著者

藤原 浩司 兼岩 憲

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.29, no.4, pp.364-374, 2014-07-01 (Released:2014-06-18)

参考文献数

21

---

As one of the core technologies of the Semantic Web, the RDF data model enables us to represent machine-readable metadata about Web resources. SPARQL is the standard query language for answering queries from RDF data. In order to support SPARQL in practice, it is important to access such metadata quickly from large scale RDF data stores, such as Gene Ontology and DBpedia. In this paper, we present an algorithm that decides the solve order of an inputed query. We formalize the search space reduction of the connected variables occurring in a query and establish the costs of different query patterns. In particular, we consider selecting one of the various orders of elements in each query that results in less computation for searching and reasoning steps, i.e., matching a subgraph of a complex RDF graph. Using the algorithm, we implement an efficient query system with a RDF store (called NodeStore) for large-scale RDF data. In the RDF store, an indexed data structure of RDF graphs is well constructed for optimizing RDF data processing, e.g., finding a set of RDF triples including a common resource. For the evaluation of deciding the solve order of a query, we show some experimental results for our query system NodeStore and the Jena framework using a LUBM dataset (a benchmarking framework for semantic repositories).

著者

込山 悠介 番野 雅城 鑓水 優行 加藤 文彦 大向 一輝 武田 英明 清水 謙多郎

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.29, no.4, pp.356-363, 2014-07-01 (Released:2014-06-18)

参考文献数

12

---

Researchers of agriculture, life science and drug design of the need to acquire information that combines two or more life science databases for problem solving. Semantic Web technologies are already necessary for data integration between those databases. This study introduces a technique of utilizing RDF (Resource Description Framework) and OWL (Web Ontology Language) as a data set for development of a machine learning predictor of interactomics. Also, for SPARQL (SPARQL Protocol and RDF Query Language) we sketched the implementing method of interactomics LOD (Linked Open Data) in the graph database. Interactomics LOD has included the pairs of protein--protein interactions of tyrosine kinase, the pairs of amino acid residues of sugar (carbohydrate) binding proteins, and cross-reference information of the protein chain among an entry of major bioscience databases since 2013. Finally, we designed three RDF schema models and made access possible using AllegroGraph 4.11 and Virtuoso 7. The number of total triples was 1,824,859,745 in these databases. It could be combined with public LOD of the life science domain of 28,529,064,366 triples and was able to be searched. We showed that it was realistic to deal with large-scale LOD on a comparatively small budget by this research. The cost cut by LOD decreased not only expense but development time. Especially RDF-SIFTS (Structure Integration with Function, Taxonomy and Sequence) that is an aggregate of 10 small LOD was constructed in the short period of BioHackathon 2013 or was developed in one week. We could say that we can obtain quickly a data set required for the machine learning of interactomics by using LOD. We set up the interactomics LOD for application development as a database. SPARQL endpoints of these databases are exhibited on the portal site UTProt (The University of Tokyo Protein, http://utprot.net).

著者

小嵜 耕平 新保 仁 小町 守 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.28, no.4, pp.400-408, 2013 (Released:2013-06-20)

参考文献数

30

---

Graph construction is an important step in graph-based semi-supervised classification. While the k-nearest neighbor graphs have been the de facto standard method of graph construction, this paper advocates using the less well-known mutual k-nearest neighbor graphs for high-dimensional natural language data. To evaluate the quality of the graphs apart from classification algortihms, we measure the assortativity of graphs. In addition, to compare the performance of these two graph construction methods, we run semi-supervised classification methods on both graphs in word sense disambiguation and document classification tasks. The experimental results show that the mutual k-nearest neighbor graphs, if combined with maximum spanning trees, consistently outperform the k-nearest neighbor graphs. We attribute better performance of the mutual k-nearest neighbor graph to its being more resistive to making hub vertices. The mutual k-nearest neighbor graphs also perform equally well or even better in comparison to the state-of-the-art b-matching graph construction, despite their lower computational complexity.

著者

長井 拓馬 兼岩 憲

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.29, no.3, pp.343-355, 2014-05-01 (Released:2014-06-05)

参考文献数

23

---

In this paper, we propose an algorithm for ALCH(D) concept learning from RDF data using minimal model reasoning. This algorithm generates concept expressions in the Description Logic ALCH(D) by giving background knowledge and positive and negative examples in the RDF form. Our method can be widely applied to RDF data on the Web, as background knowledge. An advantage of the method for RDF data is that reasoning on RDF graphs is tractable compared to logical reasoning for OWL data. We solve the problem that RDF data cannot be directly applied to the concept learning due to its less expressive power, speci.cally, the lack of negative expressions. In order to construct expressive ALCH(D) concepts from less expressive RDF data in the concept learning, we introduce (nonmonotonic) inference rules based on minimal model reasoning which derive implicit subclass and subproperty relations from the background knowledge in the RDF form. We prove the soundness, completeness and decidability of the nonmonotonic RDF reasoning in the minimal Herbrand models for RDF graphs. The process of concept learning is divided in two parts: (i) concept generation and (ii) concept evaluation. In the concept generation, minimal model reasoning enables us to derive complex concepts consisting of negation, conjunction, disjunction and quanti.ers and to exclude inconsistent concepts. In the concept evaluation, we evaluate hypothesis concepts with class and property hierarchies where minimal model reasoning is used for expressing more speci.c concepts as the answer for learning. We implement a system that learns some ALCH(D) concepts describing the features of given examples.

著者

中辻 真 藤原 靖宏 内山 俊郎 戸田 浩之

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.28, no.6, pp.457-467, 2013-11-01 (Released:2013-10-01)

参考文献数

33

被引用文献数

1

---

Tracking user interests over time is important for making accurate recommendations. However, the widely-used time-decay-based approach worsens the sparsity problem because it deemphasizes old item transactions. We introduce two ideas to solve the sparsity problem. First, we divide the users’ transactions into epochs i.e. time periods, and identify epochs that are dominated by interests similar to the current interests of the active user. Thus, it can eliminate dissimilar transactions while making use of similar transactions that exist in prior epochs. Second, we use a taxonomy of items to model user item transactions in each epoch. This well captures the interests of users in each epoch even if there are few transactions. It suits the situations in which the items transacted by users dynamically change over time; the semantics behind classes do not change so often while individual items often appear and disappear. Fortunately, many taxonomies are now available on the web because of the spread of the Linked Open Data vision. We can now use those to understand dynamic user interests semantically. We evaluate our method using a dataset, a music listening history, extracted from users’ tweets and one containing a restaurant visit history gathered from a gourmet guide site. The results show that our method predicts user interests much more accurately than the previous time-decay-based method.

著者

佐藤 正平 狩野 均

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.311-319, 2010 (Released:2010-02-25)

参考文献数

25

被引用文献数

2

---

In this paper, we propose a new method to obtain the transition rules of two-dimensional cellular automata (CA) that performs grayscale image processing. CA has the advantages of producing complex systems from the local interaction of simple elements, and has attracted increased research interest. The difficulty of designing CA's transition rules to perform a particular task has severely limited their applications. So, the evolutionary design of CA rules has been studied. In this method, an evolutionary algorithm was used to evolve CA. In recent years, this method has been applied to image processing. Rosin has studied the evolutionary design of two-dimensional CA to perform noise reduction, thinning and convex hulls. Batouche et al. and Slatnia et al. employed genetic algorithm to investigate the possibility of CA to perform edge detection. In the previous methods, 2-state CA was used for binary image processing. Unlike the previous methods, the present method uses 256-state CA rules to perform grayscale image processing. Gene Expression Programming (GEP) proposed by Ferreira is employed as a learning algorithm in which the chromosomes encode the transition rules as expression trees. Experimental results for the reduction of impulse noise, salt-and-pepper noise and gaussian noise show that the proposed method is equivalent to previous methods in performance and more than 100 times faster than the method proposed by Rosin. We show that the rule obtained by the proposed method employs symmetry-based strategy in the noise reduction process and this property can reduce complexity of CA.

著者

鳴海 真里子 今井 倫太

出版者

一般社団法人 人工知能学会

雑誌

人工知能学会全国大会論文集 第17回全国大会(2003)

巻号頁・発行日

pp.34, 2003 (Released:2004-02-03)

---

本研究では、ロボット上の音声対話システムにセンサの情報とアクチュエータの動作を反映させる機構を実装し、ロボットが演出を行うことにより人間ロボット間で共感を実現させることを目的とする。実験により、その有効性を検証する。

著者

岩橋 直人 谷口 忠大 Naoto Iwahashi Tadahiro Taniguchi

出版者

一般社団法人 人工知能学会

雑誌

人工知能学会誌 = Journal of Japanese Society for Artificial Intelligence (ISSN:09128085)

巻号頁・発行日

vol.27, no.6, pp.544-545, 2012-11-01

著者

鹿島 久嗣 馬場 雪乃 Hisashi Kashima Yukino Baba

出版者

一般社団法人 人工知能学会

雑誌

人工知能 = journal of the Japanese Society for Artificial Intelligence (ISSN:21882266)

巻号頁・発行日

vol.29, no.1, pp.4-11, 2014-01-01

著者

大島 直樹 山口 雄大 デシルバ ラビンドラ 岡田 美智男

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.29, no.3, pp.288-300, 2014-05-01 (Released:2014-04-04)

参考文献数

17

---

Understanding why people treat simple geometric animations like real agent which has intention to interact with people even if its geometry is artificial thing will aid the `agency' problems of human-agent interaction. This paper explores effects of treating simple geometric animations as a real participant to facilitate multi-party conversation in social interaction. Observational study was conducted with groups of two or three persons using simple circle (sociable spotlight) which moves based on dynamic information in the current multi-party conversation, with the goal of discovering how participants are utilizing the behaviors of sociable spotlight as an other party for organizing the conversational sequences in talk-in-interaction. In addition, we motivated to explore how the sociable spotlight is embedded within the organization of conversation and how the user's behaviors are changed according to the sociable spotlight's behaviors by investigate through conversation analysis of a video-recording. Finally, we conclude how the agency of artificial things constructed in multi-party conversation from minimal designing point of view.

著者

杉山 貴昭 駒谷 和範 佐藤 理史

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.29, no.1, pp.32-40, 2014-01-05 (Released:2014-01-07)

参考文献数

16

被引用文献数

2 1

---

We have tackled a novel problem of predicting when a user is likely to begin speaking to a humanoid robot. The generality of the prediction model should be examined to apply it to various users. We show in this paper that the following two empirical evaluations. First, our proposed model does not depend on the specific participants whose data were used in our previous experiment. Second, the model can handle variations caused by individuality and instruction. We collect a data set to which 25 human participants give labels, indicating whether or not they would be likely to begin speaking to the robot. We then train a new model with the collected data and verify its performance by cross validation and open tests. We also investigate relationship of how much each human participant felt possible to begin speaking with a model parameter and instruction given to them. This shows a possibility of our model to handle such variations.

著者

松永 健太 山田 健太 高安 秀樹 高安 美佐子

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.6, pp.365-375, 2012 (Released:2012-09-27)

参考文献数

27

被引用文献数

2

---

The dealer model is an agent based model that simulates the simplified dealer's behavior and satisfies various empirical laws of the foreign exchange markets by tuning major three parameters. In this study,we improve the dealer model to satisfy a newly established empirical law about widening of spread as a response to big market price changes. As a result when a big news occurs and the market becomes turbulent, this new model can reproduce broadening of distribution of price change.In a peculiar price change of official intervention in the foreign exchange market by Bank of Japan, this model can be used for estimation of strategies of intervention and responses of the market.

著者

杉原 貴彦 劉 欣 村田 剛志

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.28, no.1, pp.67-76, 2013 (Released:2013-01-05)

参考文献数

18

被引用文献数

2

---

Many real-world complex systems can be modeled as networks, and most of them exhibit community structures. Community detection from networks is one of the important topics in link mining. In order to evaluate the goodness of detected communities, Newman modularity is widely used. In real world, however, many complex systems can be modeled as signed networks composed of positive and negative edges. Community detection from signed networks is not an easy task, because the conventional detection methods for normal networks cannot be applied directly. In this paper, we extend Newman modularity for signed networks. We also propose a method for optimizing our modularity, which is an efficient hierarchical agglomeration algorithm for detecting communities from signed networks. Our method enables us to detect communities from large scale real-world signed networks which represent relationship between users on websites such as Wikipedia, Slashdot and Epinions.

著者

堀口 祐樹 小島 一晃 松居 辰則

出版者

一般社団法人 人工知能学会

雑誌

人工知能学会全国大会論文集 第22回全国大会(2008)

巻号頁・発行日

pp.10, 2008 (Released:2009-07-31)

---

e-learnignには教師が学習者の様子を観察できないという弱点がある。本e-learningシステムは、マウスの移動速度情報などを用いて学習者の集中の度合いなどをリアルタイムに評価するものである。

著者

山本 修身 佐藤根 寛

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.2, pp.419-426, 2011

---

The fifteen puzzle is a sliding puzzle which has fifteen pieces on which numbers from 1 to 15 are printed. Using the IDA* algorithm with an admissible evaluation function, we can obtain an optimal solution of the puzzle. The performance of the algorithm depends on the evaluation function. The most simple evaluation function is the Manhattan evaluation function, whose value is the sum of the Manhattan distances from the positions of the corresponding pieces in the goal configuration. In this paper, we propose an evaluation function whose values are greater than or equal to that of the Manhattan evaluation function. Our evaluation function refers an approximated database of the gap-2<i>n</i> set. The database is computed beforehand like pattern databases, but it is completely different from pattern databases. The belongingness of a configuration of pieces to the set has to be checked by the database. Using an evaluation function based on the gap-8 set, we were able to reduce the number of search nodes to about 2.5×10^-4 times in average with the IDA* algorithm compared with the Manhattan evaluation function. We also show that combining an evaluation function by gap-8 set and an evaluation function by additive pattern databases of disjoint seven and eight pieces, we were able to reduce the number of search nodes by about 53 compared with the evaluation function only by the additive pattern databases.

著者

鳥海 不二夫 石井 健一郎

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.6, pp.346-354, 2012 (Released:2012-10-03)

参考文献数

12

被引用文献数

1

---

The concept of the ``wisdom of crowds'' has attracted attention for finding new insights by appropriately processing the large amount of information possessed by crowds. A prediction market is one estimating method that uses the mechanisms of financial markets such as stock or exchange markets to realize the ``wisdom of crowds''. In this study, we use agent-based simulation to clarify the condition that makes prediction markets effective. An artificial market is a virtual financial market run on a computer. Agents participate in them as computer programs that play the role of virtual dealers. In the simulation, we confirm the influence of the following parameters: information transmission frequency, the retention of motivation, and the gap of information recieve abilities. The results of this study suggest that prediction markets realize more accurate results than opinion polls under the following conditions: the gap of information recieve abilities and relatively low motivation.

著者

石下 円香 佐藤 充 森 辰則

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.4, pp.339-350, 2009 (Released:2009-05-22)

参考文献数

15

被引用文献数

2

---

In this paper, we propose a method of non-factoid Web question-answering that can uniformly deal with any class of Japanese non-factoid question by using a large number of example Q&A pairs. Instead of preparing classes of questions beforehand, the method retrieves already asked question examples similar to a submitted question from a set of Q&A pairs. Then, instead of preparing clue expressions for the writing style of answers according to each question class beforehand, it dynamically extracts clue expressions from the answer examples corresponding to the retrieved question examples. This clue expression information is combined with topical content information from the question to extract appropriate answer candidates. The score of an answer candidate is measured by the density of submitted question's keywords, words associated with the question and the clue expressions. Note that we utilize the set of Q&A pairs, not to find answers from them, but to obtain clue expressions about the writing style of their answers. The information source for question answering is the Web documents retrieved by using an API of a Web search engine. Experimental results showed that the clue expressions obtained from the set of examples improved the accuracy of answer candidate extraction.

著者

市野 順子 田野 俊一

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.25, no.3, pp.504-513, 2010 (Released:2010-04-13)

参考文献数

39

被引用文献数

1

---

In this study, we investigated the relationship between phases of meeting and non-verbal speech information. We considered that conversations at the meeting must show information to phases of the meeting as non-verbal features. We attempted to discriminate between the divergence phase and the convergence phase by the decision tree method using only non-verbal speech information. We performed an experiment with a group task based on a modification of the game Twenty-Questions and recorded participants' speech data. In a discrimination test, we used the recorded speech, and defined non-verbal speech features such as switching pauses (i.e. silent intervals between the utterance of two speakers), frequency for each turn-taking pattern and duration. We conducted the two discrimination tests for using parameters with friends group, with strangers group and with both groups. From the results, the accuracy of the open test is 77.3%, 85.2% and 77.3%, respectively. Taking into account only non-verbal speech information was used, we consider these results to be fairly good.

著者

竹内 誉羽 庄野 修 辻野 広司

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.2, pp.92-102, 2012 (Released:2012-02-10)

参考文献数

23

---

Future robots/agents will perform situated behaviors for each user. Flexible behavioral learning is required for coping with diverse and unexpected users' situations. Unexpected situations are usually not tractable for machine learning systems that are designed for pre-defined problems. In order to realize such a flexible learning system, we were trying to create a learning model that can function in several different kinds of state transitions without specific adjustments for each transition as a first step. We constructed a modular neural network model based on reinforcement learning. We expected that combining a modular architecture with neural networks could accelerate the learning speed of neural networks. The inputs of our neural network model always include not only observed states but also memory information for any transition. In pure Markov decision processes, memory information is not necessary, rather it can lead to lower performance. On the other hand, partially observable conditions require memory information to select proper actions. We demonstrated that the new learning model could actually learn those multiple kinds of state transitions with the same architectures and parameters, and without pre-designed models of environments. This paper describes the performances of constructed models using probabilistically fluctuated Markov decision processes including partially observable conditions. In the test transitions, the observed state probabilistically fluctuated. The new learning model could function in those complex transitions. In addition, the learning speeds of our model are comparable to a reinforcement learning algorithm implemented with a pre-defined and optimized table-representation of states.

著者

安場 直史 長岡 諒 矢野 純史 香川 浩司 森田 哲郎 沼尾 正行 栗原 聡

出版者

一般社団法人 人工知能学会

雑誌

人工知能学会全国大会論文集 第22回全国大会(2008)

巻号頁・発行日

pp.16, 2008 (Released:2009-07-31)

---

ナビゲーションシステムのルート探索手法において、熟考型探索は最適だが計算時間が長く、即応型は計算時間は短いが解が最適でない。短時間先の渋滞予測情報が提供される環境で、相反するこれらの探索を適宜使い分けることでより精度の高いルート探索を実現した。