Language：English   Publishing type：Research paper (scientific journal)  

DOI： https://doi.org/10.5860/crl.83.4.610

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

Many emerging digital library applications rely on automated classifiers that are trained using manually assigned labels. Accurately labeling training data for text classification requires either highly trained coders or multiple annotations, either of which can be costly. Previous studies have shown that there is a quality-quantity trade-off for this labeling process, and the optimal balance between quality and quantity varies depending on the annotation task. In this paper, we present a method that learns to choose between higher-quality annotation that results from dual annotation and higher-quantity annotation that results from the use of a single annotator per item. We demonstrate the effectiveness of this approach through an experiment in which a binary classifier is constructed for assigning human value categories to sentences in newspaper editorials.

DOI： 10.1007/978-3-030-91669-5_19

Repository Public URL： http://hdl.handle.net/2324/4794468

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

Repository Public URL： http://hdl.handle.net/2324/4479593

Language：English   Publishing type：Research paper (scientific journal)  

DOI： 10.1007/978-3-030-34058-2_31

Repository Public URL： http://hdl.handle.net/2324/2927456

Language：English   Publishing type：Research paper (scientific journal)  

DOI： 10.1007/978-3-030-27618-8_5

Repository Public URL： http://hdl.handle.net/2324/2551020

Language：English   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.

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

We propose a method for generating multiple-choice test for an English text selected by a learner and its answer, that are used to make a self-assessment whether the learner comprehends the text after reading it. In our method, the system extracts several important sentences from the text, and replaces one word in each of these sentences with its synonym (if possible). One of these sentences is then selected as a correct optional sentence, while further changes to the polarities or nouns in the remaining sentences are carried out to generate distractor optional sentences for the multiple-choice test. Our method has potential to make extensive reading in English more effective.

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

This paper describes a probabilistic latent variable model that is designed to detect human values such as justice or freedom that a writer has sought to reflect or appeal to when participating in a public debate. The proposed model treats the words in a sentence as having been chosen based on specific values; values reflected by each sentence are then estimated by aggregating values associated with each word. The model can determine the human values for the word in light of the influence of the previous word. This design choice was motivated by syntactic structures such as noun+noun, adjective+noun, and verb+adjective. The classifier based on the model was evaluated on a test collection containing 102 manually annotated documents focusing on one contentious political issue --- Net neutrality, achieving the highest reported classification effectiveness for this task. We also compared our proposed classifier with human second annotator. As a result, the proposed classifier effectiveness is statistically comparable with human annotators.

Repository Public URL： http://hdl.handle.net/2324/1833092

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

Determining the topic of a document is necessary to understand the content of the document efficiently. Latent Dirichlet Allocation (LDA) is a method of analyzing topics. In LDA, a topic is treated as an unobservable variable to establish a probabilistic distribution of words. We can interpret the topic with a list of words that appear with high probability in the topic. This method works well when determining a topic included in many documents having a variety of contents. However, it is difficult to interpret the topic just using conventional LDA when determining the topic in a set of article abstracts found by a keyword search, because their contents are limited and similar. We propose a method to estimate representative words of each topic from an LDA result. Experimental results show that our method provides better information for interpreting a topic than LDA does.

Language：Japanese   Publishing type：Research paper (scientific journal)  

Web文書を利用したコーパスの構築と利用に関して，番号69の論文で述べたプロジェクトを実例として，改正著作権法（2009年改正，2010年施行）の下での取り扱いについて議論を報告した．

Language：Japanese   Publishing type：Research paper (scientific journal)  

Language：Japanese   Publishing type：Research paper (scientific journal)  

Relationship between Errors and Corrections in Verb Selection: Basic Research for Composition Support,
Journal of Natural Language Processing, Vol.18, No.1, pp.3-29

Language：Japanese   Publishing type：Research paper (scientific journal)  

A System Providing Appropriate Alternative Candidates for Japanese Writing using Word Co-occurrence,
T. Nakano, Y. Tomiura, Jpn. J. Educ. Technol., 34(3), pp.181-189 (2010)

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

Language：Japanese   Publishing type：Research paper (scientific journal)  

Method for Selecting Appropriate Sentence from Documents on the WWW for the Open-ended Conversation Dialog System

Language：English   Publishing type：Research paper (scientific journal)  

Repository Public URL： http://hdl.handle.net/2324/1833090

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

Language：Japanese   Publishing type：Research paper (scientific journal)  

Analysis of Answering Method with Probability Conversion for Internet Research

Language：Japanese   Publishing type：Research paper (scientific journal)  

Discernment of Nativeness of English Documents Based on Statistical Hypothesis Testing,
Y. Tomiura, S. Aoki, M. Shibata, K. Yukino,
Journal of Natural Language Processing, Vol.16, No.1, pp.23-46 (2009)

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

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

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

Language：Japanese   Publishing type：Research paper (scientific journal)  

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

Documents Discrimination between Native English Documents and Nonnative Ones Based on Language Identification Technique

S. Aoki, Y. Tomiura, K. Yukino, R. Tanigawa

Information Technology Letters, Vol.5, pp.85-88

Language：Japanese   Publishing type：Research paper (scientific journal)  

A Learning Method of a Layered Neural Network Whose Inputs and Outputs are Symbol Sequences,

M. Motoki, Y. Tomiura, N. Takahashi

IPSJ Journal, Vol.47, No.8, pp.2279--2791

Language：Japanese   Publishing type：Research paper (scientific journal)  

Language Identification Using Low-frequent Byte-strings

K. Yukino, S. Tanaka, Y. Tomiura, H. Matsumoto

IPSJ Journal, Vol.47, No.4, pp.1287--1294

Language：Japanese   Publishing type：Research paper (scientific journal)  

Language：English   Publishing type：Research paper (scientific journal)  

Estimating Satisfactoriness of Selectional Restriction from Corpus without Thesaurus

Y. Tomiura, S. Tanaka, T. Hitaka

ACM Transactions on Asian Language Information Processing, Vol.4, No.4, pp.400--416

Language：Japanese   Publishing type：Research paper (scientific journal)  

Estimation of Nativeness of Documents Based on Skew Divergence,
H. Fujii, Y. Tomiura, S. Tanaka,
Journal of Natural Language Processing, Vol. 12, No. 4, pp.79-96 (2005)

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

Robust Language Identification for Similar Languages and short texts using Low-Frequent Byte Strings,
Kensei Yukino, Shosaku Tanaka, Yoichi Tomiura and Hideki Matsumoto,
Pacific Association for Computational Linguistics 2005 (Pacling 2005),
pp.368-373

Language：Japanese   Publishing type：Research paper (scientific journal)  

Assisting with Translating Collocations Based on the Word Co-occurrence on the Web Texts,
M. Shibata, Y. Tomiura, S. Tanaka,
IPSJ Journal, Vol.46, No.6, pp.1480-1491 (2005)

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

Problems of FGREP Module and Their Solution,
M. Motoki, Y. Tomiura, N. Takahashi,
in Proc. of 3rd IEEE International Conference on Cognitive Informatics, pp.220-227 (2004)

DOI： 10.1109/COGINF.2004.1327479

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

Language：Japanese   Publishing type：Research paper (scientific journal)  

Placement of Nouns in a Multi-Dimensional Space Based on Words' Cooccurrency

Yoichi TOMIURA, Shosaku TANAKA, Toru HITAKA

Transactions of the Japanese Society for Artificial Intelligence, Vol.19, No.1A, pp.1-9

Language：Japanese   Publishing type：Research paper (scientific journal)  

Estimation of Words' Cooccurrency from Corpus

Yoichi TOMIURA, Toru HITAKA

IPSJ Journal, Vol.45, No.1, pp.324-332

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

Language：Japanese   Publishing type：Research paper (scientific journal)  

Context Free Grammar Expressing Dependency Constraint and its Application to Japanese Language

Toshifumi TANABE, Yoichi TOMIURA, Toru HITAKA

IPSJ Journal, Vol.41, No.1, pp.36 - 45

Language：Japanese   Publishing type：Research paper (scientific journal)  

A Parameter Estimation of PCFG Expressing Dependency Constraints on a Sparse Sample

Yoichi TOMIURA, Toru HITAKA

IPSJ Journal, Vol.40, No.11, pp.4055 - 4063

Language：Japanese   Publishing type：Research paper (scientific journal)  

Classification of Syntactic Categories of Nouns by the Scattering of Semantic Categories

Shosaku TANAKA, Yoichi TOMIURA, Toru HITAKA

IPSJ Journal, Vol.40, No.9, pp.3387 - 3396

Language：Japanese   Publishing type：Research paper (scientific journal)  

Language：Japanese   Publishing type：Research paper (scientific journal)  

Semantic Structure of Japanese Noun Phrasese "NP 'no' NP",
Yoichi TOMIURA, Teigo NAKAMURA, Toru HITAKA,
IPSJ Journal, Vol.36, No.6, pp.1441-1448 (1995)

Language：Japanese   Publishing type：Research paper (scientific journal)  

Preference on Common Sense Reasoning and Application to Contextual Processing,
Yoichi TOMIURA, Natsuki ICHIMARU, Toru HITAKA,
IPSJ Journal, Vol.35, No.11, pp.2239 - 2248 (1994)

Language：Japanese   Publishing type：Research paper (scientific journal)  

A New Data Structure for Searching Prefix Words: Prefix-Closed B-tree,
Yoichi TOMIURA, Teigo NAKAMURA, Toru Hitaka,
IPSJ Journal. Vol.35, No.5, pp.779-789 (2007)

Publishing type：Research paper (international conference proceedings)  

Semantic Validity of Japanese Noun Phrases with Adnominal Particles,
Teigo NAKAMURA, Yoichi TOMIURA, Toru HITAKA,
in Proc. of PRICAI'92, Vol.1, No.2, pp.433-437 (1992)

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

Logical Form of Hierarchical Relation on Verbs and Extracting it from Definition Sentences in a Japanese Dictionary,
Yoichi TOMIURA, Teigo NAKAMURA, Toru HITAKA, Sho YOSHIDA,
in Proc. of COLING'92, Vol.2, No.14, pp.574-580 (1992)

Publishing type：Research paper (scientific journal)  

Extracting the Superordinate-Subordinate Relation between Verbs from Definition Sentences in Japanese Dictionaries,
Yoichi TOMIURA, Toru HITAKA, Sho YOSHIDA,
IPSJ Journal, Vol.32, No.1, pp.42 - 49 (1991)

Publisher：Discover Computing  

In an academic paper search to confirm the novelty of a research project, it is important to improve the recall score for the number of search results that users can check to comprehensively collect research papers related to the user’s information need. However, it may be necessary to check papers in the lower ranks to cover the relevant papers comprehensively when using single search method. To improve the recall score for the number of papers that users can check, we considered that it would be effective to integrate ranking results using multiple search methods with different approaches. This is based on the idea that relevant papers that do not appear in the higher ranks for one method can be found using other methods, and that effect would be amplified by integrating more ranking results. As the methods to be integrated, we used the ranking methods in the vector space model, the query likelihood model, and a newly proposed method. Our method is based on a topic-based Boolean search that uses the topic analysis result from latent Dirichlet allocation, and ranked papers in descending order the number of times they are included in each search result. We performed an evaluation using the NTCIR- 1 and - 2 datasets, and confirmed that our topic-based search method showed different trends from rankings based on conventional vector space model and query likelihood model. Furthermore, we showed the best performance by the re-ranking using these three methods comparing other combinations.

DOI： 10.1007/s10791-025-09527-z

Web of Science

Scopus

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

Generative AIs have become widespread and are being used for various purposes. However, a phenomenon called hallucination, in which generative AI outputs plausible lies, is being viewed as a problem. While hallucination is a problem when it is important to be factual, in some cases, such as idea generation, it is not a problem. In fact, in some cases, it may be better to have hallucination occur to come up with surprising and diverse ideas. In this study, we compared several prompts with the aim of getting ChatGPT to output information that differs from the training data for GPT when generating ideas for a fictional historical novel. The results showed that it was more effective and led to generation of more surprising ideas to directly instruct ChatGPT to include content that differs from the facts within the prompt rather than using adversarial prompts that cause hallucinations.

Repository Public URL： https://hdl.handle.net/2324/7362190

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

Early retrieval systems were used to search physical media (e.g., paper) using manually created metadata. Modern ranked retrieval techniques are far more capable, but they require that content be either born digital or digitized. For physical content, searching metadata remains the state of the art. This paper seeks to change that, using a textual-edge graph neural network to learn relations between items from available metadata and from any content that has been digitized. Results show that substantial improvement over the best prior method can be achieved.

DOI： 10.1145/3626772.3657896

Scopus

researchmap

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

Scholars use references in books and articles to materials found in archives as one way of finding those materials, but present systems for archival access do not exploit that information. To change that, the first step is to find archival references in the scholarly literature; that is the focus of this paper. Several classifier designs are compared using a few thousand manually annotated footnotes and endnotes assembled from a large set of open access papers on history. The results indicate that fairly high recall and precision can be achieved.

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

Language：English   Publishing type：Research paper (international conference proceedings)   Publisher：Proceedings - 2022 12th International Congress on Advanced Applied Informatics, IIAI-AAI 2022  

The importance of social media, such as Twitter and Instagram, has increased as a result of COVID-19: with social media, users can easily communicate with other people. Accordingly, social media have produced both new problems and benefits; these have occasionally occurred with upsurges in reactions owing to the huge numbers of social media followers. "Flaming"is a negative aspect of an upsurge: it is a rapid increase in blame, harassment, and insults directed at a person. Conversely, it is possible to use an upsurge to identify popular topics or trends, which is a positive aspect. Predicting the reactions to a social media post is currently very difficult; however, the aim is to prevent the above problems and provide support by forecasting upsurges at an early stage. We observed that such upsurges were affected by the scatter of opinions in the responses. If there are two groups of conflicting opinions, the original social media post tended to receive both sets of opinions. Toward predicting the reactions to a social media post, this study investigate the relationship between the upsurge and scatter of opinions. We quantify the scatter of opinions in the responses, which we term the scatter of replies. We enter that scatter into multiple regression analysis by means of clustering, and we evaluated our reaction prediction method. We found that for most settings, the correlation coefficient between the predicted and actual values exceeded 0.6, which is a moderate level of correlation.

DOI： 10.1109/IIAIAAI55812.2022.00018

Scopus

researchmap

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

DOI： 10.1007/978-3-030-71305-8_40

Repository Public URL： http://hdl.handle.net/2324/4479592

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

DOI： 10.1002/pra2.388

Repository Public URL： http://hdl.handle.net/2324/4479591

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

DOI： 10.1002/pra2.301

Repository Public URL： http://hdl.handle.net/2324/4479586

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

Cost-Effective Learning for Classifying Human Values

Repository Public URL： http://hdl.handle.net/2324/4479590

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

LSPR (Localized Surface Plasmon Resonance) based 2D (2 Dimensional) gas imaging sensor system which can capture spatial distribution of each constituent of mixed gas have been developed. The gas image sensor detects the gas promoted optical changes occurred on the LSPR substrate by CCD camera. Basically, LSPR gas sensor does not have a molecular selectivity, then the identification of gas species is difficult. To overcome the disadvantage, pixelated LSPR substrate based on Au/Ag core-shell structure which has different gas response properties is fabricated by photo-induced metal growth by mask-less exposure system using a commercial video projector.

DOI： 10.1109/SENSORS43011.2019.8956635

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

Our goal is to propose an alternative retrieval system of academic documents based on researcher's behavior in practice. In this study, a questionnaire survey was conducted. Question items were developed from findings in the previous observational study for researcher's behavior. From the results of 46 respondents, the top three elements checked in the search results were title, abstract, and the full-text version. They also checked structure "Introduction" in the full-text rather than other structures when they found previous research in an unfamiliar field. These results indicate that researchers use different ways for selecting documents based on the type of documents they look for.

DOI： 10.1145/3197026.3203867

Repository Public URL： http://hdl.handle.net/2324/1912801

Language：English   Publishing type：Research paper (scientific journal)  

Progress in the molecular biology of olfaction has revealed a close relationship between the structural features of odorants and the response patterns they elicit in the olfactory bulb. Molecular feature-related response patterns, termed odor maps (OMs), may represent information related to basic odor quality. Thus, studying the relationship between OMs and the molecular features of odorants is helpful for better understanding the relationships between odorant structure and odor. Here, we explored the correlation between OMs and the molecular parameters (MPs) of odorants by taking OMs from rat olfactory bulbs and extracting feature profiles of the corresponding odorant molecules. 178 images of glomerular activities in olfactory bulb that are corresponding to odorants were taken from the OdorMapDB, a publicly accessible database. The gray value of each pixel was extracted from the images (178 × 357 pixels) to fabricate an image matrix for each odorant. Forty-six molecular feature parameters were calculated using BioChem3D software, which was used to construct a second matrix for each odorant. Correlation analysis between the two matrixes was first carried out by establishing coefficient maps. Results from hierarchical clustering showed that all parameters could be segregated into seven clusters, and each cluster showed a relatively similar response pattern in the olfactory bulb. Using the information from the OMs and MPs, we mapped odorants in 2D space by incorporating dimension-reducing techniques based on principal component analysis (PCA) and t-distributed stochastic neighbor embedding (t-SNE). Artificial neural network models based on the OM and MP feature values were proposed as a means to identify odorant functional groups. An OM-PCA-based model calibrated via extreme learning machine (ELM) was 94.81% and 93.02% accuracy for the calibration and validation sets, respectively. Similarly, an MP-t-SNE-based model calibrated by ELM was 86.67% and 93.35% accuracy for the calibration set and the validation set, respectively. Thus, this research supports a structure-odor relationship from a data-analysis perspective.

DOI： 10.1016/j.snb.2017.08.024

Repository Public URL： http://hdl.handle.net/2324/1932612

Language：English   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

Repository Public URL： http://hdl.handle.net/2324/1932614

Language：English   Publishing type：Research paper (scientific journal)  

Gas chromatography/olfactometry (GC/O) has been used in various fields as a valuable method to identify odor-active components from a complex mixture. Since human assessors are employed as detectors to obtain the olfactory perception of separated odorants, the GC/O technique is limited by its subjectivity, variability, and high cost of the trained panelists. Here, we present a proof-of-concept model by which odor information can be obtained by machine-learning-based prediction from molecular parameters (MPs) of odorant molecules. The odor prediction models were established using a database of flavors and fragrances including 1026 odorants and corresponding verbal odor descriptors (ODs). Physicochemical parameters of the odorant molecules were acquired by use of molecular calculation software (DRAGON). Ten representative ODs were selected to build the prediction models based on their high frequency of occurrence in the database. The features of the MPs were extracted via either unsupervised (principal component analysis) or supervised (Boruta, BR) approaches and then used as input to calibrate machine-learning models. Predictions were performed by various machine-learning approaches such as support vector machine (SVM), random forest, and extreme learning machine. All models were optimized via parameter tuning and their prediction accuracies were compared. A SVM model combined with feature extraction by BR-C (confirmed only) was found to afford the best results with an accuracy of 97.08%. Validation of the models was verified by using the GC/O data of an apple sample for comparison between the predicted and measured results. The prediction models can be used as an auxiliary tool in the existing GC/O by suggesting possible OD candidates to the panelists and thus helping to give more objective and correct judgment. In addition, a machine-based GC/O in which the panelist is no longer needed might be expected after further development of the proposed odor prediction technique.

DOI： 10.1021/acs.analchem.7b02389

Repository Public URL： http://hdl.handle.net/2324/1932613

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

Academic database systems are vitally important tools for enabling researchers to find relevant, useful articles. Identifying how researchers select documents from search results is an extremely useful measure for improving the functions or interfaces of academic retrieval systems. This study aims to reveal which elements are checked, and in what order, when researchers select from among search results. It consists of two steps: an observational study of search sessions performed by researchers who volunteered, and a questionnaire to confirm whether extracted elements and patterns are used. This article reports findings from the observational study and introduces questions we developed based on the study. In the observational study we obtained data on nine participants who were asked to search for documents using information retrieval systems. The search sessions were recorded using a voice recorder and by capturing screen images. The participants were also asked to state which elements they checked in selecting documents, along with the reasons for their selections. Three patterns of order of checking were found. In pattern 1, seven researchers used titles and abstracts as the primary elements. In pattern 2, the others used titles and then accessed the full text before making a decision on their selection. In pattern 3, one participant searched for images and accessed the full text from the link in those pictures. We also found participants used novel elements for selecting. We subsequently developed items for a questionnaire reflecting the findings.

DOI： 10.1007/978-3-319-70232-2_20

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

This paper describes the creation of a corpus of newspaper articles about the Fukushima nuclear disaster, a coding frame for content analysis of human values expressed or reflected in that corpus and preliminary results for automation of the content analysis. Understanding the human values that motivate sentiment towards an idea can help to characterize the basis for that sentiment, and this work is a first step toward applying that approach to positions on controversial events reported in the news.

DOI： 10.1002/pra2.2017.14505401127

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

Language：English   Publishing type：Research paper (scientific journal)  

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

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

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

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

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

In this study, we aim to extract key expressions that indicate the important sentence describing the originalities or contributions from article abstracts. The expense of searching academic information increases because of increases in the number of articles, discipline subdivisions, and promotion of interdisciplinary research. Improving the extraction and presentation of the main points from article abstracts will contribute to reducing academic information search expenses. We extracted pseudo-important sentences from each article abstract based on the ratio of the number of words in the identified sentence that appear in the article title to the number of all words in the sentence. After that, we evaluated the ratio of the number of the pseudo-important sentences including each Ngram to the number of all sentences including that N-gram. We then extracted the N-grams with a ratio as high as the key expressions.

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

From the early 1960s, English for the Specific Purposes (ESP) has become one of the prominent fields of English teaching (e.g. Hutchinson and Waters, 1987). It aims to provide learners with academic English skills in a context of tertiary education. In terms of English vocabulary, the Academic Word List (AWL) contains 570 word families selected from academic texts (Coxhead, 2000). The list has been widely known to both academic English teachers and researchers, but does not cover context-bound or topic-dependent vocabulary frequently used in each discipline and sub-discipline.

The most significant words in each academic context differ depending on the discipline, university, laboratory or researcher. Therefore, it is necessary to compile a large corpus which covers a wide range of academic domains to create a more appropriate word list for researchers in different academic fields. What is more, it is important for language learners to acquire discipline-specific expressions to achieve native-like performance in an ESP context.

The purpose of this study is to identify discipline-specific expressions, especially multi-word expressions, using institutional repositories and natural language processing techniques. An institutional repository is an online collection of the intellectual output of a research institution. The recent trend of constructing institutional repository in Japan allows researchers to share the same academic resources. The present study demonstrates the effectiveness of creating a multi-word list specific to an institution.

We used a gapped n-gram approach for the identification of discipline-specific expressions. An ‘n-gram’ is a contiguous sequence of items, such as words or part-of speech tags, and a ‘gapped n-gram,’ or ‘skipgram,’ is a refinement of the n-gram approach designed to detect non-contiguous item associations (e.g. Cheng, Greaves, and Warren, 2006). Our algorithm for extracting gapped n-grams is based on the method proposed by Kozawa, Sakai, Sugiki, and Matsubara (2010). First, we filtered the texts in the institutional repository by the number of words to exclude too short or too long writings. Second, we used TreeTagger, an automatic chunking program, to identify the basic phrase and clause structures. Third, we detected general patterns of multi-word expressions by lemmatizing words except for participial verbs and by replacing determiners and cardinal numbers into part-of-speech tags. Finally, we counted the number of gapped n-grams occurred in the writings collected from each academic disciplines.

This method enabled us to effectively detect frequent patterns characteristic of each discipline. We applied the method to the QIR, the institutional repository of Kyushu University, using the repository of the Faculty of Information Science and Electrical Engineering (ISEE), and identified discipline-specific expressions for each faculty at the university as an example of the automatic detection of gapped n-grams. We collected 229 writings in the QIR, and successfully obtained 1,061 frequent phrasal expressions specific to ISEE. We expect this study contributes to teaching English for the Specific Purposes and the application of digital library to linguistic research.

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

This paper describes an effort to improve identification of human values that are directly or indirectly invoked within the prepared statements of witnesses before legislative and regulatory hearings. We automatically code human values at the sentence level using supervised machine learning techniques trained on a few thousand annotated sentences. To simulate an actual situation, we treat a quarter of the data as labeled for training and the remaining three quarters of the data as unlabeled for test. We find that augmenting the feature space using a combination of lexical and statistical co-occurrence evidence can yield about a 6% relative improvement in F1 using a Support Vector Machine classifier.

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

Language：Japanese   Publishing type：Research paper (scientific journal)  

Language：Japanese   Publishing type：Research paper (scientific journal)  

A Platform for Context-aware Notification Service with Location-aware P2P Network

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

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

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

Language：Japanese   Publishing type：Research paper (scientific journal)  

Use of Fast Algorithm for Adaptive Background Modeling with Parzen Density Estimation to Detect Objects,
T. Tanaka, A. Shimada, D. Arita, R. Taniguchi, Y. Tomiura,
The Journal of The Institute of Image Information and Television Engineers,
Vol.62, No.12, pp.2045-2052

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

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

Developing a Dialog System for New Idea Generation Support

M. Shibata, Y. Tomiura, H. Matsumoto, T. Nishiguchi, K. Yukino, A. Hino

Proc. of 21st International Conference on Computer Processing of Oriental Language

Language：Japanese   Publishing type：Research paper (scientific journal)  

Examination about Usefulness of the Method for Translating Collocations Based on the Word Co-occurence on the Web Documents

M.Shibata, S.Tanaka, Y.Tomiura

Research Reports on Information Science and Electrical Engineering of Kyushu University, Vol.10, No.1, pp.45--49

Repository Public URL： http://hdl.handle.net/2324/1516060

Language：Japanese   Publishing type：Research paper (scientific journal)  

Implementation of Writing Support System for Translating Collocations Based on the Word Co-occurrence on the Web Documents

M.Shibata, S.Tanaka, Y.Tomiura

Research Reports on Information Science and Electrical Engineering of Kyushu University, Vol.10, No.1, pp.39--44

Repository Public URL： http://hdl.handle.net/2324/1516059

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

A System for Extensive Slash Reading Using Web,
Shosaku TANAKA, Yoichi TOMIURA, Kensei YUKINO,
An Interactive Workshop on Language e-Learning (IWLeL2004), pp.133-138

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

Discrimination of Native/Non-native Documents Based on Skew Divergence,
H. Fujii, S. Tanaka, Y. Tomiura,
Information Technology Letters, Vol.3, pp.81-83 (2004)

Language：Japanese   Publishing type：Research paper (scientific journal)  

Estimation of Synonym's Mappings and its Application to Compiling Translational Dictionary using English,
S. Tanaka, Y. Marubayashi, Y. Tomiura,
Research Reports on Information Science and Electrical Engineering of Kyushu University, Vol.9, No.2, pp.73-78

Repository Public URL： http://hdl.handle.net/2324/1516050

Language：Japanese   Publishing type：Research paper (scientific journal)  

Generation of Texts for Slash Reading Based on Chunking Process

S.Tanaka, Y.Tomiura, S.Yamamoto

Annual Report on Computing and Communication Center, Vol.4, pp.1--8

Language：Japanese   Publishing type：Research paper (scientific journal)  

Parsing Method for the Translation Grammar

M. Shibata, Y. Tomiura, T. Hitaka

Research Reports on Information Science and Electrical Engineering of Kyushu University, Vol.9, No.1, pp.31--36

Repository Public URL： http://hdl.handle.net/2324/1515923

Language：Japanese   Publishing type：Research paper (scientific journal)  

Machine Translation using Translation Grammar

Masahiro SHIBATA, Toru HITAKA, Yoichi TOMIURA

Research Reports on Information Science and Electrical Engineering of Kyushu University, Vol.8, No.1, pp.61-66

Repository Public URL： http://hdl.handle.net/2324/1515819

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

Judgement on Validity of Candidates of Translations Dirived via English Using Sets of Synonyms

Shosaku TANAKA, Yoichi TOMIURA

Forum on Information Technology, Information Technology Letters, pp.75-76

Language：Japanese   Publishing type：Research paper (scientific journal)  

mplementation of The Machine Translation System Using The Translation Context Free Grammar

Masahiro Shibata, Toru Hitaka, Yoichi Tomiura

Annual Report of Computing and Communication Center Kyushu University, No.2, pp.71-79

Language：Japanese   Publishing type：Research paper (scientific journal)  

An Efficient Parsing Algorithm for CFG with Robust Dependency Constraints

Toushinbatto D., Yoichi TOMIURA, Toru HITAKA

Research Reports on Information Science and Electrical Engineering of Kyushu University, Vol.5, No.2, pp223-227

Repository Public URL： http://hdl.handle.net/2324/1515687

Language：Japanese   Publishing type：Research paper (scientific journal)  

Disambiguation of Syntactic Structures of Japanese Noun Phrases "NP 'no' NP" by PCFG Having Cooccurence Constraints Embedded

Shosaku TANAKA, Yoichi TOMIURA, Toru HITAKA

Research Reports on Information Science and Electrical Engineering of Kyushu University, Vol.5, No.1, pp.69-74

Repository Public URL： http://hdl.handle.net/2324/1513732

Language：Japanese   Publishing type：Research paper (scientific journal)  

Estimating Semantic Relation of Japanese Noun Phrases based on k-Nearest Neighbor Estimation

Shosaku TANAKA, Yasuo YANASE, Yoichi TOMIURA, Toru HITAKA

Research Reports on Information Science and Electrical Engineering of Kyushu University, Vol.4, No.2, pp.159-164

Repository Public URL： http://hdl.handle.net/2324/1515753

Event date： 2024.2 - 2024.3

Language：Japanese   Presentation type：Oral presentation (general)  

Country：Japan  

Event date： 2024.2 - 2024.3

Language：Japanese   Presentation type：Oral presentation (general)  

Country：Japan  

Event date： 2023.12

Language：English  

Country：Japan  

In recent years, huge amounts of research data have been generated, and it has become important to search them efficiently and accurately in order to make use of research data. Existing search engines and keyword-based search methods require users to enter appropriate keywords or phrases, and it is difficult to obtain satisfactory results if users do not have detailed information about the desired data. In this study, we investigated whether ChatGPT could be used to reach the desired research data by users who are not familiar with them. Specifically, we investigated whether users could find the research data cited in a research paper by entering the abstract of the paper into ChatGPT and then asking for the data necessary to write the research paper. The results showed that research data could be found in 65% of the cases, confirming that the use of ChatGPT increases the discoverability of research data.

Event date： 2021.12

Language：English   Presentation type：Oral presentation (general)  

Venue：Online   Country：Other  

Many emerging digital library applications rely on automated classifiers that are trained using manually assigned labels. Accurately labeling training data for text classification requires either highly trained coders or multiple annotations, either of which can be costly. Previous studies have shown that there is a quality-quantity trade-off for this labeling process, and the optimal balance between quality and quantity varies depending on the annotation task. In this paper, we present a method that learns to choose between higher-quality annotation that results from dual annotation and higher-quantity annotation that results from the use of a single annotator per item. We demonstrate the effectiveness of this approach through an experiment in which a binary classifier is constructed for assigning human value categories to sentences in newspaper editorials.

Repository Public URL： http://hdl.handle.net/2324/4794468

Event date： 2019.11

Language：English  

Venue：Kuala Lumpur   Country：Malaysia  

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.

Repository Public URL： http://hdl.handle.net/2324/2927456

Event date： 2019.8

Language：English   Presentation type：Oral presentation (general)  

Venue：Linz   Country：Austria  

Repository Public URL： http://hdl.handle.net/2324/2551020

Event date： 2016.6

Language：English  

Venue：Yokohama   Country：Japan  

Repository Public URL： http://hdl.handle.net/2324/1833089

Event date： 2015.2

Language：English   Presentation type：Oral presentation (general)  

Venue：Lisbon   Country：Portugal  

We propose a method for generating multiple-choice test for an English text selected by a learner and its answer, that are used to make a self-assessment whether the learner comprehends the text after reading it. In our method, the system extracts several important sentences from the text, and replaces one word in each of these sentences with its synonym (if possible). One of these sentences is then selected as a correct optional sentence, while further changes to the polarities or nouns in the remaining sentences are carried out to generate distractor optional sentences for the multiple-choice test. Our method has potential to make extensive reading in English more effective.

Event date： 2014.12

Language：English   Presentation type：Oral presentation (general)  

Venue：Shanghai   Country：China  

This paper describes a probabilistic latent variable model that is designed to detect human values such as justice or freedom that a writer has sought to reflect or appeal to when participating in a public debate. The proposed model treats the words in a sentence as having been chosen based on specific values; values reflected by each sentence are then estimated by aggregating values associated with each word. The model can determine the human values for the word in light of the influence of the previous word. This design choice was motivated by syntactic structures such as noun+noun, adjective+noun, and verb+adjective. The classifier based on the model was evaluated on a test collection containing 102 manually annotated documents focusing on one contentious political issue --- Net neutrality, achieving the highest reported classification effectiveness for this task. We also compared our proposed classifier with human second annotator. As a result, the proposed classifier effectiveness is statistically comparable with human annotators.

Event date： 2014.3

Language：English   Presentation type：Oral presentation (general)  

Venue：Novotel Barcelona Sant Joan Despi (Barcelona)   Country：Japan  

Determining the topic of a document is necessary to understand the content of the document efficiently. Latent Dirichlet Allocation (LDA) is a method of analyzing topics. In LDA, a topic is treated as an unobservable variable to establish a probabilistic distribution of words. We can interpret the topic with a list of words that appear with high probability in the topic. This method works well when determining a topic included in many documents having a variety of contents. However, it is difficult to interpret the topic just using conventional LDA when determining the topic in a set of article abstracts found by a keyword search, because their contents are limited and similar. We propose a method to estimate representative words of each topic from an LDA result. Experimental results show that our method provides better information for interpreting a topic than LDA does.

Event date： 2011.7

Presentation type：Oral presentation (general)  

Venue：International Islamic University   Country：Malaysia  

Event date： 2010.3

Presentation type：Oral presentation (general)  

Venue：Porto   Country：Portugal  

Event date： 2010.3

Presentation type：Oral presentation (general)  

Country：Egypt  

Event date： 2009.10

Presentation type：Oral presentation (general)  

Venue：青山学院大学（東京都）   Country：Japan  

Event date： 2009.9

Presentation type：Oral presentation (general)  

Country：Japan  

Event date： 2008.1

Venue：Las Vegas   Country：United States  

Analysis of Answering Method with Probability Conversion for Internet Research,
Atsushi TAGAMI, Chikara SAKAKI, Teruyuki HASEGAWA, Shigehiro ANO, Yoichi TOMIURA,
Fifth IEEE Consumer Communications & Networking Conference (CCNC'08)

Language：English  

Country：Other  

Repository Public URL： http://hdl.handle.net/2324/4479593

Presentation type：Oral presentation (general)  

Country：United States  

Presentation type：Oral presentation (general)  

Venue：九州大学（福岡市）   Country：Japan  

Presentation type：Oral presentation (general)  

Country：Japan  

Presentation type：Oral presentation (general)  

Country：China  

A Method for Retrieving Translations of Collocation in Web Data,
M. Shibata, Y. Tomiura, S Tanaka,
IJCNLP-04 Satellite Symposium

Presentation type：Oral presentation (general)  

Venue：ビクトリア大学   Country：Canada  

Problems of FGREP Module and Their Solution,
M. Motoki, Y. Tomiura, N. Takahashi,
3rd IEEE International Conference on Cognitive Informatics,

Presentation type：Oral presentation (general)  

Venue：明星大学   Country：Japan  

Robust Language Identification for Similar Languages and short texts using Low-Frequent Byte Strings,
Kensei Yukino, Shosaku Tanaka, Yoichi Tomiura and Hideki Matsumoto,
Pacific Association for Computational Linguistics 2005 (Pacling 2005)

Presentation type：Oral presentation (general)  

Venue：福岡大学   Country：Japan  

Country：Singapore  

Presentation type：Oral presentation (general)  

Country：Finland  

Event date： 2024.12

Language：English   Presentation type：Poster presentation  

Venue：Bandar Sunway   Country：Malaysia  

Generative AIs have become widespread and are being used for various purposes. However, a phenomenon called hallucination, in which generative AI outputs plausible lies, is being viewed as a problem. While hallucination is a problem when it is important to be factual, in some cases, such as idea generation, it is not a problem. In fact, in some cases, it may be better to have hallucination occur to come up with surprising and diverse ideas. In this study, we compared several prompts with the aim of getting ChatGPT to output information that differs from the training data for GPT when generating ideas for a fictional historical novel. The results showed that it was more effective and led to generation of more surprising ideas to directly instruct ChatGPT to include content that differs from the facts within the prompt rather than using adversarial prompts that cause hallucinations.

Repository Public URL： https://hdl.handle.net/2324/7362190

Event date： 2024.7

Language：English   Presentation type：Poster presentation  

Venue：Washington D.C.   Country：United States  

DOI： 10.1145/3626772.3657896

Event date： 2023.12

Language：English  

Venue：Taipei   Country：Taiwan, Province of China  

Scholars use references in books and articles to materials found in archives as one way of finding those materials, but present systems for archival access do not exploit that information. To change that, the first step is to find archival references in the scholarly literature; that is the focus of this paper. Several classifier designs are compared using a few thousand manually annotated footnotes and endnotes assembled from a large set of open access papers on history. The results indicate that fairly high recall and precision can be achieved.

Event date： 2023.10 - 2023.11

Language：English  

Venue：Vienna   Country：Austria  

Visualizing invisible gas molecules can be a great help to our lives. At present, gas sensors can already visualize the spatial distribution of gas mixture, however, the visualization of a specific gas requires further analysis of the measurement data. In this study, matrix decomposition is used to analyze the measurement data of localized surface plasmon resonance (LSPR) gas sensor. To satisfy the linear relationship between the concentration of gas and the output of the device required for applying matrix decomposition, we formulated a procedure for processing the measurement data instead of using them directly. To obtain the diffusion trace of a specific gas, we designed a method to obtain the characteristic output of the specific gas, then by using the characteristic output as the known information, the corresponding diffusion trace can be estimated better through the matrix decomposition algorithm. We used the designed method to analyze the measurement data, and the results show that our method can obtain the spatial distribution of some gas.

Event date： 2022.7

Language：English   Presentation type：Oral presentation (general)  

Country：Japan  

Event date： 2020.10 - 2020.11

Language：English  

Venue：Online   Country：Other  

Repository Public URL： http://hdl.handle.net/2324/4479591

Event date： 2020.10 - 2020.11

Language：English   Presentation type：Oral presentation (general)  

Venue：Online   Country：Other  

Repository Public URL： http://hdl.handle.net/2324/4479586

Event date： 2019.11

Language：English   Presentation type：Oral presentation (general)  

Venue：Kuala Lumpur   Country：Malaysia  

Event date： 2019.11

Language：English  

Venue：Kuala Lumpur   Country：Malaysia  

Event date： 2019.11

Language：English  

Venue：Kuala Lumpur   Country：Malaysia  

Event date： 2019.10

Language：English   Presentation type：Oral presentation (general)  

Venue：Montreal   Country：Canada  

Event date： 2019.5

Language：English   Presentation type：Oral presentation (general)  

Venue：Fukuoka   Country：Japan  

If we can quickly recognize the distribution of dangerous gases, it will be useful in places such as disaster scene. Localized surface plasmon resonance (LSPR) gas sensor is known as a gas sensor with high response / recovery speed and high spatial resolution. However, the general LSPR gas sensor does not have a molecular selectivity and it is difficult to identify the gas species. We made gas selected pixelated LSPR substrate based on Au/Ag core-shell structure by photo-induced growth by exposure system using the photomask.

Event date： 2019.3 - 2019.4

Language：English   Presentation type：Oral presentation (general)  

Venue：Washington   Country：United States  

Event date： 2019.3 - 2019.4

Language：English   Presentation type：Oral presentation (general)  

Venue：Washington   Country：United States  

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.

Event date： 2018.12 - 2019.12

Language：English  

Venue：Fukuoka   Country：Japan  

Event date： 2018.12

Language：English  

Venue：Fukuoka   Country：Japan  

Event date： 2018.11

Language：English   Presentation type：Oral presentation (general)  

Venue：Hamilton   Country：New Zealand  

Repository Public URL： http://hdl.handle.net/2324/2230669

Event date： 2018.11

Language：English  

Venue：Hamilton   Country：New Zealand  

Event date： 2018.11

Language：English   Presentation type：Oral presentation (general)  

Venue：Hamilton   Country：New Zealand  

Repository Public URL： http://hdl.handle.net/2324/2230666

Event date： 2018.10

Language：English   Presentation type：Oral presentation (general)  

Venue：全州   Country：Korea, Republic of  

Event date： 2018.6

Language：English  

Country：United States  

Our goal is to propose an alternative retrieval system of academic documents based on researcher’s behavior in practice. In this study, a questionnaire survey was conducted. Question items were developed from findings in the previous observational study for researcher’s behavior. From the results of 46 respondents, the top three elements checked in the search results were title, abstract, and the full-text version. They also checked structure “Introduction” in the full-text rather than other structures when they found previous research in an unfamiliar field. These results indicate that researchers use different ways for selecting documents based on the type of documents they look for.

Repository Public URL： http://hdl.handle.net/2324/1912801

Event date： 2017.12

Language：English   Presentation type：Oral presentation (general)  

Venue：Singapore   Country：Singapore  

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.

Repository Public URL： http://hdl.handle.net/2324/1932614

Event date： 2017.11

Language：English   Presentation type：Oral presentation (general)  

Venue：Bangkok   Country：Thailand  

Academic database systems are vitally important tools for enabling researchers to find relevant, useful articles. Identifying how researchers select documents from search results is an extremely useful measure for improving the functions or interfaces of academic retrieval systems. This study aims to reveal which elements are checked, and in what order, when researchers select from among search results. It consists of two steps: an observational study of search sessions performed by researchers who volunteered, and a questionnaire to confirm whether extracted elements and patterns are used. This article reports findings from the observational study and introduces questions we developed based on the study. In the observational study we obtained data on nine participants who were asked to search for documents using information retrieval systems. The search sessions were recorded using a voice recorder and by capturing screen images. The participants were also asked to state which elements they checked in selecting documents, along with the reasons for their selections. Three patterns of order of checking were found. In pattern 1, seven researchers used titles and abstracts as the primary elements. In pattern 2, the others used titles and then accessed the full text before making a decision on their selection. In pattern 3, one participant searched for images and accessed the full text from the link in those pictures. We also found participants used novel elements for selecting. We subsequently developed items for a questionnaire reflecting the findings.

Event date： 2017.10 - 2017.11

Language：English  

Venue：Arlington   Country：United States  

Event date： 2016.9

Language：English  

Venue：University of Zadar   Country：Croatia  

Event date： 2016.7

Language：English   Presentation type：Oral presentation (general)  

Venue：Kumamoto   Country：Japan  

Event date： 2016.6

Language：English  

Venue：Yokohama   Country：Japan  

Event date： 2015.12

Language：English  

Venue：Seoul   Country：Korea, Republic of  

Event date： 2015.12

Language：English  

Venue：Seoul   Country：Korea, Republic of  

Event date： 2015.7

Language：English   Presentation type：Oral presentation (general)  

Venue：Okayama   Country：Japan  

Event date： 2015.7

Language：English   Presentation type：Oral presentation (general)  

Venue：Okayama   Country：Japan  

Event date： 2015.3

Language：English   Presentation type：Oral presentation (general)  

Venue：California   Country：United States  

Event date： 2014.12

Language：Japanese   Presentation type：Oral presentation (general)  

Venue：国立情報学研究所   Country：Japan  

Event date： 2014.11

Language：Japanese   Presentation type：Oral presentation (general)  

Venue：梅花女子大学   Country：Japan  

Event date： 2014.9

Language：Japanese   Presentation type：Oral presentation (general)  

Venue：筑波大学   Country：Japan  

Event date： 2014.9

Language：Japanese   Presentation type：Oral presentation (general)  

Venue：筑波大学   Country：Japan  

Event date： 2014.8 - 2014.9

Language：English   Presentation type：Oral presentation (general)  

Venue：Kitakyushu-shi   Country：Japan  

Event date： 2014.8 - 2014.9

Language：English   Presentation type：Oral presentation (general)  

Venue：Kitakyushu-shi   Country：Japan  

a

Event date： 2014.3

Language：English   Presentation type：Oral presentation (general)  

Venue：University of Western   Country：Australia  

Event date： 2014.3

Language：Japanese  

Venue：北海道大学   Country：Japan  

Event date： 2013.11

Language：Japanese   Presentation type：Oral presentation (general)  

Venue：屋久島 環境文化村センター   Country：Japan  

Event date： 2013.9

Language：Japanese   Presentation type：Oral presentation (general)  

Country：Japan  

Event date： 2013.9

Language：English   Presentation type：Oral presentation (general)  

Venue：Keio University   Country：Japan  

This paper describes an effort to improve identification of human values that are directly or indirectly invoked within the prepared statements of witnesses before legislative and regulatory hearings. We automatically code human values at the sentence level using supervised machine learning techniques trained on a few thousand annotated sentences. To simulate an actual situation, we treat a quarter of the data as labeled for training and the remaining three quarters of the data as unlabeled for test. We find that augmenting the feature space using a combination of lexical and statistical co-occurrence evidence can yield about a 6% relative improvement in F1 using a Support Vector Machine classifier.

Event date： 2013.3

Language：Japanese   Presentation type：Oral presentation (general)  

Venue：名古屋大学   Country：Japan  

Event date： 2012.12

Presentation type：Oral presentation (general)  

Venue：龍谷大学　大宮キャンパス   Country：Japan  

Event date： 2012.1

Presentation type：Oral presentation (general)  

Venue：奄美市立奄美博物館   Country：Japan  

Event date： 2012.1

Presentation type：Oral presentation (general)  

Venue：福岡大学　六本松キャンパスセミナーハウス   Country：Japan  

Event date： 2012.1

Presentation type：Oral presentation (general)  

Venue：福岡大学　六本松キャンパスセミナーハウス   Country：Japan  

Event date： 2011.11

Presentation type：Symposium, workshop panel (public)  

Venue：九州大学　箱崎キャンパス   Country：Japan  

Event date： 2011.10

Presentation type：Oral presentation (general)  

Venue：京都外国語大学   Country：Japan  

Event date： 2011.5

Presentation type：Oral presentation (general)  

Venue：同志社大学 室町キャンパス（京都府京都市）   Country：Japan  

近年、非母語話者が書く英語科学論文（以後、論文）と母語話者の論文の分類を通して、両者の様々な言語的差異を抽出することが試みられている。本発表では、まず、論文中の談話表現に着目し、その頻度を素性の候補とするランダムフォレストに基づく分類器を構築する。その分類精度は約90%で類似研究の中でも高いものであった。そして、構築された分類器の素性を分析することで、非母語話者の論文に特徴的な談話表現を抽出する。

Event date： 2010.4

Presentation type：Oral presentation (general)  

Country：Australia  

Event date： 2009.9

Presentation type：Oral presentation (general)  

Country：Japan  

Event date： 2009.4

Presentation type：Oral presentation (general)  

Venue：神戸大学（兵庫県）   Country：Japan  

Event date： 2007.10

Presentation type：Oral presentation (general)  

Venue：沖縄   Country：Japan  

Event date： 2007.10

Presentation type：Oral presentation (general)  

Venue：沖縄   Country：Japan  

Event date： 2007.9

Presentation type：Oral presentation (general)  

Venue：アルカディア市ヶ谷（東京、私学会館）   Country：Japan  

Event date： 2007.7

Presentation type：Oral presentation (general)  

Country：Japan  

Language：English  

Country：Other  

Repository Public URL： http://hdl.handle.net/2324/4479590

Language：Japanese  

Country：Japan  

Language：English   Presentation type：Oral presentation (general)  

Venue：Online   Country：Other  

Repository Public URL： http://hdl.handle.net/2324/4479592

Language：Japanese   Presentation type：Oral presentation (general)  

Country：Japan  

Presentation type：Oral presentation (general)  

Country：Japan  

Acquisition of Semantic Relations of Japanese Noun Phrases "NP 'no' NO" by using Statistical Property

Shosaku TANAKA, Yoichi TOMIURA, Toru HITAKA

Technical Report of IEICE, NLC 98-4

Presentation type：Oral presentation (general)  

Country：Japan  

A Parameter Estimation of a Probabilistic Context Free Grammar on a Sparse Sample

Yoichi TOMIURA, Takeshi NISHIDA, Toru HITAKA

Tecnical Report of IEICE, NLC 98-12

Presentation type：Oral presentation (general)  

Country：Japan  

How to Strengthen a Context Free Grammar Expressing Dependency Constraint

D. Toushinbatto, Yoichi TOMIURA, Toru HITAKA

IPSJ SIG Notes, Vol.98, No.99 (98-NL-128)

Presentation type：Oral presentation (general)  

Country：Japan  

Discriminant Analysis with Incomplete Data and its Applicatioin to Estimation of Words' Cooccurrency

Yoichi TOMIURA, Shosaku TANAKA, Toru HITAKA

Technical Report of IEICE, NLC 2000-76

Presentation type：Oral presentation (general)  

Country：Japan  

Extraction of Candidates having Semantic Relations of Japanese Noun Phrase "NP1 'no' NP2" by Dependency Structures

Shosaku TANAKA, Yoichi TOMIURA, Toru HITAKA

Technical Report of IEICE, NLC 2001-46

Presentation type：Oral presentation (general)  

Country：Japan  

PP-attachment Ambiguity Resolution Using a Neural Network wiht Modified FGREP Method

TAKAHASHI Naoto, MOTOKI Minoru, SHIMAZU Yoshio, TOMIURA Yoichi, HITAKA Toru

the 2nd Workshop on Natural Language Processing and Neural Networks (post-conference workshop of NLPRS2001)

Presentation type：Oral presentation (general)  

Country：Japan  

Judgement on Validity of Candidates of Translations Derived via English Using Sets of Synonyms

Shosaku TANAKA, Yoichi TOMIURA

Forum on Information Technology

Presentation type：Oral presentation (general)  

Venue：徳島大学（徳島市）   Country：Japan  

Placement of Nouns in n-Dimensional Space Based on Cooccurrency

Yoichi TOMIURA, Sho TANAKA, Toru HITAKA

IPSJ SIG Notes, Vol.2003, No.23 (2003-NL-154)

Presentation type：Oral presentation (general)  

Venue：同志社大学（京都府京田辺市）  

Discrimination of Native/Non-native Documents Based on Skew Divergence,
H. Fuji, S. Tanaka, Y. Tomiura,
Forum on Information Technology 2004 (FIT2004)

Venue：早稲田大学   Country：Japan  

A System for Extensive Slash Reading Using Web,
Shosaku TANAKA, Yoichi TOMIURA, Kensei YUKINO,
The 18th Pacific Asia Conference on Language, Information and Computation (PACLIC 18)

Country：Japan  

Tracing Japanese EFL Learners' Development in Productive Vocabulary

Venue：佐賀大学   Country：Japan  

Country：Japan  

Presentation type：Oral presentation (general)  

Venue：宮崎大学   Country：Japan  

Presentation type：Oral presentation (general)  

Venue：宮崎大学   Country：Japan  

Presentation type：Oral presentation (general)  

Venue：宮崎大学   Country：Japan  

Presentation type：Oral presentation (general)  

Venue：関西大学   Country：Japan  

Venue：龍谷大学   Country：Japan  

Presentation type：Oral presentation (general)  

Venue：九州大学（福岡）   Country：Japan  

Presentation type：Oral presentation (general)  

Venue：Hanoi   Country：Viet Nam  

Language：English   Presentation type：Oral presentation (general)  

DOI： 10.1109/IIAIAAI55812.2022.00018

Language：Japanese  

Language：Japanese   Publishing type：Article, review, commentary, editorial, etc. (scientific journal)  

Language：Japanese   Publishing type：Article, review, commentary, editorial, etc. (scientific journal)  

Type：Competition, symposium, etc. 

Type：Competition, symposium, etc. 

Type：Competition, symposium, etc. 

Type：Competition, symposium, etc. 

Type：Competition, symposium, etc. 

Type：Competition, symposium, etc. 

Type：Competition, symposium, etc. 

Type：Competition, symposium, etc. 

Type：Competition, symposium, etc. 

Type：Competition, symposium, etc. 

Type：Competition, symposium, etc. 

Type：Competition, symposium, etc. 

Type：Competition, symposium, etc. 

Type：Competition, symposium, etc. 

Type：Competition, symposium, etc. 

Type：Competition, symposium, etc. 

Type：Competition, symposium, etc. 

Type：Competition, symposium, etc. 

Type：Competition, symposium, etc. 

Number of participants：700

Type：Competition, symposium, etc. 

Type：Competition, symposium, etc. 

Type：Competition, symposium, etc. 

Type：Competition, symposium, etc. 

Type：Competition, symposium, etc.