著者

福岡 健太 浅原 正幸 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.22, no.1, pp.69-77, 2007 (Released:2007-01-05)

参考文献数

15

---

Linear-chain conditional random fields are a state-of-the-art machine learner for sequential labeling tasks. Altun investigated various loss functions for linear-chain conditional random fields. Tsuboi introduced smoothing method between point-wise loss function and sequential loss function. Sarawagi proposed semi-markov conditional random fields in which variable length of observed tokens are regarded as one node in lattice function. We propose a smoothing method among several loss functions for semi-markov conditional random fields. We draw a comparison among the loss functions and smoothing rate settings in base phrase chunking and named entity recognition tasks.

著者

村田 順平 岩沼 宏治 大塚 尚貴

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.3, pp.464-474, 2010 (Released:2010-04-06)

参考文献数

19

---

In this paper, we propose new methods and gave a system, called IFMAP , for extracting interesting patterns from a long sequential data based on frequency and self-information, and experimentally evaluate the proposed methods in the application of handling a newspaper article corpus.Sequential data mining methods based on frequency have intensively beenstudied so far. These methods, however, are not effective nor valuable for some applications where almost all high-frequent patterns should beregarded just as meaningless noisy patterns. An information-gain concept is quite important in order to restrain these noisy patterns, and was already studied for integrating it with a frequency criteria. Yang et.~al. gave a sequential mining system InfoMiner which can find periodic synchronous patterns being interesting and well-balanced from the both view-points of frequency and self-information. In this paper, we refine and extend the InfoMiner technologies in the following points: firstly, our method can handle ordinary, i.e., asynchronous and non-periodic patterns by using a sliding window mechanism, whereas InfoMiner cannot; secondly we give several combination measures for choosing valuable patterns based on frequency and self-information, while InfoMiner has just one measure which, we show in this paper, is not appropriate nor effective for handling newspaper article corpora; thirdly, we proposed a new unified method for pruning the search space of sequential data mining, which can uniformally be applied to any combination measures proposed here. We conduct experiments for evaluating the effectiveness and efficiency of the proposed method with respect to the runtime and the amount of excluding noisy patterns.

著者

山本 浩司 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.25, no.3, pp.442-451, 2010 (Released:2010-04-02)

参考文献数

11

---

In this paper, we propose a communication framework which combined two types of communication among wheelchairs and mobile devices. Due to restriction of range of activity, there is a problem that wheelchair users tend to shut themselves up in their houses. We developed a navigational wheelchair which loads a system that displays information on a map through WWW. However, this wheelchair is expensive because it needs a solid PC, a precise GPS, a battery, and so on. We introduce mobile devices and use this framework to provide information to wheelchair users and to facilitate them to go out. When a user encounters other users, they exchange messages which they have by short-distance wireless communication. Once a message is delivered to a navigational wheelchair, the wheelchair uploads the message to the system. We use two types of pheromone information which represent trends of user's movement and existences of a crowd of users. First, when users gather, ``crowd of people pheromone'' is emitted virtually. Users do not send these pheromones to the environment but carry them. If the density exceeds the threshold, messages that express ``people gethered'' are generated automatically. The other pheromone is ``movement trend pheromone'', which is used to improve probability of successful transmissions. From results of experiments, we concluded that our method can deliver information that wheelchair users gathered to other wheelchairs.

著者

小町 守 工藤 拓 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.25, no.2, pp.233-242, 2010 (Released:2010-01-26)

参考文献数

27

被引用文献数

5 6

---

Bootstrapping has a tendency, called semantic drift, to select instances unrelated to the seed instances as the iteration proceeds. We demonstrate the semantic drift of Espresso-style bootstrapping has the same root as the topic drift of Kleinberg's HITS, using a simplified graph-based reformulation of bootstrapping. We confirm that two graph-based algorithms, the von Neumann kernels and the regularized Laplacian, can reduce the effect of semantic drift in the task of word sense disambiguation (WSD) on Senseval-3 English Lexical Sample Task. Proposed algorithms achieve superior performance to Espresso and previous graph-based WSD methods, even though the proposed algorithms have less parameters and are easy to calibrate.

著者

山本 大介 増田 智樹 大平 茂輝 長尾 確

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.2, pp.243-251, 2010 (Released:2010-01-26)

参考文献数

19

---

In this paper, we propose a video scene annotation method based on tag clouds. First, user comments associated with a video are collected from existing video sharing services. Next, a tag cloud is generated from these user comments. The tag cloud is displayed on the video window of the Web browser. When users click on a tag included in the tag cloud while watching the video, the tag gets associated with the time point of the video. Users can share the information on the tags that have already been clicked. We confirmed that the coverage of annotations generated by this method is higher than that of the existing methods, and users are motivated to add tags by sharing tag clouds. This method will contribute to advanced video applications.

著者

岡 瑞起 松尾 豊

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.1, pp.1-8, 2010 (Released:2010-01-06)

参考文献数

17

被引用文献数

2

---

Measuring the weight of the relation between a pair of entities is necessary to use social networks for various purposes. Intuitively, a pair of entities has a stronger relation than another. It should therefore be weighted higher. We propose a method, using a Web search engine, to compute the weight of the relation existing between a pair of entities. Our method receives a pair of entities and various relations that exist between entities as input. It then outputs the weighted value for the pair of entities. The method explores how search engine results can be used as evidence for how strongly the two entities pertain to the relation.

著者

柴田 知秀 姜 ナウン 黒橋 禎夫

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.1, pp.224-232, 2010 (Released:2010-01-06)

参考文献数

13

---

The recent explosive increase of Web pages has made it possible for us to obtain a variety of information with a search engine. However, by some estimates, as many as 40% of the pages on the Web are duplicates of the other pages. Therefore, there is a problem that some search results contain duplicate pages. This paper proposes a method for finding similar pages from a huge amount of Web pages: hundred million Japanese Web pages. Similar pages are defined as two pages that share some sentences, and are classified into mirror pages, citation pages and plagiaristic pages, etc. First, in each page, its content region is extracted since sentences in a non-content region do not tend to be utilized for the similar page detection. From the content region in each page, relatively long sentences are extracted. This is because two pages tend to be relevant when they share relatively long sentences. A pair of pages that has the identical sentences is regarded as similar pages. Next, similar pages are classified based on several information such as an overlap ratio, the number of inlinks/outlinks, and the URL similarity. We conducted the similar page detection and classification on the large scale Japanese Web page collection, and can find some mirror pages, citation pages, and plagiaristic pages.

著者

小町 守 鈴木 久美

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.23, no.3, pp.217-225, 2008 (Released:2008-03-11)

参考文献数

16

被引用文献数

3 6

---

We propose a method for learning semantic categories of words with minimal supervision from web search query logs. Our method is based on the Espresso algorithm (Pantel and Pennacchiotti, 2006) for extracting binary lexical relations, but makes important modifications to handle query log data for the task of acquiring semantic categories. We present experimental results comparing our method with two state-of-the-art minimally supervised lexical knowledge extraction systems using Japanese query log data, and show that our method achieves higher precision than the previously proposed methods.

著者

乾 孝司 村上 浩司 橋本 泰一 内海 和夫 石川 正道

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.24, no.6, pp.469-479, 2009 (Released:2009-08-07)

参考文献数

28

被引用文献数

1

---

This paper presents a method for boosting the performance of the organization name recognition, which is a part of named entity recognition (NER). Although gazetteers (lists of the NEs) have been known as one of the effective features for supervised machine learning approaches on the NER task, the previous methods which have applied the gazetteers to the NER were very simple. The gazetteers have been used just for searching the exact matches between input text and NEs included in them. The proposed method generates regular expression rules from gazetteers, and, with these rules, it can realize a high-coverage searches based on looser matches between input text and NEs. To generate these rules, we focus on the two well-known characteristics of NE expressions; 1) most of NE expressions can be divided into two parts, class-reference part and instance-reference part, 2) for most of NE expressions the class-reference parts are located at the suffix position of them. A pattern mining algorithm runs on the set of NEs in the gazetteers, and some frequent word sequences from which NEs are constructed are found. Then, we employ only word sequences which have the class-reference part at the suffix position as suffix rules. Experimental results showed that our proposed method improved the performance of the organization name recognition, and achieved the 84.58 F-value for evaluation data.

著者

塚本 鋭 内田 誠 白山 晋

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.24, no.5, pp.438-445, 2009 (Released:2009-07-17)

参考文献数

14

被引用文献数

1 1

---

We study how initial network structure affects the evolution of cooperation in a spatial prisoner's dilemma game. The network structure is characterized by various statistical properties. In those properties, we focus on the variance of degree distribution, and inquire how it affects the evolution of cooperation. Some interactions between the variance of degree distribution and other statistical properties such as degree correlation and cluster coefficient are investigated. Moreover we compare results of static networks with those of dynamical networks generated in a process of replacing links by natural selection. It is found that a scale-free network does not always promote the evolution of cooperation, and there exists an appropriate value of the variance, at which the cooperation progresses strongly. In addition, we find that the effects of degree correlation and cluster coefficient for the evolution of cooperation vary with different variances of degree distribution.

著者

高野 敦子 池奥 渉太 北村 泰彦

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.24, no.3, pp.322-332, 2009 (Released:2009-04-16)

参考文献数

11

被引用文献数

2

---

Recently, the role of reputation information in on-line discussion groups and review sites has received much attention, and that has spurred a great deal of research on sentiment analysis of web documents. It is well known that collecting sentiment expressions, which tend to be domain-dependent, is useful for sentiment analysis. However, it can be prohibitively costly to manually collect expressions for each domain. The purpose of this paper is to propose an automatic method to acquire sentiment expressions on a specific subject from web documents. Our approach is based on a characteristic of sentiment expressions that often appear with their sentiment causes and both of them have cause-and-effect relationships. We develop a technique for recognizing cause-and-effect relationships between sentiment expressions and their sentiment causes using the results of dependency structure analysis. The proposed method uses this technique to extract sentiment causes starting from a small set of seed sentiment expressions, and extracts sentiment expressions from a set of sentiment causes. To evaluate this work, we conducted experiments using discussion board messages about hotels and sweets. The results demonstrate that the proposed method effectively extract diversified sentiment expressions relevant to each domain and possesses adequate precision. Precision is also found to be better for compound sentiment expressions.

著者

前野 義晴 大澤 幸生

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.24, no.5, pp.376-385, 2009 (Released:2009-06-11)

参考文献数

34

被引用文献数

1

---

Can we discover a node which is not observable directly but mediates the stochastic diffusion process in a network? We address such a node discovery and mathematically formulate the basic concept which is promising to solving the problem in general. The proposed method is tested with a node discovery in a Barabási-Albert model which the conventional method raised and partially succeeded in. Its performance is measured with the receiver operating characteristic curves and van Rijsbergen's F-measure (the harmonic mean of precision and recall). The proposed method succeeds in discovering an unobservable peripheral node, and an unobservable hub node in a less clustered network where the conventional method failed.

著者

小松 孝徳 山田 誠二

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.24, no.2, pp.232-240, 2009 (Released:2009-01-22)

参考文献数

14

被引用文献数

10 6

---

We describe an ``adaptation gap'' that indicates the differences between the functions of artificial agents users expect before starting their interactions and the functions they perceive after the interactions. We investigated the effects of this adaptation gap on users' impressions of the artificial agents because any variations in impressions before and after the start of an interaction determine whether the user feels that this agent is worth continuing an interaction. The results showed that the positive or negative signs of the adaptation gap and the subjective impression scores of the agents before the experiment affected the final users' impressions of the agents significantly.

著者

鈴木 輝彦 延澤 志保 太原 育夫

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.24, no.1, pp.178-190, 2009 (Released:2009-01-06)

参考文献数

9

---

Tierra and Avida are well-known models of digital organisms. They describe a life process as a sequence of computation codes. A linear sequence model may not be the only way to describe a digital organism, though it is very simple for a computer-based model. Thus we propose a new digital organism model based on a tree structure, which is rather similar to the generic programming. With our model, a life process is a combination of various functions, as if life in the real world is. This implies that our model can easily describe the hierarchical structure of life, and it can simulate evolutionary computation through mutual interaction of functions. We verified our model by simulations that our model can be regarded as a digital organism model according to its definitions. Our model even succeeded in creating species such as viruses and parasites.

著者

平尾 努 鈴木 潤 磯崎 秀樹

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.24, no.2, pp.223-231, 2009 (Released:2009-01-20)

参考文献数

17

被引用文献数

3 4 1

---

We derived the oracle summary with the highest ROUGE score that can be achieved by integrating sentence extraction with sentence compression from the reference abstract. The analysis results of the oracle revealed that summarization systems have to assign an appropriate compression rate for each sentence in the document. In accordance with this observation, this paper proposes a summarization method as a combinatorial optimization: selecting the set of sentences that maximize the sum of the sentence scores from the pool which consists of the sentences with various compression rates, subject to length constrains. The score of the sentence is defined by its compression rate, content words and positional information. The parameters for the compression rates and positional information are optimized by minimizing the loss between score of oracles and that of candidates. The results obtained from TSC-2 corpus showed that our method outperformed the previous systems with statistical significance.

著者

今別府 考洋 小野 智司 森重 綾太 黒瀬 元義 中山 茂

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.24, no.2, pp.250-262, 2009 (Released:2009-01-27)

参考文献数

45

被引用文献数

1 1

---

Quantum-inspired Evolutionary Algorithm (QEA) has been proposed as one of stochastic algorithms of evolutionary computation instead of a quantum algorithm. The authors have proposed Quantum-inspired Evolutionary Algorithm based on Pair Swap (QEAPS), which uses pair swap operator and does not group individuals in order to simplify QEA and reduce parameters in QEA. QEA and QEAPS imitationally use quantum bits as genes and superposition states in quantum computation. QEAPS has shown better search performance than QEA on knapsack problem, while eliminating parameters about immigration intervals and number of groups. However, QEAPS still has a parameter in common with QEA, a rotation angle unit, which is uncommon among other evolutionary computation algorithms. The rotation angle unit deeply affects exploitation and exploration control in QEA, but it has been unclear how the parameter influences QEAPS to behave. This paper aims to show that QEAPS involves few parameters and even those parameters can be adjusted easily. Experimental results, in knapsack problem and number partitioning problem which have different characteristics, have shown that QEAPS is competitive with other metaheuristics in search performance, and that QEAPS is robust against the parameter configuration and problem characteristics.

著者

三木 光範 廣安 知之 冨岡 弘志

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.20, no.4, pp.289-296, 2005 (Released:2005-04-19)

参考文献数

17

被引用文献数

3 6

---

We propose a parallel distributed interactive genetic algorithm(PDIGA) as a new design collaboration method. PDIGA uses an IGA and a parallel distributed genetic algorithm, which combines several IGA systems, and it has a scheme that best ones among the design solutions based on the subjective evaluation of each user are shared among users for every generation. A collaboration system using PDIGA is developed to make good design solutions among several people at difference locations. To verify a validity of the PDIGA system, we conducted experiments for comparing IGA and PDIGA. In particular, we examined each user's evaluation to the final design, compared the averages of the individual evaluation value, and examined the similarity of the design solution. These showed that the collaboration system using PDIGA become a consensus building system, and users' design solutions are unified in a group.

著者

内田 誠 白山 晋

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.23, no.6, pp.485-493, 2008 (Released:2008-08-26)

参考文献数

33

被引用文献数

1

---

We investigate a factor of the `network effect' that affects on communication service markets by a multi-agent based simulation approach. The network effect is one of a market characteristic, whereby the benefit of a service or a product increase with use. So far, the network effect has been studied in terms of macroscopic metrics, and interaction patterns of consumers in the market were often ignored. To investigate an infulence of structures of the interaction patterns, we propose a multi-agent based model for a communication serivce market, in which embedded complex network structures are considered as an interaction pattern of agents. Using several complex network models as the interaction patterns, we study the dynamics of a market in which two providers are competing. By a series of simulations, we show that the structural properties of the complex networks, such as the clustering coefficient and degree correlations, are the major factors of the network effect. We also discuss an adequate model of the interaction pattern for reproducing the market dynamics in the real world by performing simulations exploiting with a real data of social network.

著者

土井 晃一郎 山下 哲矢 田中 孝侑 山本 章博

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.23, no.3, pp.185-192, 2008 (Released:2008-03-11)

参考文献数

20

---

We present the intentional kernel as a new class of kernel functions for structured data. The class is highly contrasted to the convolution kernel, that is a typical class of kernel functions. That is, the convolution kernel is defined with sub-structures, while the intentional kernel is based on derivations constracting structures. We show instances of the intentional kernel for boolean functions, first-order terms, context sensitive languages, and RNA sequences. We also show some properties of the intentional kernel, and discuss the difference between the intentional kernel and the convolution kernel.

著者

上原 宏 佐藤 忠彦 吉田 健一

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.23, no.3, pp.205-216, 2008 (Released:2008-03-11)

参考文献数

20

被引用文献数

1

---

This paper proposes a method to measure the effects of TV advertisements on the Internet bulletin boards. It aims to clarify how the viewes' interests on TV advertisements are reflected on their images on the promoted products. Two kinds of time series data are generated based on the proposed method. First one represents the time series fluctuation of the interests on the TV advertisements. Another one represents the time series fluctuation of the images on the products. By analysing the correlations between these two time series data, we try to clarify the implicit relationship between the viewer's interests on the TV advertisement and their images on the promoted products. By applying the proposed method to an Internet bulletin board that deals with certain cosmetic brand, we show that the images on the products vary depending on the difference of the interests on each TV advertisement.