| LI XINBAI | D, 中間発表 | ソーシャル・コンピューティング | 荒牧 英治, 安本 慶一, 若宮 翔子, 矢田 竣太郎, She Wan Jou |
|
title: *** K-mQA: Knowledge-enhanced Medical Question Answering ***
abstract: *** While Pre-training Language Models (PLMs) are widely adopted in Question Answering (QA) tasks, such an approach requires domain knowledge when applied in a medical context. Prior work treat medical entities as special tokens when pre-training. However, pre-training of language models has high computational cost. Triples of Knowledge Bases (KBs) describe relations between medical entities, which can readily convert to natural sentences. We assume that inserting sentences converted from triples can provide external knowledge to improve the QA performance without pre-training. In this research, we propose a method, K-mQA, that adopts medical KBs to extract relevant knowledge, and insert knowledge into input text. The experimental results have corroborated that our method K-mQA attains the improvements with the comparison of T5 and entity-enriched medical QA models such as M-cERNIE. *** language of the presentation: *** English *** | |||
| PIERRE JUDE CRENER JUNIOR | M, 2回目発表 | ソーシャル・コンピューティング | 荒牧 英治, 中村 哲, 若宮 翔子, 矢田 竣太郎, She Wan Jou |
|
Title: Time Series Forecasting for Medical Research Trend Abstract: The increasing volume of research literature in the medical field presents an opportunity and a challenge to identify emerging trends for both drug and symptom studies interest. Our research applies Time Series Forecasting to monitor and predict the frequency of mentions concerning specific drugs and symptoms over time. Using a Named Entity Recognition (NER) model pre-trained on medical data, we extract drug and symptom entities from published abstracts to perform our analyses. Our study contributes to data-driven decision-making by using frequency analysis as an indicator of the importance, therapeutic benefits, or potential concerns surrounding certain drugs and symptoms. This approach can help interested parties in preventive decision-making and guide future medical research initiatives. Language of the presentation: English | |||
| 大塚 皇輝 | 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, 矢田 竣太郎, She Wan Jou |
|
title: Prediction of Patient Prognosis Using Specific Health Examination Data
abstract: In recent years, the declining birthrate and aging population have made the tightening of medical resources a problem. Under such circumstances, if the prognosis of patients can be predicted appropriately, it will lead to the optimal allocation of medical resources. For this reason, we used data from the Specified Health Examination to predict patient prognosis. While structured data such as test values are commonly used for analysis, this time we focused on the natural language part of the data. language of the presentation: Japanese 発表題目: 特定健康診査データを用いた患者予後予測 発表概要: 近年少子高齢化等の影響で医療資源の逼迫が問題となっている。このような状況で患者の予後を適切に予測できれば、医療資源の最適配置につなげることができる。そのため今回は特定健康診査データを用いて患者の予後予測を行った。従来検査値等の構造化データを用いて解析を手法が一般的であるが、今回は自然言語部分に注目して解析を行った。 | |||