受賞区分：国際学会・会議・シンポジウム等の賞  受賞国：ベトナム社会主義共和国

受賞区分：国内学会・会議・シンポジウム等の賞  受賞国：日本国

受賞区分：国内学会・会議・シンポジウム等の賞 

受賞区分：国内学会・会議・シンポジウム等の賞  受賞国：日本国

受賞区分：国内学会・会議・シンポジウム等の賞 

受賞区分：国内学会・会議・シンポジウム等の賞  受賞国：日本国

受賞区分：国内学会・会議・シンポジウム等の賞 

受賞区分：国内学会・会議・シンポジウム等の賞 

受賞区分：国内学会・会議・シンポジウム等の賞 

受賞区分：国内学会・会議・シンポジウム等の賞  受賞国：日本国

受賞区分：国内学会・会議・シンポジウム等の賞  受賞国：日本国

受賞区分：国内学会・会議・シンポジウム等の賞  受賞国：日本国

受賞区分：国際学会・会議・シンポジウム等の賞  受賞国：オランダ王国

受賞区分：国際学会・会議・シンポジウム等の賞  受賞国：オランダ王国

受賞区分：国内学会・会議・シンポジウム等の賞 

受賞区分：国内学会・会議・シンポジウム等の賞  受賞国：日本国

受賞区分：国内学会・会議・シンポジウム等の賞  受賞国：日本国

受賞区分：国内学会・会議・シンポジウム等の賞  受賞国：日本国

受賞区分：国際学会・会議・シンポジウム等の賞  受賞国：ベルギー王国

受賞区分：国内学会・会議・シンポジウム等の賞  受賞国：日本国

受賞区分：国内学会・会議・シンポジウム等の賞  受賞国：日本国

受賞区分：国内学会・会議・シンポジウム等の賞  受賞国：日本国

出版者・発行元：NarSUM 2023 - Proceedings of the 2nd Workshop on User-centric Narrative Summarization of Long Videos, Co-located with: MM 2023  

Scopus

出版者・発行元：MM 2023 - Proceedings of the 31st ACM International Conference on Multimedia  

With video capture devices becoming widely popular, the amount of video data generated per day has seen a rapid increase over the past few years. Browsing through hours of video data to retrieve useful information is a tedious and boring task. Video Summarization technology has played a crucial role in addressing this issue. It is a well-researched topic in the multimedia community. However, the focus so far has been limited to creating summary to videos which are short (only a few minutes). This workshop aims to call for researchers on relevant background to focus on novel solutions for user-centric narrative summarization of long videos. This workshop will also cover important aspects of video summarization research like what is "important"in a video, how to evaluate the goodness of a created summary, open challenges in video summarization, etc.

DOI： 10.1145/3581783.3610946

Scopus

掲載種別：研究論文（学術雑誌）   出版者・発行元：Informatica (Slovenia)  

The deep learning-based text classification methods perform better than traditional ones. In addition to the success of the deep learning technique, multi-Task learning (MTL) has come to become a promising approach for text classification; for instance, an MTL approach in text classification employs named entity recognition as an auxiliary task and has showcased that the task helps to improve the text classification performance. Existing MTL-based text classification methods depend on the auxiliary tasks using supervised labels. Obtaining such supervision labels requires additional human and financial costs in addition to those for the main text classification task. To reduce these additional costs, we propose an MTL-based text classification framework on supervised label creation by automatically labeling phrases in texts for the auxiliary recognition task. A basic idea to realize the proposed framework is to utilize phrasal expressions consisting of subwords (called subword-phrases). To the best of our knowledge, no text classification approach has been designed on top of subword-phrases because subwords only sometimes express a coherent set of meanings. The novelty of the proposed framework is in adding subword-phrase recognition as an auxiliary task and utilizing subword-phrases for text classification. It extracts subword-phrases in an unsupervised manner using the statistics approach. To construct labels for effective subword-phrase recognition tasks, extracted subword-phrases are classified based on document classes to ensure that subword-phrases dedicated to some classes can be distinguishable. Experimental evaluation for text classification using five popular datasets showcased the effectiveness of the subword-phrase recognition as an auxiliary task. It also showed that comparing various labeling schemes in recent studies indicated insights for labeling common subword-phrases among several document classes.

DOI： 10.31449/inf.v47i3.4742

Scopus

出版者・発行元：Multimedia Tools and Applications  

Sound symbolism is a well-researched topic of psycholinguistics, which tries to comprehend the connection between the sound of a word and its meanings. The Bouba-Kiki effect, one form of sound symbolism, claims that people perceive the pronunciation of “Kiki” as pointier than that of “Bouba.” There is no research that focuses on modeling such perception, i.e., how pointy a pronunciation sounds to humans, through computational and data-driven approaches. To address this, this paper first proposes the novel concept of “phonetic pointiness” defined as how pointy a shape humans are most likely to associate with a given pronunciation. We then model this phonetic pointiness from computational and data-driven approaches to calculate a score for an arbitrary pronunciation. There are three proposed models: a referential model, an expressive model, and a combined model, which integrates the previous two. The idea comes from an existing psycholinguistic classification of two types of sound symbolisms: referential symbolism and expressive symbolism, where the former relates to vocabulary knowledge, while the latter is based on pure human intuition. The proposed models are constructed only with image and language data available on the Web, therefore not requiring task-specific human annotations. We evaluate these models through a crowd-sourced user study, finding a promising correlation between human perception and the phonetic pointiness calculated by the proposed models. The results indicate that human perception can be modeled better by combining both types of sound symbolisms. Furthermore, by observing the behaviors of the models, we show several possible use-cases, such as product naming and psycholinguistic research, which can be a useful insight to further studies and applications.

DOI： 10.1007/s11042-023-15732-z

Web of Science

Scopus

記述言語：英語   出版者・発行元：The Institute of Electronics, Information and Communication Engineers  

This paper proposes a method for estimating impressions received from images according to the personal attributes of users, so that they can find the desired images based on their tastes. A previous study taking into account gender and age as personal attributes showed promising results. However, it also showed that users sharing the same gender and age do not necessarily share similar tastes. Therefore, other attributes should be considered to well capture users' personal tastes. However, taking more attributes into account leads to a problem that insufficient amounts of data are served to classifiers, due to explosion of the number of combinations of attributes. To tackle this problem, we propose an aggregation-based method to condense training data for impression estimation while personal attribute information is taken into account. For evaluation, a dataset of 4,000 carpet images annotated with 24 impression words by crowd-workers was prepared, which contained 273k annotations. Experimental results showed that the use of combinations of personal attributes improved the accuracy of impression estimation. This indicates that combinations of personal attributes are helpful to estimate impressions of individual viewers to images.

DOI： 10.34385/proc.78.p1-14

CiNii Research

記述言語：英語   出版者・発行元：The Institute of Electronics, Information and Communication Engineers  

Agriculture is the pillar industry of human sur- vival. However, various crop diseases reduce the hu- man food supply and lead to starvation and death in the worst cases. Experts perform visual symptoms ob- servation for crop disease diagnosis. Which process is time-consuming and expensive. Also, the process has significant risk of human error due to subjective per- ception. Convolutional Neural Networks (CNN) use image processing techniques to show great potential in plant disease detection. However, it requires thou- sands of channels to learn rich features, resulting in large models requiring powerful computing, power sup- ply, and high bandwidth, making it more expensive and difficult for farmers to acquire. Therefore, deploying these solutions on resource-constrained devices is de- sirable to make them more accessible. Thus, we pro- pose a lightweight object detection CNN that can run on resource-constrained devices to detect crop diseases. Channel pruning is applied to optimize resource use by removing unimportant channels and filter weights to reduce network parameters, inference time, and the number of FLOPS. Experimental results with object de- tector, Faster R-CNN with two backbones, ResNet-50, and EfficientNet-B7, show significant improvement in model efficiency, keeping high accuracy.

DOI： 10.34385/proc.78.p1-23

CiNii Research

記述言語：英語   出版者・発行元：The Institute of Electronics, Information and Communication Engineers  

Object detection is the task of detecting objects in an image. In this task, the detection of small objects is particularly difficult. Other than the small size, it is also accompanied by difficulties due to blur, occlusion, and so on. Current small object detection methods are tailored to small and dense situations, such as pedestrians in a crowd or far objects in remote sensing scenarios. However, when the target object is small and sparse, there is a lack of objects available for training, making it more difficult to learn effective features. In this paper, we propose a specialized method for detecting a specific category of small objects; birds. Particularly, we improve the features learned by the neck; the sub-network between the backbone and the prediction head, to learn more effective features with a hierarchical design. We employ Swin Transformer to upsample the image features. Moreover, we change the shifted window size for adapting to small objects. Experiments show that the proposed Swin Transformerbased neck combined with CenterNet can lead to good performance by changing the window sizes. We further find that smaller window sizes (default 2) benefit mAPs for small object detection.

DOI： 10.34385/proc.78.te-3

CiNii Research

記述言語：英語   出版者・発行元：The Institute of Electronics, Information and Communication Engineers  

Small Object Detection (SOD) is an important machine vision topic because (i) a variety of real-world applications require object detection for distant objects and (ii) SOD is a challenging task due to the noisy, blurred, and less-informative image appearances of small objects. This paper proposes a new SOD dataset consisting of 39,070 images including 137,121 bird instances, which is called the Small Object Detection for Spotting Birds (SOD4SB) dataset. The detail of the challenge with the SOD4SB dataset is introduced in this paper. In total, 223 participants joined this challenge. This paper briefly introduces the awardwinning methods. The dataset, the baseline code, and the website for evaluation on the public testset are publicly available.

DOI： 10.34385/proc.78.te-1

CiNii Research

記述言語：日本語   出版者・発行元：デジタルアーカイブ学会  

国家・社会制度に関する政策は、法令を通して制度化されるため、日本社会の動きは、法令情報を介して捉えることができる。本研究は、制定・改正などを通じた法令の連続的変遷を把握し、日本の国家・社会運営の長期的変化を調査するための研究基盤の確立を目指すものである。その最初の目標として、明治以降の全法令を検索可能なオープンデータベースシステムの構築を進めているが、現在、明治 19（1886）年から平成 29（2017）年までに公布された法律と勅令のXML文書化を完了し、それらの全文検索が可能なデータベースの構築を終えた。本報告では、既存のデータベースの問題点について述べた上で、構築したデータベースを説明する。

DOI： 10.24506/jsda.7.s2_s142

CiNii Research

出版者・発行元：Proceedings - 2023 IEEE 6th International Conference on Multimedia Information Processing and Retrieval, MIPR 2023  

This paper proposes an effective image retrieval method. Recent image retrieval approaches attempt to construct a single global feature, including local features of an image. In contrast, this paper proposes multiple features for each image. The basic idea is that a target object in a query image is not necessarily in a major part of a database image; therefore, its single feature may include noisy information from the surroundings of the target object. To deal with this, this paper proposes a Visual Passage Score Aggregation framework (VPSA). VPSA first decomposes an image into several pieces of images, called Visual Passages. Based on visual passages, VPSA aggregates relevance scores of visual passages for ranking. VPSA is efficient in the retrieval phase because an ordinary nearest neighbor search is used. The experiment revealed that VPSA showed superior or comparable performance to the state-of-the-art methods, and it takes a shorter time in the retrieval phase.

DOI： 10.1109/MIPR59079.2023.00021

Scopus

掲載種別：論文集(書籍)内論文   出版者・発行元：Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)  

Most content summarization models from the field of natural language processing summarize the textual contents of a collection of documents or paragraphs. In contrast, summarizing the visual contents of a collection of images has not been researched to this extent. In this paper, we present a framework for summarizing the visual contents of an image collection. The key idea is to collect the scene graphs for all images in the image collection, create a combined representation, and then generate a visually summarizing caption using a scene-graph captioning model. Note that this aims to summarize common contents across all images in a single caption rather than describing each image individually. After aggregating all the scene graphs of an image collection into a single scene graph, we normalize it by using an additional concept generalization component. This component selects the common concept in each sub-graph with ConceptNet based on word embedding techniques. Lastly, we refine the captioning results by replacing a specific noun phrase with a common concept from the concept generalization component to improve the captioning results. We construct a dataset for this task based on the MS-COCO dataset using techniques from image classification and image-caption retrieval. An evaluation of the proposed method on this dataset shows promising performance.

DOI： 10.1007/978-3-031-27077-2_14

Web of Science

Scopus

掲載種別：研究論文（国際会議プロシーディングス）   出版者・発行元：Proceedings of MVA 2023 - 18th International Conference on Machine Vision and Applications  

Agriculture is the pillar industry of human survival. However, various crop diseases reduce the human food supply and lead to starvation and death in the worst cases. Experts perform visual symptoms observation for crop disease diagnosis. Which process is time-consuming and expensive. Also, the process has significant risk of human error due to subjective perception. Convolutional Neural Networks (CNN) use image processing techniques to show great potential in plant disease detection. However, it requires thousands of channels to learn rich features, resulting in large models requiring powerful computing, power supply, and high bandwidth, making it more expensive and difficult for farmers to acquire. Therefore, deploying these solutions on resource-constrained devices is desirable to make them more accessible. Thus, we propose a lightweight object detection CNN that can run on resource-constrained devices to detect crop diseases. Channel pruning is applied to optimize resource use by removing unimportant channels and filter weights to reduce network parameters, inference time, and the number of FLOPS. Experimental results with object detector, Faster R-CNN with two backbones, ResNet-50, and EfficientNet-B7, show significant improvement in model efficiency, keeping high accuracy.

DOI： 10.23919/MVA57639.2023.10215815

Web of Science

Scopus

その他リンク： https://dblp.uni-trier.de/db/conf/mva/mva2023.html#PadeiroKI23

掲載種別：研究論文（国際会議プロシーディングス）   出版者・発行元：Proceedings of MVA 2023 - 18th International Conference on Machine Vision and Applications  

Object detection is the task of detecting objects in an image. In this task, the detection of small objects is particularly difficult. Other than the small size, it is also accompanied by difficulties due to blur, occlusion, and so on. Current small object detection methods are tailored to small and dense situations, such as pedestrians in a crowd or far objects in remote sensing scenarios. However, when the target object is small and sparse, there is a lack of objects available for training, making it more difficult to learn effective features. In this paper, we propose a specialized method for detecting a specific category of small objects; birds. Particularly, we improve the features learned by the neck; the sub-network between the backbone and the prediction head, to learn more effective features with a hierarchical design. We employ Swin Transformer to upsample the image features. Moreover, we change the shifted window size for adapting to small objects. Experiments show that the proposed Swin Transformer-based neck combined with CenterNet can lead to good performance by changing the window sizes. We further find that smaller window sizes (default 2) benefit mAPs for small object detection.

DOI： 10.23919/MVA57639.2023.10216093

Web of Science

Scopus

その他リンク： https://dblp.uni-trier.de/db/conf/mva/mva2023.html#HuoKLKHKI23

掲載種別：研究論文（国際会議プロシーディングス）  

DOI： 10.1145/3607541.3616818

Web of Science

その他リンク： https://dblp.uni-trier.de/db/conf/mcge/mcge2023.html#Matsuhira0KHDI23

掲載種別：研究論文（学術雑誌）  

DOI： 10.48550/arXiv.2307.09143

掲載種別：研究論文（国際会議プロシーディングス）   出版者・発行元：Proceedings of MVA 2023 - 18th International Conference on Machine Vision and Applications  

Small Object Detection (SOD) is an important machine vision topic because (i) a variety of real-world applications require object detection for distant objects and (ii) SOD is a challenging task due to the noisy, blurred, and less-informative image appearances of small objects. This paper proposes a new SOD dataset consisting of 39,070 images including 137,121 bird instances, which is called the Small Object Detection for Spotting Birds (SOD4SB) dataset. The detail of the challenge with the SOD4SB dataset 1 is introduced in this paper. In total, 223 participants joined this challenge. This paper briefly introduces the award-winning methods. The dataset 2, the baseline code 3, and the website for evaluation on the public testset 4 are publicly available.

DOI： 10.23919/MVA57639.2023.10215935

Web of Science

Scopus

その他リンク： https://dblp.uni-trier.de/db/conf/mva/mva2023.html#KondoUYHSHHHXWLHKLKHKISLLY23

掲載種別：研究論文（学術雑誌）  

DOI： 10.48550/arXiv.2303.03144

掲載種別：研究論文（国際会議プロシーディングス）   出版者・発行元：Proceedings of MVA 2023 - 18th International Conference on Machine Vision and Applications  

This paper proposes a method for estimating impressions from images according to the personal attributes of users so that they can find the desired images based on their tastes. Our previous work, which considered gender and age as personal attributes, showed promising results, but it also showed that users sharing these attributes do not necessarily share similar tastes. Therefore, other attributes should be considered to capture the personal tastes of each user well. However, taking more attributes into account leads to a problem in which insufficient amounts of data are served to classifiers due to the explosion of the number of combinations of attributes. To tackle this problem, we propose an aggregation-based method to condense training data for impression estimation while considering personal attribute information. For evaluation, a dataset of 4,000 carpet images annotated with 24 impression words was prepared. Experimental results showed that the use of combinations of personal attributes improved the accuracy of impression estimation, which indicates the effectiveness of the proposed approach.

DOI： 10.23919/MVA57639.2023.10216055

Web of Science

Scopus

その他リンク： https://dblp.uni-trier.de/db/conf/mva/mva2023.html#KojimaKKDI23

出版者・発行元：IEEE Access  

Summarization is a challenging task that aims to generate a summary by grasping common information of a given set of information. Text summarization is a popular task of determining the topic or generating a textual summary of documents. In contrast, image summarization aims to find a representative summary of a collection of images. However, current methods are still restricted to generating a visual scene graph, tags, and noun phrases, but cannot generate a fitting textual description of an image collection. Thus, we introduce a novel framework for generating a summarized caption of an image collection. Since scene graph generation shows advancement in describing objects and their relationships on a single image, we use it in the proposed method to generate a scene graph for each image in an image collection. Then, we find common objects and their relationships from all scene graphs and represent them as a summarized scene graph. For this, we merge all scene graphs and select part of it by estimating the most common objects and relationships. Finally, the summarized scene graph is input into a captioning model. In addition, we introduce a technique to generalize specific words in the final caption into common concept words incorporating external knowledge. To evaluate the proposed method, we construct a dataset for this task by extending the annotation of the MS-COCO dataset using an image retrieval method. The evaluation of the proposed method on this dataset showed promising performance compared to text summarization-based methods.

DOI： 10.1109/ACCESS.2023.3332098

Web of Science

Scopus

掲載種別：研究論文（国際会議プロシーディングス）   出版者・発行元：ACM International Conference Proceeding Series  

Text classification using deep learning, which is trained with a tremendous amount of text, has achieved superior performance than traditional methods. In addition to its success, multi-Task learning has become a promising approach for text classification; for instance, a multi-Task learning approach employs named entity recognition as an auxiliary task for text classification. The existing MTL-based text classification methods depend on auxiliary tasks using supervised labels, which require large human and/or financial efforts to create. To reduce these efforts, this paper proposes a multi-Task learning-based text classification framework which reduces the additional efforts on supervised label creation. A basic idea to realize this is that to utilize phrasal expressions consisting of subwords (called subword-phrase). To the best of our knowledge, there has been no text classification approach on top of subword-phrases, because subwords do not always express a coherent set of meanings. The proposed framework is new to add subword-phrase recognition as an auxiliary task, and to utilize subword-phrases for text classification. To realize the low-cost auxiliary recognition task, the framework extracts subword-phrases in an unsupervised manner. The experimental evaluation of the five popular datasets for text classification showcases the effectiveness of the involvement of the subword-phrase recognition as an auxiliary task. It also shows comparative results with the state-of-The-Art method.

DOI： 10.1145/3568562.3568635

Scopus

掲載種別：研究論文（国際会議プロシーディングス）   出版者・発行元：AVSS 2022 - 18th IEEE International Conference on Advanced Video and Signal-Based Surveillance  

We propose a method to detect birds in a 3D environment referring to both audio information observed from a microphone array and visual information observed from a panorama camera. In general, in panorama images, birds appear relatively too small to be detected accurately even with the state-of-the-art deep learning models. Thus, the proposed method takes a two step approach where the birds are first roughly located referring to audio information by Sound Source Localization (SSL), and then image detection is applied within its vicinity. Through evaluation on a dataset annotated with bounding boxes surrounding the birds, we show that the proposed method improves detection performance of birds that appear in relatively small sizes in the image, in both accuracy and processing speed.

DOI： 10.1109/AVSS56176.2022.9959510

Web of Science

Scopus

その他リンク： https://dblp.uni-trier.de/db/conf/avss/avss2022.html#KawanishiICMKKD22

掲載種別：研究論文（国際会議プロシーディングス）   出版者・発行元：Proceedings - 5th International Conference on Multimedia Information Processing and Retrieval, MIPR 2022  

Image captioning is one of the main goals in vision and language processing, which aims to generate proper descriptions of images. Recently, the attention mechanisms became crucial in captioning tasks, as they can capture global dependencies between modalities. Moreover, some works have used objects detected from the input image as anchor points, so called object tags, to ease such alignments resulting in good performance for this task. In this paper, we newly introduce action information as a prior to further improve this, by adding action tags for training. The action tags can learn alignment at action semantic level and catch the previously ignored dimension of action, that could be very important in image captioning. We found that training with action tags can be used to describe images in a dynamic style. Furthermore, we found it can actually lead to a significant improvement compared with other methods in captioning performance measured by common metrics.

DOI： 10.1109/MIPR54900.2022.00041

Scopus

その他リンク： https://dblp.uni-trier.de/db/conf/mipr/mipr2022.html#HuoKKI22

掲載種別：論文集(書籍)内論文   出版者・発行元：Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)  

The research field of digital libraries mainly deals with data with graph structure. Graph database management systems (GDBMSs) are suitable for managing data in the digital library because the data size is large and its structure is complex. However, when performing a non-simple search or analysis on a graph, GDBMSs cannot avoid reaching already-scanned nodes from different starting nodes by repeatedly traversing edges such as property paths pattern in SPARQL. Therefore, when a GDBMS reaches high degree nodes, the number of graph traversals increases in proportion to the number of its adjacent nodes. Consequently, the cost of traversing multiple paths extremely increases affected by nodes connected enormous the number of edges in conventional GDBMSs. In this paper, we propose a data access approach by repeatedly traversing edges belonging to a specific relationship or anything one while distinguishing between high degree nodes and low degree ones. Finally, a result of our experiment indicated our approach can increase the speed of repeat traversals by a factor of a maximum of ten.

DOI： 10.1007/978-3-031-21756-2_29

Web of Science

Scopus

掲載種別：研究論文（国際会議プロシーディングス）   出版者・発行元：Proceedings - 5th International Conference on Multimedia Information Processing and Retrieval, MIPR 2022  

Gait is one of the most familiar action for us, that is why we can distinguish slight difference of human gaits and perceive their impressions. However, the relationship has been never explored because of the absence of intuitive labels for the slight differences. In this paper, to solve this problem, we propose a intuitive gait model using Japanese mimetic-words. A mimetic-word has sound-symbolism, which means that there is an association between linguistic sounds and sensory experiences, and the phonemes of a mimetic-word is strongly related to the visual sensation. Thanks to the sound-symbolism, Japanese mimetic-words have a possibility of modeling gaits intuitively. Thus, we have previously proposed a method which describes gait with a mimetic-word. In this paper, in the opposite direction, we propose a method which generates gait from a mimetic-word, and confirm the effectiveness of the proposed intuitive gait model which consists of the phonetic-vector through evaluations of both the generation task and the description task.

DOI： 10.1109/MIPR54900.2022.00050

Scopus

その他リンク： https://dblp.uni-trier.de/db/conf/mipr/mipr2022.html#KatoHDIKKDM22

CiNii Research

掲載種別：研究論文（国際会議プロシーディングス）   出版者・発行元：Proceedings - 4th International Conference on Multimedia Information Processing and Retrieval, MIPR 2021  

Data exploration is a fundamental user task in the information seeking process. In data exploration, users have ambiguous information needs, and they traverse across the data for gathering information. In this paper, a novel data exploration system, called FPX-G, is proposed that uses virtual reality (VR) technology. VR-based data exploration (or immersive analytics) is a recent trend in data analytics, and the existing work approaches involve aggregated information in an interactive and 3D manner. However, exploration for individual pieces of data scarcely has been approached. Traditional data exploration is done on 2D displays, therefore space is limited, and there is no depth. FPX-G fully utilizes 3D space to make individual piece of data visible in the user’s line of sight. In this paper, the data structure in FPX-G is designed as a graph, and the data exploration process is modeled as graph traversal. To utilize the capability of VR, FPX-G provides a first person view-based interface from which users can look at individual pieces of data and can walk through the data (like walking in a library). In addition to the walking mechanism, to deal with limited physical space in a room, FPX-G introduces eye-tracking technology for traversing data through a graph. A simulation-based evaluation reveals that FPX-G provides a significantly efficient interface for exploring data compared with the traditional 2D interface.

DOI： 10.1109/MIPR51284.2021.00018

Scopus

その他リンク： https://dblp.uni-trier.de/db/conf/mipr/mipr2021.html#KomamizuIOT21

出版者・発行元：WAT 2021 - 8th Workshop on Asian Translation, Proceedings of the Workshop  

For updating the translations of Japanese statutes based on their amendments, we need to consider the translation “focality;” that is, we should only modify expressions that are relevant to the amendment and retain the others to avoid misconstruing its contents. In this paper, we introduce an evaluation metric and a corpus to improve focality evaluations. Our metric is called an Inclusive Score for DIfferential Translation: (ISDIT). ISDIT consists of two factors: (1) the n-gram recall of expressions unaffected by the amendment and (2) the n-gram precision of the output compared to the reference. This metric supersedes an existing one for focality by simultaneously calculating the translation quality of the changed expressions in addition to that of the unchanged expressions. We also newly compile a corpus for Japanese partially amendment translation that secures the focality of the post-amendment translations, while an existing evaluation corpus does not. With the metric and the corpus, we examine the performance of existing translation methods for Japanese partially amendment translations.

Scopus

掲載種別：研究論文（学術雑誌）  

DOI： 10.26421/JDI2.4-5

掲載種別：研究論文（学術雑誌）   出版者・発行元：Knowledge and Information Systems  

Linked Data (LD) has become a valuable source of factual records, and entity search is a fundamental task in LD. The task is, given a query consisting of a set of keywords, to retrieve a set of relevant entities in LD. The state-of-the-art approaches for entity search are based on information retrieval techniques. This paper first examines these approaches with a traditional evaluation metric, recall@k, to reveal their potential for improvement. To obtain evidence for the potentials, an investigation is carried out on the relationship between queries and answer entities in terms of path lengths on a graph of LD. On the basis of the investigation, learning representations of entities are dealt with. The existing methods of entity search are based on heuristics that determine relevant fields (i.e., predicates and related entities) to constitute entity representations. Since the heuristics require burdensome human decisions, this paper is aimed at removing the burden with a graph proximity measurement. To this end, in this paper, RWRDoc is proposed. It is an RWR (random walk with restart)-based representation learning method that learns representations of entities by using weighted combinations of representations of reachable entities w.r.t. RWR. RWRDoc is mainly designed to improve recall scores; therefore, as shown in experiments, it lacks capability in ranking. In order to improve the ranking qualities, this paper proposes a personalized PageRank-based re-ranking method, PPRSD (Personalized PageRank-based Score Distribution), for the retrieved results. PPRSD distributes relevance scores calculated by text-based entity search methods in a personalized PageRank manner. Experimental evaluations showcase that RWRDoc can improve search qualities in terms of recall@1000 and PPRSD can compensate for RWRDoc’s insufficient ranking capability, and the evaluations confirmed this compensation.

DOI： 10.1007/s10115-020-01445-4

Web of Science

Scopus

記述言語：日本語   掲載種別：研究論文（学術雑誌）   出版者・発行元：一般社団法人 人工知能学会  

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.

DOI： 10.1527/tjsai.h-j53

Scopus

CiNii Research

出版者・発行元：一般社団法人 人工知能学会  

<p>法令文書には，「者」「物」「もの」や「規定」「規程」のように互いに類似している法令用語が出現する．このような法令用語は，法制執務（法令の起草・制定・改廃など法令文書の作成・管理に関する業務）の慣習や規則によって使用法とともに定義されている．法令において，これらの法令用語はそれに従い，厳密に書き分ける必要がある．契約書や約款などの広義の法令文書においても，誤解を防ぐために，法令に準じて正しく書き分けることが望ましい．そこで，本研究では，与えられた法令文から法令用語を検出し，誤用と思われるものに対してその修正案を出力することにより，法令文書の作成を支援する手法を提案する．本手法では，このタスクを選択肢付き穴埋め問題とみなし，分類器により解決する．分類器は，一般文によって事前学習したBERTモデルから構築する．このとき，（１）法令文によるドメイン適応，（２）訓練データのアンダーサンプリング，（３）分類器の統一の三つの工夫を施すことにより性能向上を図る．実験の結果，ランダムフォレストやニューラル言語モデルによる分類器よりも本手法の方が高い性能を発揮することを明らかにした．</p>

DOI： 10.11517/pjsai.JSAI2020.0_4P3OS805

記述言語：英語   出版者・発行元：Transactions of the Japanese Society for Artificial Intelligence  

We propose a method to assist legislative drafters that locates inappropriate legal terms in Japanese statutory sentences and suggests corrections. We focus on sets of mistakable legal terms whose usages are defined in legislation drafting rules. Our method predicts suitable legal terms using a classifier based on BERT (Bidirectional Encoder Representations from Transformers). The BERT classifier is pretrained with a huge number of whole sentences; thus, it contains abundant linguistic knowledge. Classifiers for predicting legal terms suffer from two-level infrequency: term-level infrequency and set-level infrequency. The former causes a class imbalance problem and the latter causes an underfitting problem; both degrade classification performance. To overcome these problems, we apply three techniques, namely, preliminary domain adaptation, repetitive soft undersampling, and classifier unification. The preliminary domain adaptation improves overall performance by providing prior knowledge of statutory sentences, the repetitive soft undersampling overcomes term-level infrequency, and the classifier unification overcomes set-level infrequency while saving storage consumption. Our experiments show that our classifier outperforms conventional classifiers using Random Forest or language models, and that all three training techniques improve performance.

DOI： 10.1527/tjsai.E-K25

Scopus

CiNii Research

記述言語：英語   掲載種別：研究論文（その他学術会議資料等）   出版者・発行元：Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)  

Legal documents are typically hierarchically structured. This paper focuses on ordinances and rules (OR documents for short) in the local governments, which are designed for social lives under the governments. OR documents are composed of provisions for social lives in various aspects such as healthy development of youths and landscape preservation. OR documents in different local governments share common provisions but also include different provisions depending on their social situations. There is a large demand on helping governmental officers draft OR documents, especially searching “relevant parts” of OR documents. To help drafting OR documents, this paper designs the relevancy of OR documents with two basic measurements; matching ratio and provision commonality. Based on the relevancy, this paper develops a structured document search algorithm for OR documents. Experimental evaluation on real OR documents in Japan demonstrates that the proposed algorithm successfully discovers relevant parts of OR documents.

DOI： 10.1007/978-3-030-58790-1_1

Web of Science

Scopus

その他リンク： https://dblp.uni-trier.de/db/conf/jsai/jsai2019w.html#KomamizuFOT19

掲載種別：研究論文（学術雑誌）  

その他リンク： https://dblp.uni-trier.de/db/journals/corr/corr2009.html#abs-2009-06194

出版者・発行元：CEUR Workshop Proceedings  

Linked Open Data (LOD) has been proliferated over various domains, however, there are still lots of open data in various format other than RDF. Document-centric XML data are such open data that are connected with entities in LOD as supplemental documents for these entities. To utilize document-centric XML data linked from entities in LOD, in this paper, a SPARQL-based seamless access method on RDF and XML data is proposed. In particular, an extension to SPARQL, XQueryFILTER, which enables XQuery as a filter in SPARQL is proposed. For efficient query processing of the combination of SPARQL and XQuery, a query optimization is proposed. Experimental scenarios using real-world data showcase the effectiveness of XQueryFILTER and optimization efficiency.

Scopus

掲載種別：研究論文（国際会議プロシーディングス）   出版者・発行元：Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)  

Class imbalance is commonly observed in real-world data, and it is still problematic in that it hurts classification performance due to biased supervision. Undersampling is one of the effective approaches to the class imbalance. The conventional undersampling-based approaches involve a single fixed sampling ratio. However, different sampling ratios have different preferences toward classes. In this paper, an undersampling-based ensemble framework, MUEnsemble, is proposed. This framework involves weak classifiers of different sampling ratios, and it allows for a flexible design for weighting weak classifiers in different sampling ratios. To demonstrate the principle of the design, in this paper, three quadratic weighting functions and a Gaussian weighting function are presented. To reduce the effort required by users in setting parameters, a grid search-based parameter estimation automates the parameter tuning. An experimental evaluation shows that MUEnsemble outperforms undersampling-based methods and oversampling-based state-of-the-art methods. Also, the evaluation showcases that the Gaussian weighting function is superior to the fundamental weighting functions. In addition, the parameter estimation predicted near-optimal parameters, and MUEnsemble with the estimated parameters outperforms the state-of-the-art methods.

DOI： 10.1007/978-3-030-59051-2_14

Web of Science

Scopus

その他リンク： https://dblp.uni-trier.de/db/conf/dexa/dexa2020-2.html#KomamizuUOT20

記述言語：英語  

記述言語：英語   掲載種別：研究論文（その他学術会議資料等）   出版者・発行元：Proceedings - 2019 IEEE International Conference on Big Data, Big Data 2019  

We propose a method that assists legislative drafters in locating inappropriate legal terms in Japanese statutory sentences and suggests corrections. We focus on sets of mistakable legal terms whose usages are defined in legislation drafting rules. Our method predicts suitable legal terms using a classifier based on a BERT (Bidirectional Encoder Representations from Transformers) model. We apply three techniques in training the BERT classifier, specifically, preliminary domain adaptation, repetitive soft undersampling, and classifier unification. These techniques cope with two levels of infrequency: legal term-level infrequency that causes class imbalance and legal term set-level infrequency that causes underfitting. Concretely, preliminary domain adaptation improves overall performance by providing prior knowledge of statutory sentences, repetitive soft undersampling improves performance on infrequent legal terms without sacrificing performance on frequent legal terms, and classifier unification improves performance on infrequent legal term sets by sharing common knowledge among legal term sets. Our experiments show that our classifier outperforms conventional classifiers using Random Forest or a language model, and that all three training techniques contribute to performance improvement.

DOI： 10.1109/BigData47090.2019.9006511

Scopus

その他リンク： https://dblp.uni-trier.de/db/conf/bigdataconf/bigdataconf2019.html#YamakoshiKOT19

担当区分：筆頭著者   記述言語：英語  

記述言語：日本語   掲載種別：研究論文（その他学術会議資料等）  

担当区分：筆頭著者   記述言語：英語  

記述言語：日本語   掲載種別：研究論文（その他学術会議資料等）  

記述言語：英語  

記述言語：日本語  

記述言語：日本語   掲載種別：研究論文（その他学術会議資料等）  

記述言語：英語  

記述言語：日本語   掲載種別：研究論文（学術雑誌）   出版者・発行元：一般社団法人 人工知能学会  

本研究の目標は，日本法令の要約を提供することである．そのためにランダムフォレストによる重要文抽出に基づく自動要約を提案する． 従来の自動要約に関する研究においては，原文書の情報のみが用いられてきた．近年では機械学習に基づく手法なども提案されている． しかし，そうした機械学習において利用される学習データの量は，特に日本語においては，充分でなかった． それに対し，本研究の法令の要約においては，政府が作成する「法令のあらまし」を利用することにより，この問題を解決する． さらに，従来利用されてきた決定木やSVMを使った手法に代えて，ランダムフォレストを用いた重要文抽出を提案し，その性能が従来手法を上回ることを示す． 本論文の貢献は，従来よりもサイズの大きな要約用コーパスを作成した点と，重要文抽出におけるランダムフォレストの有効性を確認した点にある．

DOI： 10.11517/pjsai.jsai2019.0_4e2os7a02

CiNii Research

記述言語：英語   掲載種別：研究論文（学術雑誌）   出版者・発行元：IEICE Transactions on Information and Systems  

Multi-attributed graphs, in which each node is characterized by multiple types of attributes, are ubiquitous in the real world. Detection and characterization of communities of nodes could have a significant impact on various applications. Although previous studies have attempted to tackle this task, it is still challenging due to difficulties in the integration of graph structures with multiple attributes and the presence of noises in the graphs. Therefore, in this study, we have focused on clusters of attribute values and strong correlations between communities and attribute-value clusters. The graph clustering methodology adopted in the proposed study involves Community detection, Attribute-value clustering, and deriving Relationships between communities and attribute-value clusters (CAR for short). Based on these concepts, the proposed multi-attributed graph clustering is modeled as CAR-clustering. To achieve CAR-clustering, a novel algorithm named CARNMF is developed based on non-negative matrix factorization (NMF) that can detect CAR in a cooperative manner. Results obtained from experiments using real-world datasets show that the CARNMF can detect communities and attribute-value clusters more accurately than existing comparable methods. Furthermore, clustering results obtained using the CARNMF indicate that CARNMF can successfully detect informative communities with meaningful semantic descriptions through correlations between communities and attribute-value clusters.

DOI： 10.1587/transinf.2018DAP0022

Web of Science

Scopus

CiNii Research

Web of Science

出版者・発行元：CEUR Workshop Proceedings  

As law is a blueprint of a society and is changed over time as social environments changed, analyzing histories (change provenances) of laws can reveal important facts such as legislative facts and critical events for the society. Linked Open Data (LOD) has emerged as a preferred method for publishing and sharing open data, however, there is an ontological barrier for publishing law history data as LOD. To break through the barrier, this paper proposes an ontology for law history data of the Japanese statute law. The ontology is inspired from PROV-O and SIOC ontologies. The LOD dataset based on the proposed ontology enables wide variety of analyses on the law history data by simple SPARQL queries. The analyses include simple search, visualization, temporal analysis, data mining, etc. This paper displays parts of the analyses which indicate several legislative facts behind changes of laws. The analyses demonstrate the proposed ontology and LOD dataset are useful for legal data analysis. The proposed ontology is comparable with ELI (European Legislation Identifier) which is designed for EU laws, this paper thus discusses the comparability and future directions of the proposed ontology.

Scopus

出版者・発行元：Proceedings of the 33rd Pacific Asia Conference on Language, Information and Computation, PACLIC 2019  

We propose a method for finding and correct- ing misused Thai legal terms in Thai statu- tory sentences. Our method predicts legal terms using Random Forest classifiers, each of which is optimized for each set of similar legal terms. Each classifier utilizes outside- the-sentence features, namely, promulgation year, title keywords, and section keywords of statutes, in addition to words adjacent to the targeted legal term. Our experiment shows that our method outperformed not only a Ran- dom Forest method without the outside-the- sentence features, but also BERT (Bidirec- tional Encoder Representations from Trans- formers), a powerful language representation model, in overall accuracy.

Scopus

記述言語：英語   掲載種別：研究論文（その他学術会議資料等）   出版者・発行元：Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)  

The nagoy team participated in the NTCIR-14 QA Lab-PoliInfo’s summarization subtask. This paper describes our summarization system for assembly member speeches using random forest classifiers. Since we encountered an imbalance in the data, we were unable to achieve good results in this subtask when training on all data. To solve this problem, we developed a new summarization system that applies multiple random forest classifiers training on different-sized data sets step by step. As a result, our system achieved good performance, especially in the evaluation by ROUGE scores. In this paper, we also compare our system with a single random forest classifier using probability.

DOI： 10.1007/978-3-030-36805-0_9

Scopus

その他リンク： https://dblp.uni-trier.de/db/conf/ntcir/ntcir2019.html#OgawaSKT19

出版者・発行元：CEUR Workshop Proceedings  

Entity search is a fundamental task in Linked Data (LD). The task is, given a keyword search query, to retrieve a set of entities in LD which are relevant to the query. The state-of-the-art approaches for entity search are based on information retrieval technologies such as TF-IDF vectorization and ranking models. This paper examines the approaches by applying a traditional evaluation metrics, recall@k, and shows ranking qualities still room left for improvements. In order to improve the ranking qualities, this paper explores possibilities of graph analytical methods. LD is regarded as a large graph, graph analytical approaches are therefore appropriate for this purpose. Since query-based graph analytical approaches fit to entity search tasks, this paper proposes a personalized PageRank-based re-ranking method, PPRSD (Personalized PageRank based Score Distribution), for retrieved results by the state-of-the-art. The experimental evaluation recognizes improvements but its results are not satisfactory, yet. For further improvements, this paper reports investigations about relationship between queries and entities in terms of path lengths on the graph, and discusses future directions for graph analytical approaches.

Scopus

記述言語：英語  

担当区分：筆頭著者   記述言語：英語  

記述言語：英語  

DOI： 2-9

記述言語：英語  

DOI： 334-339

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）   出版者・発行元：Institute of Electrical and Electronics Engineers Inc.  

Analyzing and feeding back the results on real-world services are important missions in the Big Data era to realize smart city. However, analyzing real-world data is still challenging because of dirtiness of data and large variety of analytic requirements. To cope with the challenges, this paper proposes and develops an analytical toolbox for smart city applications. The analytical toolbox consists of three phases: preparation, analysis, and visualization. The preparation phase deals with the dirtiness of the data by including fundamental data cleansing techniques and data integration techniques. The analysis phase is responsible for ETL (extract, transform and load) process and analytical query processing from the next phase. The visualization phase deals with analytical requirements from users and visualization of analytical results. This paper showcases a real-world use case of the proposed analytical toolbox. The use case is now open in public with help of Fujisawa city, Japan, and this fact indicates that the proposed analytical toolbox is feasible for real-world data analysis and feeding back to citizens.

DOI： 10.1109/BigData.2017.8258429

Scopus

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）   出版者・発行元：Institute of Electrical and Electronics Engineers Inc.  

Data integration is still laboursome task when integrating data are not consistently managed. Such inconsistency can happen easily in real-world situations, such as properties of objects are managed by a central organization and trajectories (or logs) of the objects are recorded by other peripheral organizations. This paper deals with a case of missing ordering information. Integrating property data and log data without ordering information causes duplicated results. In order to solve this problem, this paper proposes a join algorithm, called implicit order join, which discovers implicit ordering information from both property data and log data with help of partial true integrated results from human assistance. With the discovered ordering information, the implicit order join enables to integrate the property data and log data. In order to discover the implicit ordering information, ordering correlation between attribute sequences of property data and log data should be found from comprehensive examination of possible attribute sequence pairs. The potential number of sequence pairs is as high as factorial order of the number of attributes. Therefore, this paper develops a heuristic approach to prune unnecessary examinations based on ordering dependency between attribute sequences. Experimental evaluation in this paper indicates that implicit order join can reduce 77% labouring tasks for integration and the pruning method reduces the number of attribute sequences in orders of magnitude.

DOI： 10.1109/BigData.2017.8258474

Scopus

記述言語：英語   出版者・発行元：Frontiers in Artificial Intelligence and Applications  

We propose a method that assists legislation officers in finding inappropriate Japanese legal terms in Japanese statutory sentences and suggests corrections. In particular, we focus on sets of similar legal terms whose usages are defined in legislation drafting rules. Our method predicts suitable legal terms in statutory sentences using Random Forest classifiers, each of which is optimized for each set of similar legal terms. Our experiment shows that our method outperformed existing modern word prediction methods using neural language models.

DOI： 10.3233/978-1-61499-935-5-161

Web of Science

Scopus

掲載種別：研究論文（国際会議プロシーディングス）   出版者・発行元：Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)  

Linked Data has become a valuable source of factual records. However, because of its simple representations of records (i.e., a set of triples), learning representations of entities is required for various applications such as information retrieval and data mining. Entity representations can be roughly classified into two categories; (1) interpretable representations, and (2) latent representations. Interpretability of learned representations is important for understanding relationship between two entities, like why they are similar. Therefore, this paper focuses on the former category. Existing methods are based on heuristics which determine relevant fields (i.e., predicates and related entities) to constitute entity representations. Since the heuristics require laboursome human decisions, this paper aims at removing the labours by applying a graph proximity measurement. To this end, this paper proposes RWRDoc, an RWR (random walk with restart)-based representation learning method which learns representations of entities by weighted combinations of minimal representations of whole reachable entities w.r.t. RWR. Comprehensive experiments on diverse applications (such as ad-hoc entity search, recommender system using Linked Data, and entity summarization) indicate that RWRDoc learns proper interpretable entity representations.

DOI： 10.1007/978-3-319-98809-2_10

Web of Science

Scopus

その他リンク： https://dblp.uni-trier.de/db/conf/dexa/dexa2018-1.html#Komamizu18

担当区分：筆頭著者   記述言語：英語  

DOI： 284-288

担当区分：筆頭著者   記述言語：英語  

DOI： 35-42

記述言語：日本語  

DOI： 138-145

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）   出版者・発行元：Knowledge Systems Institute Graduate School  

Analyzing behaviours of developers in different platforms (in particular, GitHub and Stack Overflow in this paper) can reveal interesting facts related to development activities. There are only few datasets for analysing crossplatform user behaviours, especially across GitHub and Stack Overflow. Users on GitHub and Stack Overflow are identifiable by equivalences of email addresses. In order to increase the number of identifiable users on these datasets, this paper retrieves potentially identifiable users between GitHub and Stack Overflow not relying only on email addresses. This paper employs a classification-based link prediction, which design the user identification problem as a link prediction problem on the bipartite graph consisting of users of GitHub and those of Stack Overflow. With the identification method, this paper generates a probabilistic dataset containing pairs of users with probabilities (or confidences). This paper, as well, publishes the identification tool in order to enable further data generation on appearing datasets of GitHub, Stack Overflow and others. The generated dataset and tool are highly helpful to accelerate researches on mining software repositories.

DOI： 10.18293/SEKE2017-109

Scopus

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）   出版者・発行元：Springer Verlag  

Ranking quality for keyword search over Linked Data (LD) is crucial when users look for entities from LD, since datasets in LD have complicated structures as well as much contents. This paper proposes a keyword search method, FORK, which ranks entities in LD by ObjectRank, a well-known link-structure analysis algorithm that can deal with different types of nodes and edges. The first attempt of applying ObjectRank to LD search reveals that ObjectRank with inappropriate settings gives worse ranking results than PageRank which is equivalent to ObjectRank with all the same authority transfer weights. Therefore, deriving appropriate authority transfer weights is the most important issue for encouraging ObjectRank in LD search. FORK involves a relevance feedback algorithm to modify the authority transfer weights according with users’ relevance judgements for ranking results. The experimental evaluation of ranking qualities using an entity search benchmark showcases the effectiveness of FORK, and it proves ObjectRank is more feasible raking method for LD search than PageRank and other comparative baselines including information retrieval techniques and graph analytic methods.

DOI： 10.1007/978-3-319-70145-5_5

Scopus

担当区分：筆頭著者   記述言語：英語  

DOI： 10.1109/HPCC-SmartCity-DSS.2016.0192

担当区分：筆頭著者   記述言語：英語  

DOI： 251-256

担当区分：筆頭著者   記述言語：英語  

DOI： 10.1145/2938503.2938532

担当区分：筆頭著者   記述言語：英語   掲載種別：研究論文（学術雑誌）  

DOI： 10.1108/IJWIS-03-2016-0014

担当区分：筆頭著者   記述言語：英語  

DOI： 10.1145/2837185.2837230

担当区分：筆頭著者   記述言語：英語   掲載種別：研究論文（学術雑誌）  

DOI： 10.1108/IJWIS-04-2015-0012

担当区分：筆頭著者   記述言語：英語  

DOI： 10.1145/2684200.2684294

担当区分：筆頭著者   記述言語：英語  

DOI： 10.1109/NBiS.2014.77

担当区分：筆頭著者   記述言語：英語  

DOI： 10.1145/2628194.2628241

担当区分：筆頭著者   記述言語：英語  

DOI： 10.1007/978-3-642-32597-7_40

担当区分：筆頭著者   記述言語：英語   掲載種別：研究論文（学術雑誌）  

DOI： 10.1108/17440081211282865

担当区分：筆頭著者   記述言語：英語  

DOI： 10.1145/2095536.2095544

記述言語：日本語   出版者・発行元：一般社団法人 人工知能学会  

法令文書には，「者」「物」「もの」や「規定」「規程」のように互いに類似している法令用語が出現する．このような法令用語は，法制執務（法令の起草・制定・改廃など法令文書の作成・管理に関する業務）の慣習や規則によって使用法とともに定義されている．法令において，これらの法令用語はそれに従い，厳密に書き分ける必要がある．契約書や約款などの広義の法令文書においても，誤解を防ぐために，法令に準じて正しく書き分けることが望ましい．そこで，本研究では，与えられた法令文から法令用語を検出し，誤用と思われるものに対してその修正案を出力することにより，法令文書の作成を支援する手法を提案する．本手法では，このタスクを選択肢付き穴埋め問題とみなし，分類器により解決する．分類器は，一般文によって事前学習したBERTモデルから構築する．このとき，（１）法令文によるドメイン適応，（２）訓練データのアンダーサンプリング，（３）分類器の統一の三つの工夫を施すことにより性能向上を図る．実験の結果，ランダムフォレストやニューラル言語モデルによる分類器よりも本手法の方が高い性能を発揮することを明らかにした．

DOI： 10.11517/pjsai.jsai2020.0_4p3os805

CiNii Research

開催年月日： 2018年9月

記述言語：日本語   会議種別：口頭発表（一般）  

国名：日本国  

開催年月日： 2018年9月

記述言語：日本語   会議種別：口頭発表（一般）  

国名：日本国  

開催年月日： 2018年9月

記述言語：日本語   会議種別：口頭発表（一般）  

国名：日本国  

開催年月日： 2018年9月

記述言語：日本語   会議種別：口頭発表（一般）  

国名：日本国  

開催年月日： 2018年9月

記述言語：日本語   会議種別：口頭発表（一般）  

国名：日本国  

開催年月日： 2018年3月

記述言語：日本語   会議種別：口頭発表（一般）  

国名：日本国  

開催年月日： 2018年3月

記述言語：日本語   会議種別：口頭発表（一般）  

国名：日本国  

開催年月日： 2017年4月

記述言語：日本語   会議種別：ポスター発表  

国名：日本国  

開催年月日： 2017年3月

記述言語：日本語   会議種別：口頭発表（一般）  

国名：日本国  

開催年月日： 2017年3月

記述言語：日本語   会議種別：口頭発表（一般）  

国名：日本国  

開催年月日： 2014年9月

記述言語：日本語   会議種別：口頭発表（一般）  

国名：日本国  

開催年月日： 2012年3月

記述言語：日本語   会議種別：口頭発表（一般）  

国名：日本国  

開催年月日： 2012年2月

記述言語：日本語   会議種別：公開講演，セミナー，チュートリアル，講習，講義等  

国名：日本国  

開催年月日： 2011年10月

記述言語：日本語   会議種別：口頭発表（一般）  

国名：日本国  

開催年月日： 2011年10月

記述言語：英語   会議種別：ポスター発表  

国名：オーストラリア連邦  

開催年月日： 2011年3月

記述言語：日本語   会議種別：口頭発表（一般）  

国名：日本国  

開催年月日： 2011年2月 - 2011年3月

記述言語：日本語   会議種別：口頭発表（一般）  

国名：日本国  

開催年月日： 2010年3月

記述言語：日本語   会議種別：口頭発表（一般）  

国名：日本国  

開催年月日： 2010年2月 - 2010年3月

記述言語：日本語   会議種別：口頭発表（一般）  

国名：日本国  

開催年月日： 2009年9月

記述言語：日本語   会議種別：口頭発表（一般）  

国名：日本国  

開催年月日： 2009年3月

記述言語：日本語   会議種別：口頭発表（一般）  

国名：日本国  

Scopus

Scopus

Scopus

Scopus

Scopus

Scopus