コロキアムB発表

日時: 6月11日(木)3限(13:30~15:00)


会場: L1

司会: 進藤 裕之
氏家 翔吾 M, 1回目発表 ソーシャル・コンピューティング 荒牧 英治, 渡辺 太郎, 田中 宏季, 若宮 翔子
title: Adverse drug events detection from Japanese medical articles
abstract: In order to ensure the safety of drugs, post-marketing surveillance is conducted for adverse drug events (ADE) caused by the administration of drugs. In particular, pharmaceutical companies monitor medical articles for their own drugs, which is labor-intensive. Thus, a system supporting this duty is expected. In this study, we propose a automated system for drug safety monitoring using natural language processing technology. The proposed system classifies medical articles related to ADEs and extracts ADE-suggesting sentences to support monitoring duty.
language of the presentation: Japanese
発表題目: 日本語医学論文からの有害事象抽出システム
発表概要: 医薬品の安全性確保のため,市販後の医薬品を対象に,医薬品投与によって生じる好ましくない医療上の出来事(有害事象)の監視が行われている.特に製薬会社では,自社の医薬品を対象に医学論文の監視を行っているが,多大なコストがかかっており,情報技術による支援が期待されている.そこで,本研究では自然言語処理技術を用いて,医薬品安全性監視業務の支援システムを提案する.提案システムでは,有害事象に関連する論文の分類を行うとともに,根拠となる文を抽出することで,業務の自動化を目指す.
 
AZUAJE SUAREZ GAMAR IVAN M, 1回目発表 ソーシャル・コンピューティング 荒牧 英治, 渡辺 太郎, 須藤 克仁, 若宮 翔子
title: Writing Assistant based on Image Generation
abstract: Current writing assistants can automatically identify grammar and spelling mistakes. However, they have a disadvantage by not taking semantics into account. A sentence may be grammatically correct, but its meaning may not be the one the user intended. In this research, we propose a system that can provide visual semantic feedback by incorporating text-to-image generation. We describe the components of the system and the model used for image generation. We explain how to use the system and its advantages over traditional writing assistants. Finally, we propose an experimental setting to evaluate its usefulness.
language of the presentation: English
 
GAO ZHIWEI M, 1回目発表 ソーシャル・コンピューティング 荒牧 英治, 渡辺 太郎, 田中 宏季, 若宮 翔子
title: A Preliminary Analysis of Offensive Language Transferability from Social Media to Video Live Streaming
abstract: Given the growing popularity of online games and eSports, the young generation is increasingly enjoying its video live streaming service. Streaming channels are usually combined with chat rooms, where offensive conversations often appear against the streamer or audience. Such offensive languages may cause many serious impacts on a victim’s life and even lead to teen suicide. One method of detecting offensive language is to use deep learning techniques. This research aims to detect offensive language appearing in video live streaming chats. Focusing on Twitch, the most popular live streaming platform, we created a dataset for the task of detecting offensive language. We collected 2000 chat posts across four popular game titles with genre diversity (i.e., competitive, violent, peaceful). Making use of the similarity in offensive languages among social media, we adopt the state-of-the-art models trained over the offensive language on Twitter to our Twitch data (i.e., transfer learning). Our results show that transfer from social media to live streaming is possible. However, the similarity of chat posts or target contents does not help to predict the tranferability with limited data.
language of the presentation: English
 
MUTINDA FAITH WAVINYA M, 2回目発表 ソーシャル・コンピューティング 荒牧 英治, 渡辺 太郎, 若宮 翔子, 田中 宏季
title: Detecting Redundancy in Japanese Electronic Medical Records Using BERT
abstract: Electronic Health Records (EHRs) are widely adopted to record patient's medical progress. They have improved clinical documentation and decision support, because they provide a coordinated, quick, and efficient access to patient records. However, the electronic format of the EHRs encourages copy-and-paste and use of templates. Copy-and-paste introduces redundancy which reduces the quality of the EHR data and makes it difficult to extract relevant information for decision making. Therefore, there is need to minimize redundancy so as to improve the quality of collected EHR data and make clinical decision making easier and efficient. One method for detecting redundant information is to compute the degree of semantic equivalence between clinical texts to remove texts which are highly equivalent. Although STS tasks have been widely studied in the general English domain, there exists very few resources for STS tasks in the clinical domain. In this study, we create a dataset for Japanese clinical STS, and also investigate the level of redundancy in Japanese medical documents. We present a BERT-based model for redundancy detection in Japanese medical documents. Preliminary results from our experiments show that the model can effectively detect redundancy in Japanese medical documents.
language of the presentation: English