WEKO3
アイテム
介護記録および行動認識のための対話システム活用の研究
https://doi.org/10.18997/00008415
https://doi.org/10.18997/000084157b44a3ec-b097-40ae-9833-14b34898da17
| 名前 / ファイル | ライセンス | アクション |
|---|---|---|
kou_k_525.pdf (3.7 MB)
|
|
| アイテムタイプ | 学位論文 = Thesis or Dissertation(1) | |||||||
|---|---|---|---|---|---|---|---|---|
| 公開日 | 2021-07-30 | |||||||
| 資源タイプ | ||||||||
| 資源タイプ識別子 | http://purl.org/coar/resource_type/c_db06 | |||||||
| 資源タイプ | doctoral thesis | |||||||
| タイトル | ||||||||
| タイトル | A Dialogue System for Nursing records and Activity Data Collection in Long-Term Care Settings | |||||||
| 言語 | en | |||||||
| タイトル | ||||||||
| タイトル | 介護記録および行動認識のための対話システム活用の研究 | |||||||
| 言語 | ja | |||||||
| 言語 | ||||||||
| 言語 | eng | |||||||
| 著者 |
Mairittha, Tittaya
× Mairittha, Tittaya
|
|||||||
| 抄録 | ||||||||
| 内容記述タイプ | Abstract | |||||||
| 内容記述 | Rapid population aging has led to a growth in demand for nursing care facilities in Japan as many studies reported concern over the increased workload of nurses. Moreover, most of the heavy workload involved in recording patient information. So, convenient tools can significantly improve the patient safety and productivity of nurses. The transition from paper to an electronic health record （EHR） has undoubtedly increased nurses’ productivity in the digital age. EHR helps to reduce errors due to poor handwriting and costs through decreased paperwork. It also enables quick access to patient records and shares information with others. Speech Recognition is the technology that converts spoken word into text. It is commonly faster than typing. It allows nurses to enter data while taking care of patients （or hands-free mode）. However, it often applies to note recording because it is difficult to transition from one field to the next with speech commands. For example, the user has to specify the input fields or use command words to go to the following field. With rapid advances in technology, artificial intelligence （AI）, such as natural language processing （NLP） and machine learning （ML） algorithms, have recently been applied to healthcare areas and significantly affected clinical practice and research. A spoken dialogue system is a significant advancement of speech recognition systems. It is regularly associated with the notion of AI, which can simulate a conversation with a user in natural language. It is designed to help users solve a specific task with explicit intent within minimal dialogue turns. It understands the speaker’s intent in context and extracts valuable structured data that can drive actions and analytic. The dialogue system used in various forms in personal assistants, such as Apple’s Siri, Amazon’s Alexa, and the Google Assistant, is becoming a part of our daily lives, especially on mobile devices. Combining this technology into the EHR provides a natural language for unstructured medical records and further speeds up the recording process beyond basic operations such as a keyboard. Besides, smartphones are increasingly used, integrating the dialogue system into smartphones offering the opportunity to collect health data to monitor personal health all over the day continuously, which can contribute to improvements in the quality of care and optimize the workflow and workload. Activity recognition aims to recognize human physical activity such as walking and running. This method’s end goal is to allow computers to provide assistance and guidance to a person before or while undertaking a task. For example, we can apply activity recognition to nursing care applications to recognize complex nursing care activities such as assisted walking and vital checking. By integrating activity recognition and nursing care records, we can enhance activity recognition and nurse efficiency. For example, recognizing the sensors’ activities and then suggesting nurses record them using the dialogue system. Yet, it is underused, and there are still challenges that this technology frequently fails to address. When working with EHRs, we necessitate to structure and organize them. However, the information they contain is usually available in an unstructured format. For healthcare providers, it is relatively common to read the documents and get the information they need. Notwithstanding, if we want to use the data for analysis or machine learning, it is challenging without structure. Particularly EHR for nursing records differs from usual EHR systems used to enter information on new patients or update each recent encounter. In care homes, nurses usually provide patients with nursing care up to 24 h a day. So, the record can have many different data formats, including free-text clinical narratives that correspond to different encounters generated at various points of time and heterogeneous contents across various healthcare providers. A significant barrier is that much of the records are unstructured and non-standard formats, challenging to analyze and interpret, especially if they work in different sections or facilities. It would require human interpretation due to the domain-specific vocabulary, potential spelling errors, acronyms, and abbreviations. This manual extraction may be a time-consuming task because it needs highly skilled human resources. Therefore, we need solutions to automatically fill a form and algorithms to convert unstructured text into structured in an automated way. This thesis explores data-driven methodologies based on machine learning to develop the spoken dialogue system, especially nursing record systems on smartphones. The main contributions of this thesis are the following: （1） identifying the challenges of implementing the dialogue system for nursing record systems; （2） integrating of the dialogue system, activity data collection, and nursing records on an Android smartphone; （3） developing an automatic dialogue labeling framework to train natural language understanding （NLU）, which is a core component in implementing spoken dialogue systems to understand the purpose of a user’s utterance; （4） exploring transfer learning techniques for constructing a question answering system to answer questions in the EHR to fulfill the NLU system. The thesis’s findings illustrate the feasibility and improvement of nursing recording performance through the spoken dialogue system. Furthermore, it provides recommendations and promising future research directions. | |||||||
| 目次 | ||||||||
| 内容記述タイプ | TableOfContents | |||||||
| 内容記述 | 1 Introduction||2 Background and Related Work||3 Identifying Challenges of Developing the Dialogue System for Nursing Records||4 Integrating the Dialogue System and Nursing Records on Smartphones||5 A Dialogue-Based Annotation for Activity Recognition||6 Automatic Labeled Dialogue Generation for Nursing Records||7 Improving Fine-Tuned Question Answering Models for Electronic Health Records||8 Discussion||9 Conclusions | |||||||
| 備考 | ||||||||
| 内容記述タイプ | Other | |||||||
| 内容記述 | 九州工業大学博士学位論文 学位記番号：工博甲第525号 学位授与年月日：令和3年6月28日 | |||||||
| キーワード | ||||||||
| 主題Scheme | Other | |||||||
| 主題 | Dialogue system | |||||||
| キーワード | ||||||||
| 主題Scheme | Other | |||||||
| 主題 | Nursing records | |||||||
| キーワード | ||||||||
| 主題Scheme | Other | |||||||
| 主題 | Activity recognition | |||||||
| キーワード | ||||||||
| 主題Scheme | Other | |||||||
| 主題 | Healthcare | |||||||
| キーワード | ||||||||
| 主題Scheme | Other | |||||||
| 主題 | Information extraction | |||||||
| アドバイザー | ||||||||
| 井上, 創造 | ||||||||
| 学位授与番号 | ||||||||
| 学位授与番号 | 甲第525号 | |||||||
| 学位名 | ||||||||
| 学位名 | 博士（工学） | |||||||
| 学位授与年月日 | ||||||||
| 学位授与年月日 | 2021-06-28 | |||||||
| 学位授与機関 | ||||||||
| 学位授与機関識別子Scheme | kakenhi | |||||||
| 学位授与機関識別子 | 17104 | |||||||
| 学位授与機関名 | 九州工業大学 | |||||||
| 学位授与年度 | ||||||||
| 内容記述タイプ | Other | |||||||
| 内容記述 | 令和3年度 | |||||||
| 出版タイプ | ||||||||
| 出版タイプ | VoR | |||||||
| 出版タイプResource | http://purl.org/coar/version/c_970fb48d4fbd8a85 | |||||||
| アクセス権 | ||||||||
| アクセス権 | open access | |||||||
| アクセス権URI | http://purl.org/coar/access_right/c_abf2 | |||||||
| ID登録 | ||||||||
| ID登録 | 10.18997/00008415 | |||||||
| ID登録タイプ | JaLC | |||||||
