ( Information Processing Society of Japan (IPSJ) ) 56 ( 1 ) 219 - 227 2015.01 [Refereed]

Recently, worldwide globalization has helped to improve communication among people with different native languages. However, sharing accurate information in many languages is not sufficient. In general, multilingual support systems for applications that require high accuracy use parallel texts, but the context may be incorrect in systems that use existing parallel text systems. Moreover, parallel texts are fixed phrases, i.e., a system can only select a perfectly matching sentence. In this study, we propose a paired dialogue parallel text composition model to resolve the existing problems with multilingual communication support. We also developed this system and conducted an experimental verification of the usefulness of the model. The main contributions of this study are as follows. (1) We propose a model that uses dialogue parallel texts to solve the problems of existing multilingual dialogue support system by using parallel texts. (2) This model collects dialogue parallel texts automatically. We also demonstrate that the model can support the selection of the most appropriate answers. (3) We show that the model can support unrestricted dialogue by connecting similar text with parallel dialogue texts.

Development and Evaluation of a Nutrition Management System for Elderly People with a Dish Registration Function

川島基子, 吉野孝, 紀平為子, 伊井みず穂, 岡本和士, 江上いすず, 藤原奈佳子, 石川豊美, 入江真行（Part： Corresponding author )

IPSJ Journal ( Information Processing Society of Japan (IPSJ) ) 56 ( 1 ) 171 - 184 2015.01 [Refereed]

A significant number of elderly people suffer from chronic diseases, such as diabetes and obesity. In general, the elderly are prescribed a medical diet in order to treat such illnesses. However, it is not easy to investigate their eating habits. To the best of our knowledge, the existing methods for monitoring the diet of the elderly are unsuitable. In this paper, we propose a nutrition management system that is easy to use for dietitians as well as the elderly. This system consists of a novel dish registration function that allows users to register various dishes. We evaluated our system by experimenting with 15 subjects with an average age of approximately 70 years. Based on the results of a questionnaire and our system's usage history, we could determine that 13 out of the 15 subjects were able to record their diet daily. Moreover, 11 subjects received reminders from our system to record their diet daily. We propose that the system can perform meal investigation and nutrition management for the elderly. During the experimental period, 314 out of 323 dish data were corrected when the elderly subjects entered new dishes in the system. The dietitians could view this data in the system. We show that our proposed system can provide useful feedback to elderly users.

Proposal of Cultural Difference Detection Method Using Japanese and Chinese Wikipedia

諏訪智大, 宮部真衣, 吉野孝（Part： Last author,　Corresponding author )

情報処理学会論文誌 55 ( 1 ) 257 - 266 2014.01 [Refereed]

多言語間コミュニケーションにおいて，同一の単語を用いて会話をしている場合でも，相手の文化について十分に理解していないために，誤解が生じる可能性がある．現在，文化差の有無の判断は，人が行う必要があるが，その判断には相手の文化に関する十分な知識が必要となるため，容易ではない．そのため，文化差が存在することを自動的に検出する仕組みが求められている．そこで本論文では，多言語知識のデータベースであるWikipediaを利用した文化差の検出手法を提案する．検討用データセットを分析した結果，文化差のある語句には，Wikipedia記事の本文やカテゴリに特徴が存在することが分かった．そこで，これらの特徴を考慮した手法を検討した．実験の結果，提案手法は，検討用データセットにおけるすべての文化差に関して，最も良い精度で検出することができた．また，「世界的な観点から説明されていない」「日本語版Wikipediaのみで日本について言及されている」「中国語版Wikipediaにおいて中国について偏った表記がされていない」「各言語版Wikipedia内に各国に対する言及表現が存在しない」の特徴を含む記事が，文化差判定において重要であることが分かった．There is a possibility that the misunderstanding is caused in multilingual communications, because people cannot understand enough other culture even when talking by using the same word. People should judge the presence of a cultural difference in today's situation. However, it is difficult to judge it because enough knowledge of other culture is necessary. We proposed a cultural difference detection method using Wikipedia which is a database of multilingual knowledge. From the analysis of data set for examination, We found the feature in the text and category, and others of the articles on Wikipedia. Therefore, We proposed method using these features. From the result of the experiment using data set for examination, we presented that the proposed method can detect correct differences about all kind of cultural differences. We found that the following features of an article are important for detecting cultural difference: (1) An article is not explained from a global viewpoint. (2) An article is mentioned about Japan by Japanese version Wikipedia. (3) An article of the Chinese version Wikipedia is not mentioned which inclined toward China. (4) An article in each language version Wikipedia has not mentioned each country.

Translation Repair Method for Improving Accuracy of Translated Sentences.

Taku Fukushima, Takashi Yoshino 0001

IEICE Trans. Inf. Syst. 97-D ( 6 ) 1528 - 1534 2014 [Refereed]

DOI

Docoitter: A Presence Display System Capable of Predicting Future In-the-room Information

田中優斗, 福島拓, 吉野孝（Part： Last author,　Corresponding author )

情報処理学会論文誌 54 ( 9 ) 2265 - 2275 2013.09 [Refereed]

現在，様々な手法を用いて在室管理が行われている．在室管理は「現在，誰がどこにいるか」を提示する．これにより，コミュニケーションの円滑化や協調作業を支援できる．しかし，提示する情報は，現在の在室の有無や行き先である．そのため，訪問者は目的の人物が不在の場合，次回の都合の良い訪問日時を把握することが困難である．そこで，我々は，現在の在室情報だけでなく，在室履歴と個人のスケジュールから未来の在室情報を予報する在室管理システム「Docoitter」を開発した．また，大学の研究室での実験から，従来から使用している紙の在室表との比較や，未来の在室情報を提示することによる効果の検証をそれぞれ行った．本研究の貢献は以下の4点にまとめられる．(1)未来の在室情報を予報する在室管理システムの提案を行い，実現した．(2)未来の在室情報の精度向上には，持ち込み計算機を利用する人や一時退席をする人を検出する必要があることを示した．(3)未来の在室情報の提示により，訪問者は研究室メンバの在室状況と訪問日時の手がかりとなる情報を得る可能性を示した．(4)未来の在室情報を提示するシステムは，至急のとき以外で利用される可能性を示した．Recently, a number of presence display systems have been fabricated by various techniques. A presence display system indicates the location of a person. Thus, such a system can facilitate communication and provide support for cooperative tasks. However, this system provides the presence and destination information of persons present. Therefore, if a person is absent, it is difficult to determine the date and time of that person's next visit. In this study, we have developed a system called Docoitter that can predict future in-the-room information about a person on the basis of that person's activity history and personal calendar. We compared the precision of our system with that of a paper-based destination bulletin board; moreover we conducted experiments in which we presented future in-the-room information at the laboratory of a university. The contributions of this study are as follows: (1) we proposed and developed a presence display system capable of predicting future in-the-room information; (2) we showed that to improve the accuracy of in-the-room information, it was necessary to detect persons using a carry-in personal computing device such as notebooks and those leaving momentarily; (3) the proposed system can provide hints regarding a person's in-the-room information, including the visit date and time, so that future in-the-room information on that person can be predicted; and (4) the system which predicts the future in-the-room information showed the possibility of being used in a case except emergency.

　 PREV - NEXT 　

<YOSHINO Takashi>

Published Papers - YOSHINO Takashi