• Two-step approach (Named Entity Recognition + Rules) performed better than a direct code generation approach for natural language to PROLEG translation. • Regarding Large Language Models (LLMs), code LLMs (designed for code generation) did not show better results than general-purpose LLMs. • Llama 3 70B obtained better results for direct code generation than other models, including Llama 3.1 70B. • A NER library (GliNER) showed better performance than state-of-the-art Large Language Models, among which Llama 3.1 70B achieves the best results. This study presents a comparative analysis of methodologies for automating the extraction of legal information from contract texts and expressing it as PROLEG logical clauses, focusing on two approaches: Direct Code Generation (DCG) via Large Language Model (LLM) prompting, and Named Entity Recognition (NER) with rule-based transformation. For the NER-based approach, we evaluated three implementations: (1) LLM prompting for entity recognition, (2) a fine-tuned LLM for NER, and (3) a specialized NER library. Our findings reveal that while LLMs demonstrate versatility in general NLP tasks, the highest precision and contextual adaptation were achieved using a NER library specifically tailored to handle named entities. For finding named entities relevant to the contract (including, for instance, the buyer or the rescission date of a contract), this approach outperformed both DCG and NER using LLMs. The results underscore the importance of tailored lexical resources and rule-based post-processing in legal NLP applications, suggesting a paradigm for optimizing automated contract analysis systems, as opposed to the current trend of using general-purpose LLMs for most NLP tasks.

Knowledge graphs (KGs) are powerful tools for representing and reasoning over structured information. Their main components include schema, identity, and context. While schema and identity matching are well-established in ontology and entity matching research, context matching remains largely unexplored. This is particularly important because real-world KGs often vary significantly in source, size, and information density—factors not typically represented in the datasets on which current entity matching methods are evaluated. As a result, existing approaches may fall short in scenarios where diverse and complex contexts need to be integrated. To address this gap, we propose a novel KG integration method consisting of label matching and triple matching. We use string manipulation, fuzzy matching, and vector similarity techniques to align entity and predicate labels. Next, we identify mappings between triples that convey comparable information, using these mappings to improve entity-matching accuracy. Our approach demonstrates competitive performance compared with leading systems in the OAEI competition and against supervised methods, achieving high accuracy across diverse test cases. Additionally, we introduce a new dataset derived from the benchmark dataset to evaluate the triple-matching step more comprehensively.

This paper presents an outline of our research project developed at our research center for "Juris-Informatics". "Juris-Informatics" is a research field based on two main topics; "Law by AI" and "Law of AI". "Law by Ai" is a research field where we investigate a support tool by AI for legal activities such as legal reasoning and legal document processing. "Law of AI" is a research field where we conduct research on legal control of AI such as considering the legal responsibility of AI and legal compliance of AI.

Most of the previous methods for table recognition rely on training datasets containing many richly annotated table images. Detailed table image annotation, e.g., cell or text bounding box annotation, however, is costly and often subjective. In this paper, we propose a weakly supervised model named WSTabNet for table recognition that relies only on HTML (or LaTeX) code-level annotations of table images. The proposed model consists of three main parts: an encoder for feature extraction, a structure decoder for generating table structure, and a cell decoder for predicting the content of each cell in the table. Our system is trained end-to-end by stochastic gradient descent algorithms, requiring only table images and their ground-truth HTML (or LaTeX) representations. To facilitate table recognition with deep learning, we create and release WikiTableSet, the largest publicly available image-based table recognition dataset built from Wikipedia. WikiTableSet contains nearly 4 million English table images, 590K Japanese table images, and 640k French table images with corresponding HTML representation and cell bounding boxes. The extensive experiments on WikiTableSet and two large-scale datasets: FinTabNet and PubTabNet demonstrate that the proposed weakly supervised model achieves better, or similar accuracies compared to the state-of-the-art models on all benchmark datasets.

To exhaustively understand the impact of rule amendments and unforeseen cases on existing norms, it requires connecting their rule-based and case-based representations. However, those connections have not been explored in depth, especially for norms that are represented as soft constraints. This paper aims to explore the connection between constraint hierarchies and case models as representative formalisms of rule-based and case-based representations of soft-constraint norms respectively. To explore the connection, we express norm scopes and preferences in both formalisms as diagrams. Based on tightening and arranging diagrams, we found the translation of constraint hierarchies with one constraint per level into case models. This provides new insights into understanding prototypical cases made by rule-based soft-constraint norms.

Wikidata is the largest collaborative general knowledge graph supported by a worldwide community. It includes many helpful topics for knowledge exploration and data science applications. However, due to the enormous size of Wikidata, it is challenging to retrieve a large amount of data with millions of results, make complex queries requiring large aggregation operations, or access too many statement references. This paper introduces our preliminary works on Wikidata-lite, a toolkit to build a database offline for knowledge extraction and exploration, e.g., retrieving item information, statements, provenances, or searching entities by their keywords, attributes. Wikidata-lite has high performance and memory efficiency, much faster than the official Wikidata SPARQL endpoint for big queries. The Wikidata-lite repository is available at https://github.com/phucty/wikidb.

Thanks to the open government initiative movement, many base registries have been made available and beneficial for all citizens to ensure accountability and transparency. All data are easy to access and process by people for effective, efficient public oversight. However, most base registries publish their data in a different data model or structure. As a result, building an AI system to utilize these data is challenging since they are machine-unreadable, poorly managed, and scattered. This paper proposes a design method for base registry data structures that uses unification and federation to improve the accessibility and availability of data. We also use Wikibase as an underlying knowledge graph to inherit common sense knowledge available in Wikidata to improve its usability.

Entity representations are useful in natural language tasks involving entities.In this paper, we propose new pretrained contextualized representations of words and entities based on the bidirectional transformer (Vaswani et al., 2017).The proposed model treats words and entities in a given text as independent tokens, and outputs contextualized representations of them.Our model is trained using a new pretraining task based on the masked language model of BERT (Devlin et al., 2019).The task involves predicting randomly masked words and entities in a large entity-annotated corpus retrieved from Wikipedia.We also propose an entity-aware self-attention mechanism that is an extension of the self-attention mechanism of the transformer, and considers the types of tokens (words or entities) when computing attention scores.The proposed model achieves impressive empirical performance on a wide range of entity-related tasks.In particular, it obtains state-of-the-art results on five well-known datasets: Open Entity (entity typing), TACRED (relation classification), CoNLL-2003 (named entity recognition), ReCoRD (cloze-style question answering), and SQuAD 1.1 (extractive question answering).Our source code and pretrained representations are available at https: //github.com/studio-ousia/luke.

We describe a neural network model that jointly learns distributed representations of texts and knowledge base (KB) entities. Given a text in the KB, we train our proposed model to predict entities that are relevant to the text. Our model is designed to be generic with the ability to address various NLP tasks with ease. We train the model using a large corpus of texts and their entity annotations extracted from Wikipedia. We evaluated the model on three important NLP tasks (i.e., sentence textual similarity, entity linking, and factoid question answering) involving both unsupervised and supervised settings. As a result, we achieved state-of-the-art results on all three of these tasks. Our code and trained models are publicly available for further academic research.

Named Entity Disambiguation (NED) refers to the task of resolving multiple named entity mentions in a document to their correct references in a knowledge base (KB) (e.g., Wikipedia).In this paper, we propose a novel embedding method specifically designed for NED.The proposed method jointly maps words and entities into the same continuous vector space.We extend the skip-gram model by using two models.The KB graph model learns the relatedness of entities using the link structure of the KB, whereas the anchor context model aims to align vectors such that similar words and entities occur close to one another in the vector space by leveraging KB anchors and their context words.By combining contexts based on the proposed embedding with standard NED features, we achieved state-of-theart accuracy of 93.1% on the standard CoNLL dataset and 85.2% on the TAC 2010 dataset.

By decoupling the network function from the underlying dedicated hardware, enterprises can reap the benefit of outsourcing network functions to the cloud. Due to a lack of analytical tools, one cannot simply justify the worthiness of outsourcing network functions. This paper proposes a decision support framework for network function outsourcing, called Location-aware Network Function Outsourcing (LawNFO), aiming to identify which network function is worth to be outsourced to the cloud. LawNFO constructs a network-function graph and maps the cost of network function processing units and the data transmission between any network-function pairs onto the graph. The network function outsourcing problem is formulated as an optimization problem targeting on the minimum network cost. To solve the optimization problem based on the graph input, we develop an efficient and fast heuristic algorithm, called Node-weighted Contraction (NowCont), that encompasses the graph partitioning and graph augmentation techniques. This paper performs simulation to evaluate the performance of LawNFO with NowCont algorithm in two aspects: 1) the closeness ratio to the optimal solutions and 2) the network cost comparing to that of all-in-house and most-in-cloud network function placement models. The simulation results illustrate that the NowCont algorithm yields the solution closest to the optimal solution. The evaluation through two use cases show that LawNFO is of benefit to the enterprises for the decision of network function outsourcing.

The structure of scientific co-authorship networks can reveal patterns of scientific collaboration. Our study focused on developing a methodology for community finding in co-authorship networks that could reveal the real, i.e., physical, patterns of scientific collaboration, by using two datasets with different collaboration scopes: one considering only collaborations within Japanese research institutions from different research areas, and the other comprising worldwide collaborations in the area of Computer Science.

This paper presents the study of pattern of interaction in research work (co-authorship in publications and co-partnership in research projects) using the techniques of social network analysis (SNA) and visualization. Along with computation of various network features at local and global levels to find the nature of collaboration, and important positions of researchers in collaboration network, we intersect the networks from co-authorship in publications and co-partnership in research projects with an attempt to find the important researchers, both with regard to collaboration in paper and projects. Moreover, we show the difference in nature of collaboration regarding publications and projects.

Recommendation systems have extracted items that users may be interested in. However, most of the recommendation applications have restricted on recommending only items in a specific domain (e.g., movies, books, and musics). In this paper, we propose a novel approach to enable the existing recommendation systems to extract addition items in various other domains. Thereby, this work is focusing on integrating all available Linked Open Data (LOD) for selecting more relevant attributes which can improve the performance of recommendation processes.

Wikipedia is the user contributed encyclopedia edited collaboratively by a wide-range of people. Wikipedia usually determines contents of article and editorial policies through discussion among participants. It requires a lot of effort in deducing its conclusion often due to protracted discussion. We need some measurement to examine how discussion is valuable to reach conclusion. So, we call it discussion validity and define it with the model with discussion capability of participants. Discussion capability of participants consists of three features each of which represents characteristic aspect of users' behavior in discussion, and approximated from the corresponding three features of their utterances. We conducted the experiments with the subjects to verify the model and found that our model resulted better than the conventional model and proposed the automatic prediction of discussion validity using the text analysis. Then we estimated discussion validity through the model with the estimated values. It turned out that the estimation was well fitted with the values by the subjects.

We propose a novel style of social tagging to construct knowledge collaboratively called Social Property Tagging and introduce the prototype system Social Infobox. Structured data is useful for computer system, however defining structure of knowledge for representing data semantics is usually a costly and time consuming task. In general, data structures are constructed by experts of knowledge engineering. Our method aims to construct not only structured data but also structure of data collaboratively by simple user input.

User-generated content can help the growth of linked data. However, we lack interfaces enabling ordinary people to author linked data. Secondly, people have multiple perspectives on the same concept and different contexts. Thirdly, not enough ontologies exist to model various data. Therefore, we propose an approach to enable people to share various data through an easy-to-use social platform. Users define their own concepts and multiple conceptualizations are allowed. These are consolidated using semi-automatic schema alignment techniques supported by the community. Further, concepts are grouped semi-automatically by similarity. As a result of consolidation and grouping, informal lightweight ontologies emerge gradually. We have implemented social software, called StYLiD, to realize our approach. It can serve as a platform motivating people to bookmark and share different things. It may also drive vertical portals for specific communities with integrated data from multiple sources. Experimental observations support the validity of our approach.

Recently, designers have been required solving comprehensive problem becoming greater and more complicated. In relation to this background, we have proposed Universal Abduction Studio (UAS); a computer environment that synthetically supports creative design. However, it is difficult for a designer to manually acquire multiple domain knowledge required for UAS. Therefore, we propose a Web-based knowledge database construction method for supporting design by UAS in this paper.

Social aspects are critical in the decision making process for social actors (human beings). Social aspects can be categorized into social interaction, social communities, social groups or any kind of behavior that emerges from interlinking, overlapping or similarities between interests of a society. These social aspects are dynamic and emergent. Therefore, interlinking them in a social structure, based on bipartite affiliation network, may result in isolated graphs. The major reason is that as these correspondences are dynamic and emergent, they should be coupled with more than a single affiliation in order to sustain the interconnections during interest evolutions. In this paper we propose to interlink actors using multiple tripartite graphs rather than a bipartite graph which was the focus of most of the previous social network building techniques. The utmost benefit of using tripartite graphs is that we can have multiple and hierarchical links between social actors. Therefore in this paper we discuss the extraction, plotting and analysis methods of tripartite relations between authors, articles and categories from Wikipedia. Furthermore, we also discuss the advantages of tripartite relationships over bipartite relationships. As a conclusion of this study we argue based on our results that to build useful, robust and dynamic social networks, actors should be interlinked in one or more tripartite networks.

As information technology progress, we are able to obtain much information about the advanced research of others. As a result, researchers and research managers need to track the current research trends amid the information flood. In order to support these efforts to gather knowledge of current research, we propose a research trend mining method. The method utilizes an author-topic model for establishing the relationships between authors, topics, and papers by probabilities, and interactively visualizes the relationships using self-organizing maps. We implemented a research area mapping system and validated it with a case study. In addition, we conducted experiments to show the performance of our system. The experimental results indicate that this system can induce the appropriate relationships for finding research trends.

Since research trends can change dynamically, researchers have to keep up with these new trends and undertake new research topics. Therefore, research communities for new research domains are important. In this paper, we propose a method to discover research communities. The key features of our method are a network model of papers and a word assignment technique for the communities obtained. We show our system based on the proposed method and discuss our system through case studies and experiments

We set out to reveal that arrangement of embodied agents' body orientation can influence the perception of body schema and the thought of users. Embodied agents are one type of media that can socially appeal to the user's intuitive thought, especially through their body expression. We focused on embodied agents' body orientation as a means to induce the user to perceive his/her body schema inside virtual space and to accept the perceived thought of an embodied agent. A psychological experiment was conducted, since arrangement of body orientation between two people or between a user and an embodied agent often influences the social relationship between them. The result of the experiment suggested that an embodied agent whose thought was different from the user's at first and whose body orientation corresponded with the user's could trigger the user's consideration of its opinions more strongly than when this agent showed the side of its body to the user. However, difference in the perception of the body schema in virtual space by the agent's body orientation arrangement was not observed. We investigated the relation between body orientation of embodied agents and change in user's perspective, and suggested a design principle embodied agents' body orientation for enhanced association between the agent and the user

A social network can become bases for information infrastructure in the future. It is important to extract social networks that are not biased. Providing a simple means for users to register their social relation is also important. We propose a method that combines various approaches to extract social networks. Especially, three kinds of networks are extracted; user-registered Know link network, Web-mined Web link network, and face-to-face Touch link network. In this paper, the combination of social network extraction for communities is described, and the analysis on the extracted social networks is shown.

Social networks play important roles in the Semantic Web: knowledge management, information retrieval, ubiquitous computing, and so on. We propose a social network extraction system called POLYPHONET, which employs several advanced techniques to extract relations of persons, detect groups of persons, and obtain keywords for a person. Search engines, especially Google, are used to measure co-occurrence of information and obtain Web documents.Several studies have used search engines to extract social networks from the Web, but our research advances the following points: First, we reduce the related methods into simple pseudocodes using Google so that we can build up integrated systems. Second, we develop several new algorithms for social networking mining such as those to classify relations into categories, to make extraction scalable, and to obtain and utilize person-to-word relations. Third, every module is implemented in POLYPHONET, which has been used at four academic conferences, each with more than 500 participants. We overview that system. Finally, a novel architecture called Super Social Network Mining is proposed; it utilizes simple modules using Google and is characterized by scalability and Relate-Identify processes: Identification of each entity and extraction of relations are repeated to obtain a more precise social network.

Research communities are very important for researchers undertaking new research topics. In this paper, we propose a community mining system using bibliography data in order to find communities of researchers. The basic concept of our study is to provide interactive visualization of both local and global research communities. We implement this concept using actual bibliography data and present a case study using the proposed system.

We propose a collaborative approach for personal task management which is modeled as an integration of alliance and human-in-the-loop model. Alliance model is based on information sharing and collaboration of several persons. They disclose their task condition and maintain to be updatable by their friends. To avoid privacy issues we propose emergent group discovery algorithm to control the level of disclosure. Human-in-the-loop model consists of three subsystems to support decision-making activities. Visualizer indicates the attributes associated with each task such as the deadline, the subjective priority, and the workload, which are determined by the user. Optimizer generates executable schedules from these tasks by active scheduler and multiobjective genetic algorithm. Recommender evaluates these alternatives by analytic hierarchy process. We implement client/server system called social scheduler on cell-phones environment. We remark the advantages of our approach with an experiment.

A method for behavior acquisition that considers embodiment by associating tactile information to visual input is described. An agent that acts in the physical world always suffers constraints derived from embodiment. On the other hand, embodiment plays a very important role in the formation of visual function. Philosophical and clinical medicine findings assert that vision does not function without learning through experiences of haptic motion. In this paper, we discuss the relation between vision, embodiment, and behavior. We develop a method for behavior acquisition through associating vision and tactile sensors. We perform experiments of obstacle avoidance using a computer simulation and a real agent to test the validity of our method.

In this paper, we are concerned with the problem of how a physical robot can get an appropriate internal representation to its task and environment. Learning from experience is effective for the problem, but it is very time-consuming to learn a representation from the beginning in a real environment. On the other hand, the representation learned only in a simulated environment has the risk of not serving the purpose in a real environment because of the uncertainty in sensors, actuators, and the environment. In, order to have the best of both worlds, it is effective to transplant the learned state representation of a virtual agent to a physical robot. For this purpose., we improved our developed incremental learning architecture for use in the real environment and developed a new architecture, called STNS-R. In this architecture, inappropriate negative instances caused by uncertainties are found on the basis of the distribution of instances and removed in order to correct the distorted shapes of the states. The effectiveness of STNS-R is shown in the experimental results.

We discuss roles of robots as autonomous knowledge media and show our prototype system of an office work assistant robot based on this approach. We are surrounded by a huge amount of artifacts and information that is difficult for humans to deal with. Robots can help people by gathering and arranging information intelligently instead of people themselves. Our prototype system called Kappa III is an office work assistant robot that can tell people the location of daily goods in the office. It firstly looks around to capture images of desks in order to remember what and where such goods are. Then it can identify them by cutting them out from the background and categorizing them by color, shape, and figure. People can ask it to take goods in the office by either specifying names or features such as color. We also realized discovery of objects three-dimensionally by comparing captured scenes with expected ones.

We propose a new architecture for recognizing objects based on a concept of "embodiment" as one of primitive functions for a cognitive robot. We define the term "embodiment" as the size and shape of the agent's body, locomotive ability and its sensor. According to embodiment, an object is represented by reaching action paths, which correspond to a set of sequences of movements taken by the agent for reaching the object. Visual information is used to obtain sensorimotor mapping which represents the relationship between the change of object's appearance and the movement of the agent. On the other hands, tactile information is utilized to evaluate the change of physical condition of the object caused by such movement. By means of this method, the agent can recognize an object without depending on its position and orientation in the environment. The experimental result of computer simulation is shown to validate the method.

This paper proposes a method for acquiring and utilizing our personal knowledge in computer systems. We developed a system called MindHeap that helps us to acquire personal knowledge by browsing WWW (World Wide Web) hypertexts. Our approach for acquiring a user's personal knowledge is to use situational context which has important effects for acquiring and reminding knowledge. Situational context means information which characterizes her/his situation of information activities. MindHeap can help the user to organize knowledge on WWW pages by finding and recording related words as situational context for topics specified by the user. The system extracts a list of related words as situational context, and records it with the topics. Thus, topics and their contexts are accumulated, then they can work as knowledge for the user, e.g., she/he can retrieve pages of which situation is similar to the current page.

Reinforcement learning is very useful for robots with little a priori knowledge in acquiring appropriate behavior. This paper describes a learning system which can learn a state representation and a behavior policy simultaneously while executing the task. We call the system - the situation transition network system. As cognitive learning, it extracts "situations" and maintains them dynamically in the continuous state space on the basis of rewards from the environment. As behavior learning, it leads to a Markov decision model of environment and performs partial planning on the model. This is a kind of reinforcement learning. The results of computer simulations are given.

To promote robotic soccer research, we need a low cost and portable robot with some sensors and a communication device. To date, there is no platform for robotic soccer. Therefore, each research must build his own robots or utilize robots which are commercially available. This paper describes how to construct a robot system which includes a lightweight and low-cost mobile robot with visual, tactile sensors, TCP/IP communication device, and portable PC where Linux is running. An example of the developed soccer robot system and preliminary experimental results are also shown.

We discuss the requirements for information sharing in a community. We propose the weak information structures to integrate heterogeneous information such as static information (e.g. local sites information) and dynamic information created in word-of-mouth communication. We have developed a system using the weak information structures for community information sharing called InfoCommon and evaluated it at the ICMAS'96 Mobile Assistant Project.

The long-term goal of the Knowledgeable Community project is to develop a computational framework for facilitating the knowledge process by humans and computers. We model the knowledgeable community as an open system of natural and artificial agents interacting in knowledgeable environments where real or virtual entities are associated with conceptualization. First, we study associative representation based on information representation with variable degrees of formality which permits raw information materials to be incrementally formalized into a well-defined system of knowledge and even incomplete knowledge pieces to be accessed and utilized by humans and computers. Second, we present the aspect theory for describing the world from multiple perspectives and its application to ontology-based intelligent document base system for human knowledge sharing. Third, we describe a computational framework using agent development, ontology-grounding, and ontology-based mediation. We have implemented several complete prototypes which can demonstrate the significance of our idea. Experimental evaluation is in progress.

Engineers have to refer to various technical information to perform their job. Technical information have been accumulated in an electronic form as documents which contain texts, pictures and movies in enterprises. However, it needs expert knowledge to make use of these technical information. In this paper, we describe an ontology-centric framework, called intelligent corporate base (ICoB) which organizes a technical information into documents, assists a comprehension of documents and helps share information. We explain a prototype system for planning on repair work of transformers. Furthermore, we describe development and use of ontology for technical information sharing.

An agent must acquire internal representation appropriate for its task, environment, and sensors. As a learning algorithm, reinforcement learning is often utilized to acquire the relation between sensory input and action. Learning agents in the real world using visual sensors are often confronted with the critical problem of how to build a necessary and sufficient state space for the agent to execute the task. We propose the acquisition of a relation between vision and action using the visual state-action map (VSAM). VSAM is the application of a self-organizing map (SOM). Input image data is mapped on the node of the learned VSAM. Then VSAM outputs the appropriate action for the state. We applied VSAM to a real robot. The experimental result shows that a real robot avoids the wall while moving around the environment.

In this paper, we discuss how ontology plays roles in building a distributed and heterogeneous knowledge-base system. First, we discuss relationship between ontology and agents in the [Formula: see text] which is a framework of knowledge sharing and reuse based on a multi-agent architecture. Ontology is a minimum requirement for each agent to join the [Formula: see text]. Second, we explain mediation by ontology to show how ontology is used in the [Formula: see text]. A special agent called mediator analyzes undirected messages and infer candidates of recipient agents by consulting ontology and relationship between ontology and agents. Third, we model ontology as combination of aspects each of which can represent a way of conceptualization. Aspects are combined either as combination aspect which means integration of aspects or category aspect which means choice of aspects. Since ontology by aspect allows heterogeneous and multiple descriptions for phenomenon in the world, it is appropriate for heterogeneous knowledge-base systems. We also show translation of messages as a way of interpreting multiple aspects. A translation agent can translate a message with some aspect to one with another aspect by analyzing dependency of aspects. Mediation and translation of messages are important to build agents easily and naturally because less knowledge on other agents is requested for each agent.