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

researchmap

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

researchmap

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

researchmap

Authorship：Lead author   Language：English   Publishing type：Research paper (international conference proceedings)  

researchmap

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

researchmap

researchmap

researchmap

Language：Japanese   Publishing type：Research paper (conference, symposium, etc.)  

researchmap

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

researchmap

researchmap

Publishing type：Research paper (international conference proceedings)  

Understanding how people's interest and emotional state change in response to news coverage of a particular topic and elucidating the characteristics of these changes can reveal the shifting nature of attention and emotion. We analyzed people's interest and emotional responses expressed via Twitter in response to news coverage of announcements of new cases of coronavirus disease 2019 (COVID-19) as a case study. As a measure of interest, we examined replies to tweets of news items posted by media outlets on Twitter, and classified the emotional content of each reply tweet using Plutchik's wheel of emotion. The analysis suggested that people were most interested in COVID-19 case numbers in April 2020, when the first wave of cases occurred and the first emergency declaration was issued, and in July 2020, when the second wave of cases emerged. The results revealed that fear was the most commonly expressed emotion. The ratio of fear-related tweets was highest in February and March 2020, a time at which new COVID-19 cases were confirmed in various locations and there was substantial public discussion regarding whether Japan would declare a state of emergency for the first time.

DOI： 10.1109/SCISISIS55246.2022.10001983

Scopus

researchmap

Language：Japanese   Publisher：Japan Society for Fuzzy Theory and Intelligent Informatics  

The developers of new vaccines against SARS-CoV-2 and governments have provided information on vaccine effectiveness and status on a daily basis to reassure people about vaccination against COVID-19. However, because the interest in vaccines and vaccination status varies by country and region, people do not always feel reassured. In this paper, we analyzed tweets posted on Twitter to elucidate the emotions people have toward COVID-19 vaccines and factors that cause such emotions to be expressed. We selected six countries for our analysis: Japan, the United States, the Great Britain, Canada, Australia, and India, and applied an emotion classification method using machine learning based on the eight types of emotions defined in Plutchik’s wheel of emotions. We also used a text analysis approach using dependency analysis and burst detection methods. The results of our emotion classification showed that fear was the most common emotion in Japan whereas anger and disgust were most common in the United States, Great Britain, Canada, and Australia; joy was most common in India. We also analyzed tweets during the period when a particular emotion was increased in the changes of the emotions represented as a time series based on the burst-detected dependency relations, and found several characteristics: many users posted vaccine-related news, one user would often post a large number of tweets with the same content, and the same event related to vaccines could arouse different emotions depending on the individual’s situation.

DOI： 10.3156/jsoft.34.3_592

CiNii Research

researchmap

Publishing type：Research paper (international conference proceedings)  

Researchers and developers search for patents in fields related to their own research to obtain information on issues and effective technologies in those fields for use in their research. However, it is impossible to read through the full text of many patents, so a method that enables patent information to be grasped briefly is needed. In this study, we analyze the structure of U.S. patents with the aim of extracting important information. Using Japanese patents with structural tags such as "field", "problem", "solution", and "effect", and corresponding U.S. patents (patent families), we automatically created a dataset of 81,405 U.S. patents with structural tags. Furthermore, using this dataset, we conduct an experiment to assign structural tags to each sentence in the U. S. patents automatically. For the embedding layer, we use a language representation model, Bidirectional Encoder Representations from Transformer, pretrained on patent documents and construct a multi-label classifier that classifies a given sentence into one of four categories: "field", "problem", "solution", or "effect". Using a loss function that considers the unbalanced amount of data for each structural tag, we are able to classify sentences related to "field", "problem", "solution", and "effect"with precision of 0.6994, recall of 0.8291, and F-measure of 0.7426.

DOI： 10.1109/IIAIAAI55812.2022.00038

Scopus

researchmap

Publishing type：Research paper (international conference proceedings)  

DOI： 10.1109/IIAIAAI55812.2022.00024

researchmap

Other Link： https://dblp.uni-trier.de/db/conf/iiaiaai/iiaiaai2022.html#FukudaNS22

researchmap

researchmap

researchmap

researchmap

researchmap

researchmap

researchmap

Publishing type：Research paper (scientific journal)  

This paper presents results of experiments in which annotators were asked to selectively reexamine their decisions when those decisions seemed inconsistent. The annotation task was binary topic classification. To operationalize the concept of annotation consistency, a text classifier was trained on all manual annotations made during a complete first pass and then used to automatically recode every document. Annotators were then asked to perform a second manual pass, limiting their attention to cases in which their first annotation disagreed with the text classifier. On average across three annotators, each working independently, 11% of first pass annotations were reconsidered, 46% of reconsidered annotations were changed in the second pass, and 71% of changed annotations agreed with decisions made independently by an experienced fourth annotator. The net result was that for an 11% average increase in annotation cost it was possible to increase overall chance corrected agreement with the annotation decisions of an experienced annotator (as measured by kappa) from 0.70 to 0.75.

DOI： 10.1002/pra2.301

Scopus

researchmap

Language：Japanese  

In an academic paper search, it is important that the search returns comprehensive results that are relevant to the user's information need by creating a Boolean query. However, it is difficult to anticipate all possible terms that authors of relevant papers might have used. We propose a Boolean-based search method based on topic analysis using latent Dirichlet allocation. Our method considers synonyms and expressions similar to the search terms, which a user might not anticipate. The sets retrieved by our method and by general ranking method are different because the purpose of the ranking method is to high rank the relevant papers and our method is focused on comprehensively collecting relevant papers. Therefore, it is able to that more high performance search result could be obtained by combining both sets.

CiNii Books

CiNii Research

researchmap

Publishing type：Research paper (international conference proceedings)  

Archives hold printed historical documents, many of which have deteriorated. It is difficult to extract text from such images without errors using optical character recognition (OCR). This problem reduces the accuracy of information retrieval. Therefore, it is necessary to improve the performance of OCR for images of deteriorated documents. One approach is to convert images of deteriorated documents to clear images, to make it easier for an OCR system to recognize text. To perform this conversion using a neural network, data is needed to train it. It is hard to prepare training data consisting of pairs of a deteriorated image and an image from which deterioration has been removed; however, it is easy to prepare training data consisting of pairs of a clear image and an image created by adding noise to it. In this study, PDFs of historical documents were collected and converted to text and JPEG images. Noise was added to the JPEG images to create a dataset in which the images had noise similar to that of the actual printed documents. U-Net, a type of neural network, was trained using this dataset. The performance of OCR for an image with noise in the test data was compared with the performance of OCR for an image generated from it by the trained U-Net. An improvement in the OCR recognition rate was confirmed.

DOI： 10.1007/978-3-030-34058-2_31

Scopus

researchmap

Publishing type：Research paper (international conference proceedings)  

Prior work on automated annotation of human values has sought to train text classification techniques to label text spans with labels that reflect specific human values such as freedom, justice, or safety. This confounds three tasks: (1) selecting the documents to be labeled, (2) selecting the text spans that express or reflect human values, and (3) assigning labels to those spans. This paper proposes a three-stage model in which separate systems can be optimally trained for each of the three stages. Experiments from the first stage, document selection, indicate that annotation diversity trumps annotation quality, suggesting that when multiple annotators are available, the traditional practice of adjudicating conflicting annotations of the same documents is not as cost effective as an alternative in which each annotator labels different documents. Preliminary results for the second stage, selecting value sentences, indicate that high recall (94%) can be achieved on that task with levels of precision (above 80%) that seem suitable for use as part of a multi-stage annotation pipeline. The annotations created for these experiments are being made freely available, and the content that was annotated is available from commercial sources at modest cost.

DOI： 10.1007/978-3-030-15742-5_18

Scopus

researchmap

Publishing type：Research paper (international conference proceedings)  

When conducting a search for research papers, the search should return comprehensive results related to the user’s query. In general, a user inputs a Boolean query that reflects the information need, and the search engine ranks the research papers based on the query. However, it is difficult to anticipate all possible terms that authors of relevant papers might have used. Moreover, general query-based ranking methods emphasize how to rank the relevant documents at the top of the results, but require some means of guaranteeing the comprehensiveness of the results. Therefore, two ranking methods that consider the comprehensiveness of relevant papers are proposed. The first uses a topic-based Boolean query search. This search converts every word in the abstract set and query into a topic via topic analysis by Latent Dirichlet Allocation (LDA) and conducts a search at the topic level. The topic assigned to synonyms of a search term is expected to be the same as that assigned to the search term. Each paper is ranked based on the number of times it is matched with each topic-based Boolean query search executed for various LDA parameter settings. The second is a hybrid method that emphasizes better results from our topic-based ranking result and a general query-based ranking result. This method is based on the observation that the paper sets retrieved by our method and by a general ranking method will be different. Through experiments using the NTCIR-1 and -2 datasets, the effectiveness of our topic-based and hybrid methods are demonstrated.

DOI： 10.1007/978-3-030-27618-8_5

Scopus

researchmap

researchmap

Language：English  

researchmap

Language：English  

researchmap

Language：English  

researchmap

Publishing type：Research paper (international conference proceedings)  

In an academic paper search to confirm the originality of a user's research, it is important that the search returns comprehensive results relevant to the user's information need. To achieve comprehensive search results, users often relax initially restrictive search formula by adding synonyms and expressions similar to the search words with operator OR, and/or replacing AND with OR operations. However, it is difficult to anticipate all the terms that authors of relevant papers might have used. In addition, the replacement of AND with OR in search phrases can return a large number of unrelated papers. To overcome these issues, we propose a research paper search method based on topic analysis, which uses Boolean search based on the topics assigned to the search words in the search formula and the abstracts that contain any search word. Our method considers synonyms and expressions similar to the search words, which a user might not anticipate, while limiting the number of papers unrelated to the information need in the search result. To investigate the effectiveness of our method, we conducted experiments using the NTCIR-1 and 2 datasets, and confirmed that our method shows a reduction effect on unrelated papers, while maintaining high coverage.

DOI： 10.1109/IALP.2017.8300606

Scopus

researchmap

Publishing type：Research paper (international conference proceedings)  

We are investigating ways of evaluating research impact as soon as possible after publication. Traditionally, the research impact or importance of academic journals has been evaluated using citation relations, such as the impact factor and the citation half-life. However, these citation-based methods require long periods to evaluate research impact and therefore are not suitable for evaluating the current impact of research papers at conferences. To solve this problem, we are studying the automatic evaluation of research impact using Twitter. Researchers participating in academic conferences often post their opinions or comments on Twitter. Here, research papers (presentations) that have many comments are considered to be outstanding and to have strong impact during the conference. In this paper, we propose a method for automatically aligning tweets with research papers. The procedure consists of the following three steps: (1) detecting valuable tweets, (2) aligning each valuable tweet with a research paper, and (3) calculating the research impact of each research paper by the number of aligned tweets. We conducted some experiments to confirm the effectiveness of our method. From the results, we obtained an MRR score of 0.223, which outperformed a baseline method.

DOI： 10.1109/DEXA.2015.64

Scopus

researchmap

Language：Japanese   Publisher：Japan Society of Library and Information Science  

We propose a method for the automatic classification of research papers in terms of the KAKEN classification index using a machine learning method. This classification index was originally devised to classify reports for the KAKEN research fund in Japan, and it is organized as a three-level hierarchy: Area, Discipline, and Research Field. Traditionally, researcher and conference names are used as cue phrases to classify the research paper efficiently. In addition to these cue phrases, we focus on elemental technologies and their effects, as discussed in each research paper. Examining the use of elemental technology terms used in each research paper and their effects is important for characterizing the research field to which a given research paper belongs. Therefore, we use elemental technology terms and their effects as additional cue phrases for machine-learning-based text classification. To investigate the effectiveness of our method, we conducted some experiments using the KAKEN and CiNii articles data. From the experimental results, we obtained average precision scores of 0.853, 0.712, and 0.615 for the Area, Discipline, and Research Field levels in the KAKEN classification index, respectively. These scores are higher than those for the method not using elemental technologies and their effects. From these results, we confirmed the effectiveness of using elemental technology terms and their effects as cue phrases.

DOI： 10.20651/jslis.62.3_145

CiNii Books

CiNii Research

researchmap

Publishing type：Research paper (international conference proceedings)  

In scientometrics and citation analysis, several measures for evaluating the industrial relevance or the impact of academic research fields have been proposed. What seems to be lacking, however, is that these measures could not evaluate the recent industrial relevance or impact of each field, because most of them rely on citations of research papers in patents and vice versa. In this paper, we attempt to evaluate the industrial and social impact of research fields using document classification techniques. Our method classifies research papers and news articles using systems including the International Patent Classification (IPC) and the KAKEN classification index. Then it evaluates the industrial and social impact of each field by comparing the number of research papers with the number of patents or articles in the IPC categories and the projects funded in each KAKEN category.

DOI： 10.1109/IIAI-AAI.2014.29

Scopus

researchmap

Language：Japanese  

CiNii Books

CiNii Research

researchmap

Other Link： http://id.ndl.go.jp/bib/024421636

Publishing type：Research paper (scientific journal)  

To a researcher in a field with high industrial relevance, retrieving and analyzing research papers and patents are important aspects of assessing the scope of the field. Knowledge of the history and effects of the elemental technologies is important for understanding trends. We propose a method for automatically creating a technical trend map from both research papers and patents by focusing on the elemental (underlying) technologies and their effects. We constructed a method that can be used in any research field. To investigate the effectiveness of our method, we conducted an experiment using the data in the NTCIR-8 Workshop Patent Mining Task. The results of our experiment showed recall and precision scores of 0.254 and 0.496, respectively, for the analysis of research papers, and recall and precision scores of 0.455 and 0.507, respectively, for the analysis of patents. Those results indicate that our method for mapping technical trends is both useful and sound. © 2012 Satoshi Fukuda, Hidetsugu Nanba and Toshiyuki Takezawa.

DOI： 10.1045/july2012-fukuda

Scopus

researchmap