著者
土斐崎 龍一 飯場 咲紀 岡谷 貴之 坂本 真樹
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.30, no.1, pp.124-137, 2015-01-06 (Released:2015-01-06)
参考文献数
51
被引用文献数
1 1

With the widespread use of online shopping in recent years, consumer search requests for products have become more diverse. Previous web search methods have used adjectives as input by consumers. However, given that the number of adjectives that can be used to express textures is limited, it is debatable whether adjectives are capable of richly expressing variations of product textures. In Japanese, tactile and visual textures are easily and frequently expressed by onomatopoeia, such as ``fuwa-fuwa'' for a soft and light sensation and ``kira-kira'' for a glossy texture. Onomatopoeia are useful for understanding not only material textures but also a user's intuitive, sensitive, and even ambiguous feelings evoked by materials. In this study, we propose a system to rank FMD images corresponding to texture associated with Japanese onomatopoeia based on their symbolic sound associations between the onomatopoeia phonemes and the texture sensations. Our system quantitatively estimates the texture sensations of input onomatopoeia, and calculates the similarities between the users' impressions of the onomatopoeia and those of the images. Our system also suggests the images which best match the input onomatopoeia. An evaluation of our method revealed that the best performance was achieved when the SIFT features, the colors of the images, and text describing impressions of the images were used.
著者
幸島 匡宏 松林 達史 澤田 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.30, no.6, pp.745-754, 2015-11-01 (Released:2015-10-27)
参考文献数
18
被引用文献数
4

In this paper, we propose a new Non-negative Matrix Factorization (NMF) method for consumer behavior pattern extraction. NMF is one of the pattern extraction method and is formulated to factorize a non-negative matrix into the product of two factor matrices. Since various types of datasets are represented by non-negative matrices, NMF could be applied in wide range of research fields including marketing science, natural language processing and brain signal processing. However, more effective extension method is required in a purchase log analysis in marketing operation since marketer needs to extract interpretable patterns from sparse matrix in which most of the elements are zero. Therefore, we propose Non-negative Micro Macro Mixed Matrix Factorization (NM4F) which uses attribution information of both users and items to improve interpretability and capability to deal with sparsity. NM4F is formulated as a method which could simultaneously factorize multiple matrices using shared factor matrices and linear constraint between factor matrices. This formulation enables to increase an amount of available information and to extract consistent patterns with several different aspect. We derive the parameter estimation algorithm by multiplicative update rules. We confirmed the effectiveness of the proposed method in terms of both quality and quantity by using real consumer panel dataset. In addition, we discuss a relation between extracted patterns by the visualization results using graph drawing.
著者
田和辻 可昌 村松 慶一 松居 辰則
出版者
一般社団法人 人工知能学会
雑誌
人工知能学会論文誌 (ISSN:13460714)
巻号頁・発行日
vol.30, no.5, pp.626-638, 2015-09-01 (Released:2015-08-27)
参考文献数
20
被引用文献数
1

In the research field of Human-agent interaction, the uncanny valley is the crucial issue to realize co-existence of human and artificial agents. It is referred to as the phenomenon that human can feel repulsive against the agents whose appearance is considerably humanlike. There has been just theoretical based but not verifiable model providing an explanation for how it occurs. We hypothesized that when human observes that humanlike agent, s/he perceives it as both human and non-human, and the contradiction between the perceptions causes him or her to elicit negative response against it. We conducted the experiment where the participants were asked to judge whether face of agents or a person was depicted as that of a real person, with their eye tracked and their gaze direction estimated. The results indicated that observers had two-step information processing to the agent. Above all, we proposed a model generating the human negative response against humanlike agents, taking into consideration of the function of the brain regions such as amygdala, prefrontal cortex, hippocampus, and striatum. To verify the model, the transition of the emotional value (namely, positive or negative) was simulated on the basis of a qualitative description for the model. It is suggested that the model be proposed which is verifiable with many findings in the field of neuroscience.
著者
山腰 貴大 小川 泰弘 駒水 孝裕 外山 勝彦
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.35, no.1, pp.H-J53_1-14, 2020-01-01 (Released:2020-01-01)
参考文献数
20
被引用文献数
3

We propose a method that assists legislation drafters in finding inappropriate use of Japanese legal terms and their corrections from Japanese statutory sentences. In particular, we focus on sets of similar legal terms whose usages are strictly defined in legislation drafting rules that have been established over the years. In this paper, we first define input and output of legal term correction task. We regard it as a special case of sentence completion test with multiple choices. Next, we describe a legal term correction method for Japanese statutory sentences. Our method predicts suitable legal terms using Random Forest classifiers. The classifiers in our method use adjacent words to a target legal term as input features, and are optimized in various parameters including the number of adjacent words to be used for each legal term set. We conduct an experiment using actual statutory sentences from 3,983 existing acts and cabinet orders that consist of approximately 47M words in total. As for legal term sets, we pick 27 sets from legislation drafting manuals. The experimental result shows that our method outperformed existing modern word prediction methods using neural language models and that each Random Forest classifier utilizes characteristics of its corresponding legal term set.
著者
古川 康一 植野 研 尾崎 知伸 神里 志穂子 川本 竜史 渋谷 恒司 白鳥 成彦 諏訪 正樹 曽我 真人 瀧 寛和 藤波 努 堀 聡 本村 陽一 森田 想平
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.2, pp.117-128, 2005 (Released:2005-02-04)
参考文献数
52
被引用文献数
4 4

Physical skills and language skills are both fundamental intelligent abilities of human being. In this paper, we focus our attention to such sophisticated physical skills as playing sports and playing instruments and introduce research activities aiming at elucidating and verbalizing them. This research area has been launched recently. We introduce approaches from physical modeling, measurements and data analysis, cognitive science and human interface. We also discuss such issues as skill acquisition and its support systems. Furthermore, we consider a fundamental issue of individual differences occurring in every application of skill elucidation. Finally we introduce several attempts of skill elucidation in the fields of dancing, manufacturing, playing string instruments, sports science and medical care.
著者
松林 達史 清武 寛 幸島 匡宏 戸田 浩之 田中 悠介 六藤 雄一 塩原 寿子 宮本 勝 清水 仁 大塚 琢馬 岩田 具治 澤田 宏 納谷 太 上田 修功
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.5, pp.wd-F_1-11, 2019-09-01 (Released:2019-09-01)
参考文献数
29

Forming security plans for crowd navigation is essential to ensure safety management at large-scale events. The Multi Agent Simulator (MAS) is widely used for preparing security plans that will guide responses to sudden and unexpected accidents at large events. For forming security plans, it is necessary that we simulate crowd behaviors which reflects the real world situations. However, the crowd behavior situations require the OD information (departure time, place of Origin, and Destination) of each agent. Moreover, from the viewpoint of protection of personal information, it is difficult to observe the whole trajectories of all pedestrians around the event area. Therefore, the OD information should be estimated from the several observed data which is counted the number of passed people at the fixed points.In this paper, we propose a new method for estimating the OD information which has following two features. Firstly, by using Bayesian optimization (BO) which is widely used to find optimal hyper parameters in the machine learning fields, the OD information are estimated efficiently. Secondly, by dividing the time window and considering the time delay due to observation points that are separated, we propose a more accurate objective function.We experiment the proposed method to the projection-mapping event (YOYOGI CANDLE 2020), and evaluate the reproduction of the people flow on MAS. We also show an example of the processing for making a guidance plan to reduce crowd congestion by using MAS.
著者
古崎 晃司 山縣 友紀 国府 裕子 今井 健 大江 和彦 溝口 理一郎
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.29, no.4, pp.396-405, 2014-07-01 (Released:2014-06-18)
参考文献数
23
被引用文献数
5 2

Publishing open data as linked data is a significant trend in not only the Semantic Web community but also other domains such as life science, government, media, geographic research and publication. One feature of linked data is the instance-centric approach, which assumes that considerable linked instances can result in valuable knowledge. In the context of linked data, ontologies offer a common vocabulary and schema for RDF graphs. However, from an ontological engineering viewpoint, some ontologies offer systematized knowledge, developed under close cooperation between domain experts and ontology engineers. Such ontologies could be a valuable knowledge base for advanced information systems. Although ontologies in RDF formats using OWL or RDF(S) can be published as linked data, it is not always convenient to use other applications because of the complicated graph structures. Consequently, this paper discusses RDF data models for publishing ontologies as linked data. As a case study, we focus on a disease ontology in which diseases are defined as causal chains.
著者
鳥海 不二夫 山本 仁志 諏訪 博彦 岡田 勇 和泉 潔 橋本 康弘
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.25, no.1, pp.78-89, 2010
被引用文献数
3 9

In recent years, application of Social Networking Services (SNS) and Blogs are growing as new communication tools on the Internet. Several large-scale SNS sites are prospering; meanwhile, many sites with relatively small scale are offering services. Such small-scale SNSs realize small-group isolated type of communication while neither mixi nor MySpace can do that. However, the studies on SNS are almost about particular large-scale SNSs and cannot analyze whether their results apply for general features or for special characteristics on the SNSs. From the point of view of comparison analysis on SNS, comparison with just several types of those cannot reach a statistically significant level. We analyze many SNS sites with the aim of classifying them by using some approaches. Our paper classifies 50,000 sites for small-scale SNSs and gives their features from the points of network structure, patterns of communication, and growth rate of SNS. The result of analysis for network structure shows that many SNS sites have small-world attribute with short path lengths and high coefficients of their cluster. Distribution of degrees of the SNS sites is close to power law. This result indicates the small-scale SNS sites raise the percentage of users with many friends than mixi. According to the analysis of their coefficients of assortativity, those SNS sites have negative values of assortativity, and that means users with high degree tend to connect users with small degree. Next, we analyze the patterns of user communication. A friend network of SNS is explicit while users' communication behaviors are defined as an implicit network. What kind of relationships do these networks have? To address this question, we obtain some characteristics of users' communication structure and activation patterns of users on the SNS sites. By using new indexes, friend aggregation rate and friend coverage rate, we show that SNS sites with high value of friend coverage rate activate diary postings and their comments. Besides, they become activated when hub users with high degree do not behave actively on the sites with high value of friend aggregation rate and high value of friend coverage rate. On the other hand, activation emerges when hub users behave actively on the sites with low value of friend aggregation rate and high value of friend coverage rate. Finally, we observe SNS sites which are increasing the number of users considerably, from the viewpoint of network structure, and extract characteristics of high growth SNS sites. As a result of discrimination on the basis of the decision tree analysis, we can recognize the high growth SNS sites with a high degree of accuracy. Besides, this approach suggests mixi and the other small-scale SNS sites have different character trait.
著者
古崎 晃司 溝口 理一郎
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.34, no.1, pp.C-I52_1-13, 2019-01-01 (Released:2019-01-07)
参考文献数
14
被引用文献数
2

Ontological considerations about part-of relations have been extensively investigated because they are basic and important relationships for ontology building. Although there are various discussions on kinds of part-of and their ontological characteristics, there remains some room for discussing a couple of fundamental issues such as “What is a part?” and “When is a part-of relation composed?” This paper discusses ontology patterns of descriptions of part-of relationships on the basis of ontological theories in order to provide practitioners with useful guidelines for descriptions of part-of structurers. This paper focuses on ontology patterns which capture commonality and special characteristics of parts so that complicated structures of physical objects are described appropriately. We discuss four problems related to descriptions of parts. 1) interdependence between the whole and its parts, 2) kinds of parts such as components, portions and materials, 3) multiple inheritance according to substance and properties of parts, 4) the commonality and specificity of parts. To cope with these problems, this paper introduces a part representation model based on ontological theory of roles. The main idea of the part representation model is to distinguish between a part dependent on its whole and the context-independent properties of the part. The former is defined as the role-holder which plays roles and the latter is defined as the player of the role. The role defines properties of the part which is dependent on its whole. These three kinds of definitions enable to describe differences of various properties of parts according to their context dependence. We show how this model is used to describe various parts through practical examples of the anatomical structure of human body developed in the medical ontology project in Japan.
著者
山本 賢太 井上 昂治 中村 静 高梨 克也 河原 達也
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.33, no.5, pp.C-I37_1-9, 2018-09-01 (Released:2018-09-03)
参考文献数
20

This paper addresses character expression for humanoid robots that play a given social role such as a lab guide or a counselor via spoken dialogue so that the character matches to the social role. While most conventional methods of character expression aim to change the style of utterance texts, this study focuses on dialogue features that may affect the impression of spoken dialogue. Specifically, we use five features: utterance amount, backchannel frequency, backchannel variety, filler frequency, and switching pause length. We adopt three character traits of extroversion, emotional instability, and politeness for a character expression, and investigate the relationship with the dialogue features. A statistical analysis of subjective evaluations shows that the dialogue features except for the backchannel variety are related to either of the traits. By using the subjective evaluation scores on the relevant traits, we can train models to control the dialogue features and behaviors according to the desired character. An experimental evaluation demonstrates the feasibility of character expression with regard to the traits of extroversion and politeness.

3 0 0 0 OA J-GLOBAL knowledge

著者
木村 考宏 川村 隆浩 渡邊 勝太郎 松本 尚也 佐藤 智宣 櫛田 達矢 松邑 勝治
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.N-F73_1-12, 2016-03-01 (Released:2016-03-31)
参考文献数
9
被引用文献数
1 1

In order to develop innovative solutions in science and technology, Japan Science and Technology Agency (JST) has built J-GLOBAL knowledge (JGk), which provides papers, patents, researchers' information, technological thesaurus, and scientific data as Linked Data, which have been accumulated by JST since 1957. The total size of all datasets is about 15.7 billion triples, and the JGk website provides a SPARQL endpoint to access part of the datasets. This paper describes several issues on schema design to construct a large-scale Linked Data, and construction methods, especially for linking to external datasets, such as DBpedia Japanese. Finally, we describe performance problems and the future works.
著者
則 のぞみ ボレガラ ダヌシカ 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.29, no.1, pp.168-176, 2014-01-05 (Released:2014-01-07)
参考文献数
24
被引用文献数
1 1

Many phenomena in the real world can be represented as multinomial relations, which involve multiple and heterogeneous objects. For instance, in social media, users' various actions such as adding annotations to web resources or sharing news with their friends can be represented by multinomial relations which involve multiple and heterogeneous objects such as users, documents, keywords and locations. Predicting multinomial relations would improve many fundamental applications in various domains such as online marketing, social media analyses and drug development. However, the high-dimensional property of such multinomial relations poses one fundamental challenge, that is, predicting multinomial relations with only a limited amount of data. In this paper, we propose a new multinomial relation prediction method, which is robust to data sparsity. We transform each instance of a multinomial relation into a set of binomial relations between the objects and the multinomial relation of the involved objects. We then apply an extension of a low-dimensional embedding technique to these binomial relations, which results in a generalized eigenvalue problem guaranteeing global optimal solutions. We also incorporate attribute information as side information to address the ``cold start"problem in multinomial relation prediction. Experiments with various real-world social web service datasets demonstrate that the proposed method is more robust against data sparseness as compared to several existing methods, which can only find sub-optimal solutions.
著者
松村 真宏 河原 大輔 岡本 雅史 黒橋 禎夫 西田 豊明
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.93-102, 2007 (Released:2007-01-05)
参考文献数
18
被引用文献数
1 3

To overcome the limitation of conventional text-mining approaches in which frequent patterns of word occurrences are to be extracted to understand obvious user needs, this paper proposes an approach to extracting questions behind messages to understand potential user needs. We first extract characteristic case frames by comparing the case frames constructed from target messages with the ones from 25M sentences in the Web and 20M sentences in newspaper articles of 20 years. Then we extract questions behind messages by transforming the characteristic case frames into interrogative sentences based on new information and old information, i.e., replacing new information with WH-question words. The proposed approach is, in other words, a kind of classification of word occurrence pattern. Qualitative evaluations of our preliminary experiments suggest that extracted questions show problem consciousness and alternative solutions -- all of which help to understand potential user needs.
著者
柴田 雅博 冨浦 洋一 西口 友美
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.24, no.6, pp.507-519, 2009 (Released:2009-09-04)
参考文献数
19
被引用文献数
5 7

We propose an open-ended dialog system that generates a proper sentence to a user's utterance using abundant documents on the World Wide Web as sources. Existing knowledge-based dialog systems give meaningful information to a user, but they are unsuitable for open-ended input. The system Eliza can handle open-ended input, but it gives no meaningful information. Our system lies between the above two dialog systems; it converses on various topics and gives meaningful information related to the user's utterances. The system selects an appropriate sentence as a response from documents gathered through the Web, on the basis of surface cohesion and shallow semantic coherence. The surface cohesion follows centering theory and the semantic coherence is calculated on the basis of the conditional distribution and inverse document frequency of content words (nouns, verbs, and adjectives.) We developed a trial system to converse about movies and experimentally found that the proposed method generated 66% appropriate responses.
著者
杉山 弘晃 目黒 豊美 東中 竜一郎 南 泰浩
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.183-194, 2015-01-06 (Released:2015-01-06)
参考文献数
18
被引用文献数
1 1

The development of open-domain conversational systems is difficult since user utterances are too flexible for such systems to respond properly. To address this flexibility, previous research on conversational systems has selected system utterances from web articles based on word-level similarity with user utterances; however, the generated utterances, which originally appeared in different contexts from the conversation, are likely to contain irrelevant information with respect to the input user utterance. To leverage the variety of web corpus in order to respond to the flexibility and suppress the irrelevant information simultaneously, we propose an approach that generates system utterances with two strongly related phrase pairs: one that composes the user utterance and another that has a dependency relation to the former. By retrieving the latter one from the web, our approach can generate system utterances that are related to the topics of user utterances. We examined the effectiveness of our approach with following two experiments. The first experiment, which examined the appropriateness of response utterances, showed that our proposed approach significantly outperformed other retrieval and rule-based approaches. The second one was a chat experiment with people, which showed that our approach demonstrated almost equal performance to a rule-based approach and outperformed other retrieval-based approaches.
著者
白松 俊 トッサヴァイネン テーム 大囿 忠親 新谷 虎松
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.LOD-C_1-11, 2016-01-06 (Released:2016-01-08)
参考文献数
25

To address social issues about the sustainability of local societies, inter-organizational collaboration in public sphere is important. Although conventional social networking services (SNSs) have recently been used for public collaboration, the SNSs are not suitable to look for potential collaborators because the conventional SNSs emphasize recency of information and lack a function for sharing information about ``who are trying to address what kind of social issues''. We designed a data model for structuring public issues and goals and built a linked open dataset (LOD) based on the above data model. Moreover, we developed a method for calculating similarities of public goals and implemented a Web service for matching public goals for finding potential collaborators. Our method for similarity calculation incorporates surficial features, semantic features, and contextual features. We conducted an experiment to investigate an optimal balance of parameters of the contextual features, which is suitable for facilitating public collaboration. Furthermore, we held a participatory event for trial use by citizens and observed positive feedbacks from the participants.
著者
杉山 貴昭 船越 孝太郎 中野 幹生 駒谷 和範
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.31, no.3, pp.C-FB2_1-9, 2016-05-01 (Released:2016-05-25)
参考文献数
18
被引用文献数
1

When a robot interacts with users in public spaces, it receives various sounds such as surrounding noises and users' voices. And furthermore, the robot needs to interact with multiple people at the same time. If the robot incorrectly determines whether it should respond to these sounds, it will erroneously respond to surrounding noises or ignore user utterances directed to the robot. In this paper, we present a machine learning-based method to estimate a response obligation, i.e., whether the robot should respond to an input sound. We address a problem setting that is more similar to interactions in public spaces than those assumed in previous studies. While previous studies assume only utterances directed to one of interlocutors as input sounds, we deal with not only those utterances but also noises and monologues. To deal with various sounds, our method uses the results of input sound classification and user behaviors both in an input sound interval and after the interval. In particular, the user behaviors after the interval are introduced as a key factor for improving the estimation accuracy of response obligation, such as a tendency that a user stands and keeps still after he/she talks to the robot. We demonstrate the new features significantly improved the estimation performance. We also investigate performances with various combinations of features and reveal that the results of input sound classification and the user behaviors after the interval are helpful for the estimation.
著者
秋葉 拓哉 林 孝紀 則 のぞみ 岩田 陽一
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
pp.B-F71, (Released:2015-10-27)
参考文献数
39

Estimating the relevance or proximity between vertices in a network is a fundamental building block of network analysis and is useful in a wide range of important applications such as network-aware searches and network structure prediction. In this paper, we (1) propose to use top-k shortest-path distance as a relevance measure, and (2) design an efficient indexing scheme for answering top-k distance queries. Although many indexing methods have been developed for standard (top-1) distance queries, no methods can be directly applied to top-k distance. Therefore, we develop a new framework for top-k distance queries based on 2-hop cover and then present an efficient indexing algorithm based on the recently proposed pruned landmark labeling scheme. The scalability, efficiency and robustness of our method are demonstrated in extensive experimental results. It can construct indices from large graphs comprising millions of vertices and tens of millions of edges within a reasonable running time. Having obtained the indices, we can compute the top-k distances within a few microseconds, six orders of magnitude faster than existing methods, which require a few seconds to compute these distances. Moreover, we demonstrate the usefulness of top-k distance as a relevance measure by applying them to link prediction, the most fundamental problem in graph data mining. We emphasize that the proposed indexing method enables the first use of top-k distance for such tasks.
著者
山田 広明 橋本 敬
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.491-497, 2015-03-01 (Released:2015-03-17)
参考文献数
11
被引用文献数
2

In order to investigate the formation mechanism of community activity, we constructed an agent-based model based on a scenario driven by subjective norm and self-efficacy utilizing a community task game. The model demonstrated the spontaneous formation of community activity. The formation and expansion were driven by two mechanisms: (1) self-efficacy maintained participation of agents having a neutral attitude towards community activity, (2) subjective norm caused an increase in participation by involving other adjacent neutral attitude agents. We suggest a reasonable strategy promoting the spontaneous formation of community activity on the basis of this mechanism.
著者
小島 諒介 佐藤 泰介
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.29, no.3, pp.301-310, 2014-05-01 (Released:2014-04-04)
参考文献数
23

In this paper we propose a new plan recognition method from observations of incomplete action sequences by regarding them as prefixes in a probabilistic context-free grammar (PCFG). In previous work that uses a PCFG for plan recognition, the PCFG receives a sentence, i.e. an observation of complete action sequences to recognize the plan behind it. However, when we deal with real plan recognition problems such as the Web access log analysis, we often cannot have complete sequences of actions and the traditional PCFG approach is not applicable. To overcome this difficulty, we extend the probability computation of PCFGs to prefix probability computation though it requires an infinite sum of probabilities. We applied the proposed method to infer the intended goals of Web site visitors from the online and partial observations of their actions. Also we compared the performance of plan recognition from observations of initial sequences of visitors' actions with that from full observations.