108年智慧應答機器人研發年智慧應答機器人研發

全文

(1)109 年度. 108 年智慧應答機器人研發.

(2) 109 年度. 2019 Development of Intelligence Response Robot 主管單位：經濟部水利署蔡孟涵 1 楊亦東 1 莊子毅 1 蔡芸琤 2 Tsai, Meng-Han Yang, I-Tung Chuang , Tzu-Yi Tsai, Yun-Cheng 邱睿宏 3 Chiu, Jui-Hung 1. 許倍銘 3 Hsu, Pei-Ming. 國立臺灣科技大學營建工程系 2 東吳大學巨量資料管理學院 3 酷必資訊股份有限公司.

(3) 109 年度. 摘要「109 年智慧應答機器人研發」(以下簡稱本計畫)延續 108 年之開發成果及經驗，以水利防災專業人員為目標使用者，發展智慧應答機器人，透過輸入文字或點擊按鈕的方式，使用者可取得防災所需的資料。本計畫執行四大工作項目的系統研發及專業服務包含：系統維運、應用探討、精進研發及協助辦理教育訓練。截至 109 年 11 月 30 日本計畫完成工作項目如下所述。在智慧應答機器人系統維運部份，已完成防汛應變資料庫盤點及更新，包含：氣象整備、應變、復原、其他、意見回饋等六大類共 437 筆資訊。並針對 59 筆靜態防災資料進行資料更新及視覺化設計。另定期更新與測試智慧應答機器人之平台及系統，已確保正常運作。亦已於黃蜂颱風等 9 場事件應變期間，進駐應變中心及隨時監控系統穩定性，以確保系統及操作問題能即時解決。在智慧防災應答機器人應用探討部分，本計畫每週提供使用者行為之資料，做為調整及更新智慧應答機器人系統之參考。並增建「抽水機調度資訊」等 4 組關鍵字資訊，以強化智慧防災應答機器人資料的完整性。本計畫於 0522 豪雨等 4 起事件應變期間針對應變人員與決策者共訪談 57 人次，另外本計畫已設計 3 種客製化選單，包含經濟部災害緊急應變小組旱災開設、經濟部災害緊急應變小組風災開設與經濟部災害緊急應變小組水災開設，經實際運作與調整，讓智慧應答機器人能因應不同需求提供不同防汛應變資訊。另亦依實際防汛應變需求，定義各式資料更新頻率及建置資料更新流程，並建立對應之資料庫總表。在智慧應答機器人精進研發部份，本計畫已開發一個視覺化且直覺之管理後台，包含帳號權限管理等 5 大類功能。汛期問卷調查部分本計畫共取得 33 份問卷，其中整體滿意度達 4.3 分。本計畫針對即時性防災作業資訊研發 10 項訂閱推播與 10 項歷史推播紀錄，另針對大雨/豪雨特報等 5 大類圖資進行自動繪圖功能研發作業，以視覺化圖示提供決策參考。本計畫針對防汛應變需求設計「地方政府開設概況」等共 4 種對話設計回應形式。本計畫另於 0522 豪雨等 4 起事件應變期間針對應變人員與決策者共訪談 57 人次，並有 35 位應變同仁提供滿意度建議，平均滿意度為 4.5 分。在教育訓練部份，已完成協助辦理 7 場次教育訓練辦理，並已於 6 月 12 日交付及側錄影片，及完成問卷統計。亦配合水利署綜合企劃組之國際水週活動規劃，於 9 月 24 日至 9 月 26 日於台北世界貿易中心一館進行擺攤事宜。關鍵詞：應答機器人、決策系統、語意分析. Abstract The “2020 Development of Intelligence Response Robot” project follows up on the achievements and experience of 2019 and continues providing professional services for the four major topics: system maintenance, further system discussions and applications, advanced research and development, and assistance in staff training. By November 30 of 2020, we have achieved the following objectives:.

(4) 109 年度. For system maintenance, we have updated the inventory of the flood prevention response database, covering 437 items of information, divided into the six categories: weather, preparedness, response, recovery, miscellany, and feedback. Also, the visualization design of 59 static disaster prevention information is updated. Regular updates and testings of the maintenance platform and the robot system are performed to ensure service availability. During the nine disaster events in the year to date, including Typhoon Vongfong (2020), the project staff stationed in the emergency operation center (EOC) to monitor the system at any time to ensure that the system and user problems can be solved immediately. For further system discussions and applications, weekly user behavior reports are provided as a reference for system adjustment and update. Four information items, such as the pump allocation, are added to the database to strengthen the information integrity of the robot. During the four disaster events in the year to date, including 0522 Heavy Rain Event, we interviewed 57 interviewers and decision-makers for data optimization. We have also completed three designs of menus for the Disaster Emergency Response Task Force of the Ministry of Economic Affairs targeting different types of events, including drought events, typhoon events, and flood events. Through actual operations and adjustments, the robot actively provides different flood prevention information according to different requirements. Meanwhile, several data update frequencies and processes have been established according to the actual flood control requirements. A database master list has been additionally created. For advanced research and development, we have developed a visual and intuitive management platform, featuring five major functionalities such as user authorization setting. In the flood season survey, 33 questionnaires were collected, concluding the overall satisfaction of 4.3 points out of 5. Ten notification items for subscription and record lookup panels are provided for real-time disaster situation grasping. The feature of automatic image generation is implemented in 5 notification items such as the heavy rain alert for providing decision reference. We have also designed four responses, such as the overview of local governments’ EOCs, according to the flood control requirements. During the four disaster events in the year to date, including 0522 Heavy Rain Event, 57 interviewers and decisionmakers were interviewed for data optimization, and 35 of them provided feedbacks with an average satisfaction score of 4.5 out of 5. In the assistance in staff training, we have assisted seven training sessions, submitted recording videos on June 12, and completed the questionnaire statistics. Following the International Water Week event plan of the Comprehensive Planning Section of the Water Resources Agency, participate in the event at Taipei World Trade Center from September 24 to September 26. Keywords：Intelligence Response Robot, decision system, semantic analysis..

(5)