著者

椿 真史 新保 仁 松本 裕治

出版者

一般社団法人 人工知能学会

雑誌

人工知能学会論文誌 (ISSN:13460714)

巻号頁・発行日

vol.31, no.2, pp.O-FA2_1-10, 2016-03-01 (Released:2016-06-09)

参考文献数

40

---

The notion of semantic similarity between text data (e.g., words, phrases, sentences, and documents) plays an important role in natural language processing (NLP) applications such as information retrieval, classification, and extraction. Recently, word vector spaces using distributional and distributed models have become popular. Although word vectors provide good similarity measures between words, phrasal and sentential similarities derived from composition of individual words remain as a difficult problem. To solve the problem, we focus on representing and learning the semantic similarity of sentences in a space that has a higher representational power than the underlying word vector space. In this paper, we propose a new method of non-linear similarity learning for compositionality. With this method, word representations are learnedthrough the similarity learning of sentences in a high-dimensional space with implicit kernel functions, and we can obtain new word epresentations inexpensively without explicit computation of sentence vectors in the high-dimensional space. In addition, note that our approach differs from that of deep learning such as recursive neural networks (RNNs) and long short-term memory (LSTM). Our aim is to design a word representation learning which combines the embedding sentence structures in a low-dimensional space (i.e., neural networks) with non-linear similarity learning for the sentence semantics in a high-dimensional space (i.e., kernel methods). On the task of predicting the semantic similarity of two sentences (SemEval 2014, task 1), our method outperforms linear baselines, feature engineering approaches, RNNs, and achieve competitive results with various LSTM models.

著者

宮崎 勝 藤沢 寛 中川 俊夫 岡田 将吾 新田 克己

出版者

一般社団法人 人工知能学会

雑誌

人工知能学会論文誌 (ISSN:13460714)

巻号頁・発行日

vol.30, no.2, pp.429-439, 2015-03-01 (Released:2015-01-30)

参考文献数

17

被引用文献数

1

---

The spread of Video on Demand (VOD) services is driving the creation of video viewing environments in which users can watch whatever they like, whenever they like. The recent appearance of social networking services (SNSs), moreover, is bringing big changes to the world of media by enabling anyone to become a disseminator of information. We are studying a platform that combines VOD and SNS to create ``horizontal links'' between program viewers, and to facilitate encounters with new programs. To investigate user behaviors on this platform, we built an SNS site called ``teleda,'' that enables program viewing by VOD, and conducted a large-scale, three-month verification trial with about 1000 participants. In this paper, we report on the relational analysis of teleda users' viewing and communication behaviors. In order to clarify how the users' communication structures relate to the viewing and posting behaviors on this system, we described the communication structures in terms of network structures, and ran a correlation analysis of network indicators and user behavior indicators such as viewing and posting. As a result, we revealed that relationships between the users' communication structures, viewing behaviors, and posting actions are characteristics of each program's genre.

著者

ジメネス フェリックス 加納 政芳 吉川 大弘 古橋 武

出版者

一般社団法人 人工知能学会

雑誌

人工知能学会論文誌 (ISSN:13460714)

巻号頁・発行日

vol.31, no.3, pp.A-F93_1-10, 2016-05-01 (Released:2016-05-13)

参考文献数

33

被引用文献数

3

---

This paper reports feasibility of collaborative learning with educational-support robots between human. We designed a robot to prompt the constructive interaction. Constructive interaction has been regarded as a foundation of collaborative learning. Constructive interaction occurs when two students are solving the same question. Therefore, the robot is designed that alternately perform the speaker motion and listener motion to occur constructive interaction with human.In the speaker motion, the robot explains a solving method to the partner and solves a question. Moreover, the robot improves its accuracy rate as learning progress. In the listener motion, the robot does not solve a question and paying attention to a partner who is solving the questions. The robot learns while solving a question issued by learning system with a college student. The college students learned in the learning system with the robot for one month and were videoed during that time to see how they learned. This results of the study suggest that robot, which alternately solves the question with a human and improves its accuracy rate as learning progress, prompts learners to learning by constructive interaction with robot in collaborative learning. This constructive interaction indicates that learners alternately solve the question with robot and listen to robot's speaking. However, learners interest in robot decrease when robot improves its accuracy rate at 100% and performs the same action. Additionally, the screen agent, which is designed same action as robot, does not prompt some learners to learning by constructive interaction because they feel lousy that screen agent could not solve the question correctly, so they ignore what the screen agent says. The same situation was occurred in some learners learning with robot

著者

吉田 光男 荒瀬 由紀

出版者

一般社団法人 人工知能学会

雑誌

人工知能学会論文誌 (ISSN:13460714)

巻号頁・発行日

vol.30, no.1, pp.161-171, 2015-01-06 (Released:2015-01-06)

参考文献数

27

---

Query classification is an important technique for web search engines, allowing them to improve users' search experience. Specifically, query classification methods classify queries according to topical categories, such as celebrities and sports. Such category information is effective in improving web search results, online advertisements, and so on. Unlike previous studies, our research focuses on trend queries that have suddenly become popular and are extensively searched. Our aim is to classify such trend queries in a timely manner, i.e., classify the queries on the same day when they become popular, in order to provide a better search experience. To reduce the expensive manual annotation costs to train supervised learning methods, we focus on a label propagation method that belongs to the semi-supervised learning family. Specifically, the proposed method is based on our previous method that constructs a graph using a corpus, and propagates a small number of ground-truth categories of labeled queries in order to estimate the categories of unlabeled queries. We extend this method to cut ineffective edges to improve both classification accuracy and computational efficiency. Furthermore, we investigate in detail the effects of different corpora, i.e., web/blog/news search results, Tweets, and news pages, on the trend query classification task. Our experiments replicate the situation of an emerging trend query; the results show that web search results are the most effective for trend query classification, achieving a 50.1% F-score, which significantly outperforms the state-of-the-art method by 7.2 points. These results provide useful insights into selecting an appropriate dataset for query classification from the various types of data available.

著者

前田 安里紗 上間 大生 白水 菜々重 松下 光範

出版者

一般社団法人 人工知能学会

雑誌

人工知能学会論文誌 (ISSN:13460714)

巻号頁・発行日

vol.30, no.1, pp.204-215, 2015-01-06 (Released:2015-01-06)

参考文献数

18

被引用文献数

2

---

The objective of this study is to support learning of Japanese onomatopoeia for foreigners who learn Japanese. In recent years, the number of such foreigners is increasing. There are a lot of onomatopoeia words in Japanese and many of them are difficult to translate because only the number of onomatopoeia words in foreign languages (e.g., Mandarin, Cantonese) are fewer than these in Japanese. To overcome the cultural difference, this paper proposes a digital picture book system for learning Japanese onomatopoeia. The system presents 32 onomatopoeia words to a user. The design criteria of the system is that: (1) adopts an interface of user participation, (2) presents a tiny story in which onomatopoeic words are associated with pictures, and (3) enables comparison of two synonymous/antomynic onomatopoeias. We conducted a user study with foreign learners and revealed that the proposed system improves understanding of semasiological differences between two confusing onomatopoeic words.

著者

荒堀 拓哉 片上 大輔 角所 考

出版者

一般社団法人 人工知能学会

雑誌

人工知能学会論文誌 (ISSN:13460714)

巻号頁・発行日

vol.28, no.2, pp.179-186, 2013 (Released:2013-02-01)

参考文献数

17

被引用文献数

1

---

In this research, we develop the operation information dissemination agent "smart caster 24 (TWENTY FOUR)" for the purpose of the design of the expressional model as a life-like agent which enables to watch for a long time. The smart caster 24 is a life-like agent who disseminates news information on the Internet in the form of human newscaster. We analyzed total of 90 news reading motions of 18 human active newscasters and designed the nonverbal expression model for the smart caster. In the results of news reading experiments, we found that the smart caster 24 presented same impression with human newscasters in some evaluation items.

著者

梶野 洸 鹿島 久嗣

出版者

一般社団法人 人工知能学会

雑誌

人工知能学会論文誌 (ISSN:13460714)

巻号頁・発行日

vol.27, no.3, pp.133-142, 2012 (Released:2012-03-27)

参考文献数

16

被引用文献数

18 32

---

It has attracted considerable attention to use crowdsourcing services to collect a large amount of labeled data for machine learning, since crowdsourcing services allow one to ask the general public to label data at very low cost through the Internet. The use of crowdsourcing has introduced a new challenge in machine learning, that is, coping with low quality of crowd-generated data. There have been many recent attempts to address the quality problem of multiple labelers, however, there are two serious drawbacks in the existing approaches, that are, (i) non-convexity and (ii) task homogeneity. Most of the existing methods consider true labels as latent variables, which results in non-convex optimization problems. Also, the existing models assume only single homogeneous tasks, while in realistic situations, clients can offer multiple tasks to crowds and crowd workers can work on different tasks in parallel. In this paper, we propose a convex optimization formulation of learning from crowds by introducing personal models of individual crowds without estimating true labels. We further extend the proposed model to multi-task learning based on the resemblance between the proposed formulation and that for an existing multi-task learning model. We also devise efficient iterative methods for solving the convex optimization problems by exploiting conditional independence structures in multiple classifiers.

著者

谷口 忠大

出版者

一般社団法人 人工知能学会

雑誌

人工知能学会論文誌 (ISSN:13460714)

巻号頁・発行日

vol.28, no.1, pp.77-87, 2013 (Released:2013-01-05)

参考文献数

12

被引用文献数

1

---

In this paper, we propose a new learning model for decentralized autonomous smart grid involving adaptive trading agents which can sell and buy electric power effectively in a local electric power network. We name the electric power network i-Rene (inter intelligent renewable energy network). The trading agents manage the amount of electric power generated by solar panels or other renewable energies by trading electric power stored in a storage battery in a house. The agent learns a trading strategy by maximizing its utility. Based on the proposed system, we evaluated its price formation and effectiveness of the adaptive trading method through simulations. Additionally, we propose a new variable consumption model for decentralized autonomous smart grid involving living people consuming electric power and the adaptive trading agents. To model demand side management which can control the amount of electric power consumption, developing variable consumption model is essential. We added a variable consumption model to the i-Rene model. We evaluated its price formation and effectiveness of the decentralized autonomous smart grid to equalize fluctuating demand.

著者

中島 秀之

出版者

一般社団法人 人工知能学会

雑誌

人工知能学会論文誌 (ISSN:13460714)

巻号頁・発行日

vol.21, no.6, pp.502-513, 2006 (Released:2006-09-14)

参考文献数

45

被引用文献数

3 5

---

As computer scientists, we have been trained in the methodology of natural science, which is analytic in its essence. Informatics, and particularly Artificial Intelligence, is not an analytic discipline. It is required to establish a constructive methodology.

著者

柳瀬 利彦 廣木 桂一 伊藤 昭博 柳井 孝介

出版者

一般社団法人 人工知能学会

雑誌

人工知能学会論文誌 (ISSN:13460714)

巻号頁・発行日

vol.26, no.5, pp.621-637, 2011 (Released:2011-07-20)

参考文献数

34

---

We propose a computing platform for parallel machine learning. Learning from large-scale data has become common, so that parallelization techniques are increasingly applied to machine learning algorithms in order to reduce calculation time. Problems of parallelization are implementation costs and calculation overheads. Firstly, we formulate MapReduce programming model specialized in parallel machine learning. It represents learning algorithms as iterations of following two phases: applying data to machine learning models and updating model parameters. This model is able to describe various kinds of machine learning algorithms, such as k-means clustering, EM algorithm, and linear SVM, with comparable implementation cost to the original MapReduce. Secondly, we propose a fast machine learning platform which reduces the processing overheads at iterative procedures of machine learning. Machine learning algorithms iteratively read the same training data in the data application phase. Our platform keeps the training data in local memories of each worker during iterative procedures, which leads to acceleration of data access. We evaluate performance of our platform on three experiments. Our platform executes k-means clustering 2.85 to 118 times faster than the MapReduce approach, and shows 9.51 times speedup with 40 processing cores against 8 cores. We also show the performance of Variational Bayes clustering and linear SVM implemented on our platform.

著者

中崎 寛之 川場 真理子 横本 大輔 宇津呂 武仁 福原 知宏

出版者

一般社団法人 人工知能学会

雑誌

人工知能学会論文誌 (ISSN:13460714)

巻号頁・発行日

vol.25, no.5, pp.613-622, 2010 (Released:2010-08-06)

参考文献数

12

被引用文献数

1

---

The overall goal of this paper is to cross-lingually analyze multilingual blogs collected with a topic keyword. The framework of collecting multilingual blogs with a topic keyword is designed as the blog feed retrieval procedure. In this paper, we take an approach of collecting blog feeds rather than blog posts, mainly because we regard the former as a larger information unit in the blogosphere and prefer it as the information source for cross-lingual blog analysis. In the blog feed retrieval procedure, we also regard Wikipedia as a large scale ontological knowledge base for conceptually indexing the blogosphere. The underlying motivation of employing Wikipedia is in linking a knowledge base of well known facts and relatively neutral opinions with rather raw, user generated media like blogs, which include less well known facts and much more radical opinions. In our framework, first, in order to collect candidates of blog feeds for a given query, we use existing Web search engine APIs, which return a ranked list of blog posts, given a topic keyword. Next, we re-rank the list of blog feeds according to the number of hits of the topic keyword as well as closely related terms extracted from the Wikipedia entry in each blog feed. We compare the proposed blog feed retrieval method to existing Web search engine APIs and achieve significant improvement. We then apply the proposed blog distillation framework to the task of cross-lingually analyzing multilingual blogs collected with a topic keyword. Here, we cross-lingually and cross-culturally compare less well known facts and opinions that are closely related to a given topic. Results of cross-lingual blog analysis support the effectiveness of the proposed framework.

著者

倉橋 節也 横幕 春樹 矢嶋 耕平 永井 秀幸

出版者

一般社団法人 人工知能学会

雑誌

人工知能学会論文誌 (ISSN:13460714)

巻号頁・発行日

vol.37, no.1, pp.C-L42_1-9, 2022-01-01 (Released:2022-01-01)

参考文献数

21

被引用文献数

2

---

In this paper, we propose a new SEIR model for COVID-19 infection prediction using mobile statistics and evolutionally optimisation, which takes into account the risk of influx. The model is able to predict the number of infected people in a region with high accuracy, and the results of estimation in Sapporo City and Tokyo Metropolitan show high prediction accuracy. Using this model, we analyse the impact of the risk of influx to Sapporo City and show that the spread of infection in November could have been reduced to 0.6 if the number of influxes had been limited after the summer. We also examine the preventive measures called for in the emergency declaration in the Tokyo metropolitan area. We found that comprehensive measures are highly effective, and estimated the effect of vaccination and circuit breakers on the spread of infection after the spring of 2021 using the effective reproduction reduction rate of infection control measures obtained from the individual-based model and the SEIR model.

著者

池田 圭佑 榊 剛史 鳥海 不二夫 風間 一洋 野田 五十樹 諏訪 博彦 篠田 孝祐 栗原 聡

出版者

一般社団法人 人工知能学会

雑誌

人工知能学会論文誌 (ISSN:13460714)

巻号頁・発行日

vol.31, no.1, pp.NFC-C_1-13, 2016-01-06 (Released:2016-09-26)

参考文献数

19

被引用文献数

1

---

During the 2011 East Japan Great Earthquake Disaster, some people used social media such as Twitter to get information important to their lives. However, the spread of groundless rumor information was big social problem. Therefore, social media users pay attention to prevent wrong information from diffusing. The way to stop the spread of a false rumor is needed, so we have to understand a diffusion of information mechanism. We have proposed information diffusion model which is based on SIR model until now. This model is represented by the stochastic state transition model for whether to propagate the information, and its transition probability is defined as the same value for all agents. People ’s thinking or actions are not the same. To solve this problem, we adopted three elements in our model: A new internal state switching model, user diversity and multiplexing of information paths. In this paper, we propose a novel information diffusion model, the Agent-based Information Diffusion Model (AIDM). We reproduce two kinds of false rumor information diffusion using proposed model. One is “single burst type false rumor spread ”, and another is “multi burst type false rumor spread. ”Proposal model is estimated by comparing real data with a simulation result.

著者

林 勝悟 谷本 啓 鹿島 久嗣

出版者

一般社団法人 人工知能学会

雑誌

人工知能学会論文誌 (ISSN:13460714)

巻号頁・発行日

vol.35, no.5, pp.B-K33_1-9, 2020-09-01 (Released:2020-09-01)

参考文献数

26

---

The recent rapid and significant increase of big data in our society has led to major impacts of machine learningand data mining technologies in various fields ranging from marketing to science. On the other hand, there still existareas where only small-sized data are available for various reasons, for example, high data acquisition costs or therarity of targets events. Machine learning tasks using such small data are usually difficult because of the lack ofinformation available for training accurate prediction models. In particular, for long-term time-series prediction, thedata size tends to be small because of the unavailability of the data between input and output times in training. Suchlimitations on the size of time-series data further make long-term prediction tasks quite difficult; in addition, thedifficulty that the far future is more uncertain than the near future.In this paper, we propose a novel method for long-term prediction of small time-series data designed in theframework of generalized distillation. The key idea of the proposed method is to utilize the middle-time data betweenthe input and output times as “privileged information,” which is available only in the training phase and not in thetest phase. We demonstrate the effectiveness of the proposed method on both synthetic data and real-world data. Theexperimental results show the proposed method performs well, particularly when the task is difficult and has highinput dimensions.

著者

藤井 樹 伊藤 靖展 鍋島 英知

出版者

一般社団法人 人工知能学会

雑誌

人工知能学会論文誌 (ISSN:13460714)

巻号頁・発行日

vol.34, no.3, pp.A-I91_1-16, 2019-05-01 (Released:2019-05-01)

参考文献数

14

---

We consider the laboratory assignment problem in which laboratories have minimum and maximum quotas. MSDA proposed by Fragiadakis, et al., is an efficient algorithm to solve the laboratory assignment problem, but it is incomplete to find a fair assignment. In this paper, we show three extensions of the laboratory assignment problem and translations from the extensions to constraint optimization problems. The first extension enables a completeness to find a fair assignment if it exists, but loses strategy-proofness. The original and first extended laboratory assignment problem may have no fair assignment. This is caused by redundant claims of an empty seat. The second extension based on the first one ensures that the laboratory assignment problem has at least one fair assignment by making a claim of an empty seat stricter. The third extension introduces tied ranks to students' preferences over laboratories, for example, students can specify multiple laboratories as their first choice. This extension gives the Žexibility to specify students' preferences and makes it possible to find more desirable assignments. The experimental results show that our approaches requires more computational time compared with MSDA, but can always find a fair assignment and a more desirable one.

著者

Kentaro Kanamori Takuya Takagi Ken Kobayashi Hiroki Arimura

出版者

The Japanese Society for Artificial Intelligence

雑誌

人工知能学会論文誌 (ISSN:13460714)

巻号頁・発行日

vol.36, no.6, pp.C-L44_1-12, 2021-11-01 (Released:2021-11-01)

参考文献数

51

---

Post-hoc explanation methods for machine learning models have been widely used to support decision-making. Counterfactual Explanation (CE), also known as Actionable Recourse, is one of the post-hoc explanation methods that provides a perturbation vector that alters the prediction result obtained from a classifier. Users can directly interpret the perturbation as an “action” to obtain their desired decision results. However, actions extracted by existing methods often become unrealistic for users because they do not adequately consider the characteristics corresponding to the data distribution, such as feature-correlations and outlier risk. To suggest an executable action for users, we propose a new framework of CE, which we refer to as Distribution-Aware Counterfactual Explanation (DACE), that extracts a realistic action by evaluating its reality on the empirical data distribution. Here, the key idea is to define a new cost function based on the Mahalanobis distance and the local outlier factor. Then, we propose a mixed-integer linear optimization approach to extracting an optimal action by minimizing the defined cost function. Experiments conducted on real datasets demonstrate the effectiveness of the proposed method compared with existing CE methods.

著者

武田 龍 駒谷 和範 中島 圭祐 中野 幹生

出版者

一般社団法人 人工知能学会

雑誌

人工知能学会論文誌 (ISSN:13460714)

巻号頁・発行日

vol.37, no.3, pp.IDS-B_1-9, 2022-05-01 (Released:2022-05-01)

参考文献数

24

---

Dialogue system development involves a variety of factors and requires multifaceted consideration, so design guidelines would be helpful. Although a neural-based approach can be used, it requires a vast amount of dialogue data and would take too much effort to collect them to develop a system for a specific and fixed-length dialogue. Furthermore, errors in automatic speech recognition and language understanding should be explicitly considered in the design because they are inevitable when the system talks with general users and would lower their impressions. We propose design guidelines for developing dialogue systems. Our systems developed with the aid of these guidelines took first place in two dialogue system competitions: the situation track of the second Dialogue System Live Competition and a pre-preliminary test of the Dialogue Robot Competition. Our proposed design guidelines are to (1) make the system take initiative, (2) avoid dialogue flows from relying too much on user utterances, and (3) include in system utterances that the system understands what the user said. We also show more details regarding the systems designed for each of the two competitions with examples, such as the dialogue examples in the two competitions and the scores of questionnaire by real users.

著者

川久保 佐記 和泉 潔

出版者

一般社団法人 人工知能学会

雑誌

人工知能学会論文誌 (ISSN:13460714)

巻号頁・発行日

vol.31, no.6, pp.AG-D-1-10, 2016-11-01 (Released:2016-11-02)

参考文献数

29

---

The effect of option markets on their underlying markets has been studied intensively since the first option market launched. Despite considerable efforts, including the development of theoretical and empirical approaches, we do not yet have conclusive evidence on this effect. We investigate the effect of option markets, especially that of dynamic hedging, on their underlying markets by using an artificial market. We propose a two-market model in which an option market and its underlying market interact. We confirmed that trading behaviors on expire date are not effect on its underlying market, but dynamic hedging, arbitrage trading changed volatility on the price of underlying asset under certain conditions.

著者

松添 静子 田中 文英

出版者

一般社団法人 人工知能学会

雑誌

人工知能学会論文誌 (ISSN:13460714)

巻号頁・発行日

vol.28, no.2, pp.170-178, 2013 (Released:2013-01-17)

参考文献数

16

被引用文献数

2 9

---

In case of educational support using a partner robot, its individual robot character is considered to play an important role. Based on this consideration, we focus on the excellence property of the robot, assuming that the difference of robot excellence affects the performance of a human learner who joins a learning activity together with the robot. In this paper, we report on a field experiment that was conducted at an English learning school for Japanese children (4--8 years of age). We divided 19 participants into three experimental conditions in which we introduced a small humanoid robot with three different excellence levels one for each condition: Condition A robot was designed to be with the highest excellence level. The robot could answer to any question given by a human teacher correctly. In contrast, Condition B robot could not answer correctly at the beginning but could learn from partner humans who could take the robot by the hand and teach it step by step. Finally, Condition C robot was designed to be with the lowest excellence level. Condition C robot could neither answer to any question from beginning nor learn from partner humans. Experimental results showed that the learning performance of participants who joined a drawing-game lesson for English names of various shapes together with Condition A or Condition B robot was more increased than the case with Condition C robot. In addition, Condition B and Condition C robots were found to be more effective in motivating participants in the learning activity. Those findings would be valuable for a better future design of a partner robot whose goal is to enrich and support educational activities in classrooms.

著者

藤澤 瑞樹 齋藤 豪 奥村 学

出版者

The Japanese Society for Artificial Intelligence

雑誌

人工知能学会論文誌 = Transactions of the Japanese Society for Artificial Intelligence : AI (ISSN:13460714)

巻号頁・発行日

vol.19, pp.483-492, 2004-11-01

参考文献数

9

被引用文献数

3

---

Previous commentary systems generate commentaries only from the viewpoint of a commentator. However there are various viewpoints for comments, and such different viewpoints invoke various comments. The amount of information about a situation may differ between the viewpoints, and the understandings of the situation may also differ between them. In this paper, we propose a method to generate commentaries automatically so that users can easily understand situations by taking into account the different understandings of the situations between viewpoints. Our method is composed of two parts. The first is generation of comment candidates about the current situation, unexpected actions, intentions of players by using a game tree. The second is comment selection which chooses comments related to the prior one so that listeners can compare the situations from different viewpoints. Based on our approach, we implemented an experimental system that generates commentaries on mahjong games. We discuss the output of the system.