Award type：Award from international society, conference, symposium, etc.  Country：Viet Nam

Award type：Award from Japanese society, conference, symposium, etc.  Country：Japan

Award type：Award from Japanese society, conference, symposium, etc. 

Award type：Award from Japanese society, conference, symposium, etc.  Country：Japan

Award type：Award from Japanese society, conference, symposium, etc. 

Award type：Award from Japanese society, conference, symposium, etc.  Country：Japan

Award type：Award from Japanese society, conference, symposium, etc. 

Award type：Award from Japanese society, conference, symposium, etc. 

Award type：Award from Japanese society, conference, symposium, etc. 

Award type：Award from Japanese society, conference, symposium, etc.  Country：Japan

Award type：Award from Japanese society, conference, symposium, etc.  Country：Japan

Award type：Award from Japanese society, conference, symposium, etc.  Country：Japan

Award type：Award from international society, conference, symposium, etc.  Country：Netherlands

Award type：Award from international society, conference, symposium, etc.  Country：Netherlands

Award type：Award from Japanese society, conference, symposium, etc. 

Award type：Award from Japanese society, conference, symposium, etc.  Country：Japan

Award type：Award from Japanese society, conference, symposium, etc.  Country：Japan

Award type：Award from Japanese society, conference, symposium, etc.  Country：Japan

Award type：Award from international society, conference, symposium, etc.  Country：Belgium

Award type：Award from Japanese society, conference, symposium, etc.  Country：Japan

Award type：Award from Japanese society, conference, symposium, etc.  Country：Japan

Award type：Award from Japanese society, conference, symposium, etc.  Country：Japan

Publishing type：Part of collection (book)   Publisher：Springer Nature Switzerland  

DOI： 10.1007/978-3-031-78090-5_5

Scopus

Publisher：Knowledge-Based Systems  

Cross-modal recipe retrieval is vital for transforming visual food cues into actionable cooking guidance, making culinary creativity more accessible. Existing methods separately encode the recipe Title, Ingredient, and Instruction using different text encoders, then aggregate them to obtain recipe feature, and finally match it with encoded image feature in a joint embedding space. These methods perform well but require significant computational cost. In addition, they only consider matching the entire recipe and the image but ignore the fine-grained correspondence between recipe components and the image, resulting in insufficient cross-modal interaction. To this end, we propose Unified Text Encoder with Fine-grained Contrastive Learning (UTE-FCL) to achieve a simple but efficient model. Specifically, in each recipe, UTE-FCL first concatenates each of the Ingredient and Instruction texts composed of multiple sentences as a single text. Then, it connects these two concatenated texts with the original single-phrase Title to obtain the concatenated recipe. Finally, it encodes these three concatenated texts and the original Title by a Transformer-based Unified Text Encoder (UTE). This proposed structure greatly reduces the memory usage and improves the feature encoding efficiency. Further, we propose fine-grained contrastive learning objectives to capture the correspondence between recipe components and the image at Title, Ingredient, and Instruction levels by measuring the mutual information. Extensive experiments demonstrate the effectiveness of UTE-FCL compared to existing methods.

DOI： 10.1016/j.knosys.2024.112641

Web of Science

Scopus

Publisher：ACM International Conference Proceeding Series  

Scopus

Language：Japanese   Publisher：一般社団法人 人工知能学会  

DOI： 10.11517/jsaisigtwo.2024.challenge-064_07

CiNii Research

Publishing type：Research paper (scientific journal)  

DOI： 10.1007/s11042-023-17657-z

Web of Science

Scopus

Publishing type：Research paper (scientific journal)  

DOI： 10.1007/s11042-023-15732-z

Web of Science

Scopus

Publishing type：Research paper (scientific journal)  

DOI： 10.1109/ACCESS.2024.3360113

Web of Science

Scopus

Publisher：IEEE Access  

Patients' safety is paramount in the healthcare industry, and reducing medication errors is essential for improvement. A promising solution to this problem involves the development of automated systems capable of assisting patients in verifying their pill intake mistakes. This paper investigates a Pill-Prescription matching task that seeks to associate pills in a multi-pill photo with their corresponding names in the prescription. We specifically aim to overcome the limitations of existing pill detection methods when faced with unseen pills, a situation characteristic of zero-shot learning. We propose a novel method named Zero-PIMA (Zero-shot Pill-Prescription Matching), designed to match pill images with prescription names effectively, even for pills not included in the training dataset. Zero-PIMA is an end-to-end model that includes an object localization module to determine and extract features of pill images and a graph convolutional network to capture the spatial relationship of the pills' text in the prescription. After that, we leverage the contrastive learning paradigm to increase the distance between mismatched pill images and pill name pairs while minimizing the distance between matched pairs. In addition, to deal with the zero-shot pill detection problem, we leverage pills' metadata retrieved from the DrugBank database to fine-tune a pre-trained text encoder, thereby incorporating visual information about pills (e.g., shape, color) into their names, making them more informative and ultimately enhancing the pill image-name matching accuracy. Extensive experiments are conducted on our collected real-world VAIPEPP dataset of multi-pill photos and prescriptions. Through a series of comprehensive experiments, the proposed method outperforms other methods for both seen and unseen pills in terms of mean average precision. These results indicate that the proposed method could reduce medication errors and improve patients' safety.

DOI： 10.1109/ACCESS.2024.3390153

Web of Science

Scopus

Publisher：Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)  

In image retrieval tasks, although efficient methods based on pre-computing information related to retrieval and effective methods utilizing re-ranking have been proposed, developing a method that achieves both efficiency and effectiveness at the same time, remains challenging. To develop an efficient and effective image retrieval method, we propose a simple-yet-effective novel image retrieval framework; R-DiP (Re-ranking based Diffusion Pre-computation). It incorporates an effective re-ranking model into the pre-computation step of an existing efficient method, namely, Offline Diffusion that pre-computes the diffusion process in the offline step and provides a simple linear combination-based retrieval in the online step. Experimental results on standard benchmarks shows that R-DiP performs comparable to the State-Of-The-Art (SOTA) image retrieval method, namely SuperGlobal, while maintaining the efficiency of Offline Diffusion. Notably, in million-scale datasets, R-DiP improves the mAP (mean Average Precision) by about 2.0%, and reduces the speed by about 75% on average, surpassing SOTA methods. These results indicate that R-DiP is a promising solution to the efficiency-effectiveness trade-off in image retrieval, that offers the flexibility to incorporate any advanced re-ranking method in the future.

DOI： 10.1007/978-3-031-68312-1_18

Web of Science

Scopus

Publisher：2024 IEEE 18th International Conference on Automatic Face and Gesture Recognition, FG 2024  

Open-vocabulary Temporal Action Detection (Open-vocab TAD) is an advanced video analysis approach that expands Closed-vocabulary Temporal Action Detection (Closed-vocab TAD) capabilities. Closed-vocab TAD is typically confined to localizing and classifying actions based on a predefined set of categories. In contrast, Open-vocab TAD goes further and is not limited to these predefined categories. This is particularly useful in real-world scenarios where the variety of actions in videos can be vast and not always predictable. The prevalent methods in Open-vocab TAD typically employ a 2-stage approach, which involves generating action proposals and then identifying those actions. However, errors made during the first stage can adversely affect the subsequent action identification accuracy. Additionally, existing studies face challenges in handling actions of different durations owing to the use of fixed temporal processing methods. Therefore, we propose a L-stage approach consisting of two primary modules: Multi-scale Video Analysis (MVA) and Video-Text Alignment (VTA). The MVA module captures actions at varying temporal resolutions, overcoming the challenge of detecting actions with diverse durations. The VTA module leverages the synergy between visual and textual modalities to precisely align video segments with corresponding action labels, a critical step for accurate action identification in Open-vocab scenarios. Evaluations on widely recognized datasets THUMOSl4 and ActivityNet-I.3, showed that the proposed method achieved superior results compared to the other methods in both Open-vocab and Closed-vocab settings. This serves as a strong demonstration of the effectiveness of the proposed method in the TAD task.

DOI： 10.1109/FG59268.2024.10581896

Web of Science

Scopus

Publishing type：Research paper (scientific journal)  

DOI： 10.1109/ACCESS.2024.3378095

Web of Science

Scopus

Publisher：ICMR 2024 - Proceedings of the 2024 International Conference on Multimedia Retrieval  

Our workshop aims to provide a platform for both academic and industrial professionals engaged in the analysis and retrieval of cross-data from diverse perspectives, with a particular emphasis on wearable and ambient sensors, lifelog cameras, social networks, and surrounding sensors. Despite numerous studies exploring individual viewpoints, there remains a significant gap in the analysis and retrieval of cross-data to maximize benefits for humanity. Additionally, challenges such as data security and distributed learning for cross-modal model training and inference arise when dealing with large and distributed datasets. We invite researchers to contribute to this initiative, with the overarching goal of fostering the development of a smart and sustainable society through the efficient utilization of intelligent cross-data analysis and retrieval techniques.

DOI： 10.1145/3652583.3659999

Web of Science

Scopus

Publisher：CMCL 2024 - 13th Edition of the Workshop on Cognitive Modeling and Computational Linguistics, Proceedings of the Workshop  

Understanding human perception of nonsense words is helpful to devise product and character names that match their characteristics. Previous studies have suggested the usefulness of Large Language Models (LLMs) for estimating such human perception, but they did not focus on its emotional aspects. Hence, this study aims to elucidate the relationship of emotions evoked by nonsense words between humans and LLMs. Using a representative LLM, GPT-4, we reproduce the procedure of an existing study to analyze evoked emotions of humans for nonsense words. A positive correlation of 0.40 was found between the emotion intensity scores reproduced by GPT-4 and those manually annotated by humans. Although the correlation is not very high, this demonstrates that GPT-4 may agree with humans on emotional associations to nonsense words. Considering that the previous study reported that the correlation among human annotators was about 0.68 on average and that between a regression model trained on the annotations for real words and humans was 0.17, GPT-4’s agreement with humans is notably strong.

Scopus

Publishing type：Research paper (international conference proceedings)  

DOI： 10.1145/3664647.3681202

Scopus

Other Link： https://dblp.uni-trier.de/db/conf/mm/mm2024.html#Matsuhira0KHI24

Publisher：IEEE Computer Society Conference on Computer Vision and Pattern Recognition Workshops  

Traditional crop disease diagnosis, reliant on expert visual observation, is expensive, time-consuming, and prone to error. While Convolutional Neural Networks (CNNs) offer promising alternatives, their high resource demands limit their accessibility to farmers, particularly those in resource-constrained settings. Lightweight models that operate on resource-limited devices without network access are crucial to address this gap. This paper proposes a Similarity-Preserving Quantization (SPQ) method to convert high-precision CNNs into lower-precision models while maintaining similar feature representations. While quantization offers a promising approach for building lightweight CNNs for crop disease detection, the quality of quantized models often suffers. SPQ addresses this challenge by ensuring equivalent activation patterns for similar crop images in both the original and quantized models. Experimental evaluation using MobileNetV2 and ResNet-50 demonstrates that SPQ improves throughput, inference, and memory footprint more than 3 times while preserving the detection performance.

DOI： 10.1109/CVPRW63382.2024.00216

Scopus

Publishing type：Research paper (international conference proceedings)   Publisher：Association for Computational Linguistics  

DOI： 10.18653/v1/2024.customnlp4u-1.6

Publisher：1st Workshop on Customizable NLP: Progress and Challenges in Customizing NLP for a Domain, Application, Group, or Individual, CustomNLP4U 2024 - Proceedings of the Workshop  

When distributional differences exist between pre-training and fine-tuning data, language models (LMs) may perform poorly on downstream tasks. Recent studies have reported that multi-task learning of downstream task and masked language modeling (MLM) task during the fine-tuning phase improves the performance of the downstream task. Typical MLM tasks (e.g., random token masking (RTM)) tend not to care tokens corresponding to the knowledge already acquired during the pre-training phase, therefore LMs may not notice the important clue or not effective to acquire linguistic knowledge of the task or domain. To overcome this limitation, we propose a new masking strategy for MLM task, called L3Masking 1, that leverages lessons (specifically, token-wise likelihood in a context) learned from the vanilla language model to be fine-tuned. L3Masking actively masks tokens with low likelihood on the vanilla model. Experimental evaluations on text classification tasks in different domains confirms a multi-task text classification method with L3Masking performed task adaptation more effectively than that with RTM. These results suggest the usefulness of assigning a preference to the tokens to be learned as the task or domain adaptation.

Scopus

Publisher：Proceedings of the 5th ACM International Conference on Multimedia in Asia, MMAsia 2023  

Recipe is a set of instructions that describes how to make food. It can help people from the preparation of ingredients, food cooking process, etc. to prepare the food, and increasingly in demand on the Web. To help users find the vast amount of recipes on the Web, we address the task of recipe recommendation. Due to multiple data types and relationships in a recipe, we can treat it as a heterogeneous network to describe its information more accurately. To effectively utilize the heterogeneous network, metapath was proposed to describe the higher-level semantic information between two entities by defining a compound path from peer entities. Therefore, we propose a metapath-enhanced recipe recommendation framework, RecipeMeta, that combines GNN (Graph Neural Network)-based representation learning and specific metapath-based information in a recipe to predict User-Recipe pairs for recommendation. Through extensive experiments, we demonstrate that the proposed model, RecipeMeta, outperforms state-of-the-art methods for recipe recommendation.

DOI： 10.1145/3595916.3626430

Scopus

Publisher：NarSUM 2023 - Proceedings of the 2nd Workshop on User-centric Narrative Summarization of Long Videos, Co-located with: MM 2023  

Scopus

Publishing type：Research paper (scientific journal)   Publisher：Slovenian Association Informatika  

DOI： 10.31449/inf.v47i3.4742

Scopus

Language：English   Publisher：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

Language：English   Publisher：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

Language：English   Publisher：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

Language：English   Publisher：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

Language：Japanese   Publisher：Japan Society for Digital Archive  

DOI： 10.24506/jsda.7.s2_s142

CiNii Research

Publisher：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

Publishing type：Part of collection (book)   Publisher：Springer International Publishing  

DOI： 10.1007/978-3-031-27077-2_14

Web of Science

Scopus

Publishing type：Research paper (international conference proceedings)  

DOI： 10.23919/MVA57639.2023.10215815

Web of Science

Scopus

Other Link： https://dblp.uni-trier.de/db/conf/mva/mva2023.html#PadeiroKI23

Publishing type：Research paper (international conference proceedings)  

DOI： 10.23919/MVA57639.2023.10216093

Web of Science

Scopus

Other Link： https://dblp.uni-trier.de/db/conf/mva/mva2023.html#HuoKLKHKI23

Publishing type：Research paper (international conference proceedings)  

DOI： 10.1145/3607541.3616818

Web of Science

Other Link： https://dblp.uni-trier.de/db/conf/mcge/mcge2023.html#Matsuhira0KHDI23

Publisher：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

Web of Science

Scopus

Publishing type：Research paper (scientific journal)  

DOI： 10.48550/arXiv.2307.09143

Publishing type：Research paper (international conference proceedings)  

DOI： 10.23919/MVA57639.2023.10215935

Web of Science

Scopus

Other Link： https://dblp.uni-trier.de/db/conf/mva/mva2023.html#KondoUYHSHHHXWLHKLKHKISLLY23

Publishing type：Research paper (scientific journal)  

DOI： 10.48550/arXiv.2303.03144

Publishing type：Research paper (international conference proceedings)  

DOI： 10.23919/MVA57639.2023.10216055

Web of Science

Scopus

Other Link： https://dblp.uni-trier.de/db/conf/mva/mva2023.html#KojimaKKDI23

Publishing type：Research paper (scientific journal)  

DOI： 10.1109/ACCESS.2023.3332098

Web of Science

Scopus

Publishing type：Research paper (international conference proceedings)   Publisher：ACM  

DOI： 10.1145/3568562.3568635

Scopus

Publishing type：Research paper (international conference proceedings)   Publisher：IEEE  

DOI： 10.1109/AVSS56176.2022.9959510

Web of Science

Scopus

Other Link： https://dblp.uni-trier.de/db/conf/avss/avss2022.html#KawanishiICMKKD22

Publishing type：Research paper (international conference proceedings)   Publisher：IEEE  

DOI： 10.1109/MIPR54900.2022.00050

Scopus

Other Link： https://dblp.uni-trier.de/db/conf/mipr/mipr2022.html#KatoHDIKKDM22

Publishing type：Research paper (international conference proceedings)   Publisher：IEEE  

DOI： 10.1109/MIPR54900.2022.00041

Scopus

Other Link： https://dblp.uni-trier.de/db/conf/mipr/mipr2022.html#HuoKKI22

Publishing type：Part of collection (book)   Publisher：Springer International Publishing  

DOI： 10.1007/978-3-031-21756-2_29

Web of Science

Scopus

CiNii Research

Publishing type：Research paper (international conference proceedings)   Publisher：IEEE  

DOI： 10.1109/MIPR51284.2021.00018

Scopus

Other Link： https://dblp.uni-trier.de/db/conf/mipr/mipr2021.html#KomamizuIOT21

Publisher：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

Publishing type：Research paper (scientific journal)  

DOI： 10.26421/JDI2.4-5

Publishing type：Research paper (scientific journal)  

DOI： 10.1007/s10115-020-01445-4

Web of Science

Scopus

Language：Japanese   Publishing type：Research paper (scientific journal)   Publisher：The Japanese Society for Artificial Intelligence  

<p>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.</p>

DOI： 10.1527/tjsai.h-j53

Scopus

CiNii Research

Language：English   Publisher：The Japanese Society for Artificial Intelligence  

<p>We propose a method to assist legislative drafters that locates inappropriate legal terms in Japanese statutorysentences and suggests corrections. We focus on sets of mistakable legal terms whose usages are defined in legislationdrafting rules. Our method predicts suitable legal terms using a classifier based on BERT (Bidirectional EncoderRepresentations 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 causesan underfitting problem; both degrade classification performance. To overcome these problems, we apply threetechniques, namely, preliminary domain adaptation, repetitive soft undersampling, and classifier unification. Thepreliminary 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-levelinfrequency while saving storage consumption. Our experiments show that our classifier outperforms conventionalclassifiers using Random Forest or language models, and that all three training techniques improve performance.</p>

DOI： 10.1527/tjsai.E-K25

Scopus

CiNii Research

Publisher：The Japanese Society for Artificial Intelligence  

<p>Legal documents contain legal terms that have similar meaning or pronunciation each other. Japanese legislation defines their usage on the basis of traditional customs and rules. In accordance with the definition, we need to use these legal terms properly and strictly in a statute. We are also encouraged to follow the definition in writing broad-sense legal documents, such as contracts and terms of use. To assist in writing legal documents, we propose a method that locates inappropriate legal terms in Japanese statutory sentences and suggests corrections. We solve this task with a classifier by regarding the task as a sentence completion test. Our classifier is based on a pretrained BERT model trained by using a large amount of general sentences. To raise performance, we apply three training techniques: domain adaptation, undersampling, classifier unification. Our experiments show that our classifier achieved better performance than Random Forest-based ones and language model-based ones.</p>

DOI： 10.11517/pjsai.JSAI2020.0_4P3OS805

Language：English   Publishing type：Research paper (other academic)   Publisher：Springer  

DOI： 10.1007/978-3-030-58790-1_1

Web of Science

Scopus

Other Link： https://dblp.uni-trier.de/db/conf/jsai/jsai2019w.html#KomamizuFOT19

Publishing type：Research paper (international conference proceedings)   Publisher：Springer  

DOI： 10.1007/978-3-030-59051-2_14

Web of Science

Scopus

Other Link： https://dblp.uni-trier.de/db/conf/dexa/dexa2020-2.html#KomamizuUOT20

Publishing type：Research paper (scientific journal)  

Other Link： https://dblp.uni-trier.de/db/journals/corr/corr2009.html#abs-2009-06194

Publisher：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

Language：English  

Language：English   Publishing type：Research paper (other academic)   Publisher：IEEE  

DOI： 10.1109/BigData47090.2019.9006511

Scopus

Other Link： https://dblp.uni-trier.de/db/conf/bigdataconf/bigdataconf2019.html#YamakoshiKOT19

Authorship：Lead author   Language：English  

Language：Japanese   Publishing type：Research paper (other academic)  

Authorship：Lead author   Language：English  

Language：Japanese   Publishing type：Research paper (other academic)  

Language：English  

Language：Japanese  

Language：Japanese   Publishing type：Research paper (other academic)  

Language：English  

Language：Japanese   Publishing type：Research paper (scientific journal)   Publisher：The Japanese Society for Artificial Intelligence  

Our purpose is to provide an automatic summarization for Japanese acts and we propose a sententence extraction method with Random Forest. While the traditional automatic summarization methods have used the information of summarizing source data, in recent years, the methods based on machine learning use the summarization results. However, in such a method, the amount of learning corpus is small, especially in Japanese text. In this research, we solve this problem by using "Outlines of Japanese Statutes," which are official summaries of statutes published by the Japanese government. Furthermore, we show that the sentence extraction method with Random Forest has higher performance rather than with decision trees or with support vector machines.

DOI： 10.11517/pjsai.jsai2019.0_4e2os7a02

CiNii Research

Language：English   Publishing type：Research paper (scientific journal)   Publisher：The Institute of Electronics, Information and Communication Engineers  

<p>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 <u><b>C</b></u>ommunity detection, <u><b>A</b></u>ttribute-value clustering, and deriving <u><b>R</b></u>elationships 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.</p>

DOI： 10.1587/transinf.2018DAP0022

Web of Science

Scopus

CiNii Research

Web of Science

Publisher：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

Publisher：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

Publisher：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

Language：English   Publishing type：Research paper (other academic)   Publisher：Springer  

DOI： 10.1007/978-3-030-36805-0_9

Scopus

Other Link： https://dblp.uni-trier.de/db/conf/ntcir/ntcir2019.html#OgawaSKT19

Language：English  

Authorship：Lead author   Language：English  

Language：English  

DOI： 2-9

Language：English  

DOI： 334-339

Language：English   Publishing type：Research paper (international conference proceedings)   Publisher：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

Language：English   Publishing type：Research paper (international conference proceedings)   Publisher：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

Language：English  

DOI： 10.3233/978-1-61499-935-5-161

Web of Science

Scopus

Publishing type：Research paper (international conference proceedings)   Publisher：Springer  

DOI： 10.1007/978-3-319-98809-2_10

Web of Science

Scopus

Other Link： https://dblp.uni-trier.de/db/conf/dexa/dexa2018-1.html#Komamizu18

Authorship：Lead author   Language：English  

DOI： 284-288

Authorship：Lead author   Language：English  

DOI： 35-42

Language：Japanese  

DOI： 138-145

Language：English   Publishing type：Research paper (international conference proceedings)   Publisher：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

Language：English   Publishing type：Research paper (international conference proceedings)   Publisher：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

Authorship：Lead author   Language：English  

DOI： 10.1109/HPCC-SmartCity-DSS.2016.0192

Authorship：Lead author   Language：English  

DOI： 251-256

Authorship：Lead author   Language：English  

DOI： 10.1145/2938503.2938532

Authorship：Lead author   Language：English   Publishing type：Research paper (scientific journal)  

DOI： 10.1108/IJWIS-03-2016-0014

Authorship：Lead author   Language：English  

DOI： 10.1145/2837185.2837230

Authorship：Lead author   Language：English   Publishing type：Research paper (scientific journal)  

DOI： 10.1108/IJWIS-04-2015-0012

Authorship：Lead author   Language：English  

DOI： 10.1145/2684200.2684294

Authorship：Lead author   Language：English  

DOI： 10.1109/NBiS.2014.77

Authorship：Lead author   Language：English  

DOI： 10.1145/2628194.2628241

Authorship：Lead author   Language：English  

DOI： 10.1007/978-3-642-32597-7_40

Authorship：Lead author   Language：English   Publishing type：Research paper (scientific journal)  

DOI： 10.1108/17440081211282865

Authorship：Lead author   Language：English  

DOI： 10.1145/2095536.2095544

Language：Japanese   Publisher：The Japanese Society for Artificial Intelligence  

Legal documents contain legal terms that have similar meaning or pronunciation each other. Japanese legislation defines their usage on the basis of traditional customs and rules. In accordance with the definition, we need to use these legal terms properly and strictly in a statute. We are also encouraged to follow the definition in writing broad-sense legal documents, such as contracts and terms of use. To assist in writing legal documents, we propose a method that locates inappropriate legal terms in Japanese statutory sentences and suggests corrections. We solve this task with a classifier by regarding the task as a sentence completion test. Our classifier is based on a pretrained BERT model trained by using a large amount of general sentences. To raise performance, we apply three training techniques: domain adaptation, undersampling, classifier unification. Our experiments show that our classifier achieved better performance than Random Forest-based ones and language model-based ones.

DOI： 10.11517/pjsai.jsai2020.0_4p3os805

CiNii Research

Event date： 2018.9

Language：Japanese   Presentation type：Oral presentation (general)  

Country：Japan  

Event date： 2018.9

Language：Japanese   Presentation type：Oral presentation (general)  

Country：Japan  

Event date： 2018.9

Language：Japanese   Presentation type：Oral presentation (general)  

Country：Japan  

Event date： 2018.9

Language：Japanese   Presentation type：Oral presentation (general)  

Country：Japan  

Event date： 2018.9

Language：Japanese   Presentation type：Oral presentation (general)  

Country：Japan  

Event date： 2018.3

Language：Japanese   Presentation type：Oral presentation (general)  

Country：Japan  

Event date： 2018.3

Language：Japanese   Presentation type：Oral presentation (general)  

Country：Japan  

Event date： 2017.4

Language：Japanese   Presentation type：Poster presentation  

Country：Japan  

Event date： 2017.3

Language：Japanese   Presentation type：Oral presentation (general)  

Country：Japan  

Event date： 2017.3

Language：Japanese   Presentation type：Oral presentation (general)  

Country：Japan  

Event date： 2014.9

Language：Japanese   Presentation type：Oral presentation (general)  

Country：Japan  

Event date： 2012.3

Language：Japanese   Presentation type：Oral presentation (general)  

Country：Japan  

Event date： 2012.2

Language：Japanese   Presentation type：Public lecture, seminar, tutorial, course, or other speech  

Country：Japan  

Event date： 2011.10

Language：Japanese   Presentation type：Oral presentation (general)  

Country：Japan  

Event date： 2011.10

Language：English   Presentation type：Poster presentation  

Country：Australia  

Event date： 2011.3

Language：Japanese   Presentation type：Oral presentation (general)  

Country：Japan  

Event date： 2011.2 - 2011.3

Language：Japanese   Presentation type：Oral presentation (general)  

Country：Japan  

Event date： 2010.3

Language：Japanese   Presentation type：Oral presentation (general)  

Country：Japan  

Event date： 2010.2 - 2010.3

Language：Japanese   Presentation type：Oral presentation (general)  

Country：Japan  

Event date： 2009.9

Language：Japanese   Presentation type：Oral presentation (general)  

Country：Japan  

Event date： 2009.3

Language：Japanese   Presentation type：Oral presentation (general)  

Country：Japan  

Scopus

Scopus

Scopus

Scopus

Scopus

Scopus