AUTHOR=Liu Ning , Luo Kexue , Yuan Zhenming , Chen Yan TITLE=A Transfer Learning Method for Detecting Alzheimer's Disease Based on Speech and Natural Language Processing JOURNAL=Frontiers in Public Health VOLUME=Volume 10 - 2022 YEAR=2022 URL=https://www.frontiersin.org/journals/public-health/articles/10.3389/fpubh.2022.772592 DOI=10.3389/fpubh.2022.772592 ISSN=2296-2565 ABSTRACT=Objective: Alzheimer’s disease is a neurodegenerative disease and not easy to be detected with a convenient and reliable method. Language changes in AD patients are an important signal of their cognitive status being affected which may lead to early diagnosis potentially. We develop a Bert-based deep learning model using Natural Language Processing (NLP) technology for early diagnosis of AD from picture description task. Methods: The lack of large datasets poses the limitation for using complex models which do not need feature engineering. Transfer learning can better solve the problem, the model is pre-trained on large datasets for understanding the text properly, then text classification task is performed on small training sets with pre-trained language model. Results: The model was evaluated on ADReSS datasets in 2020, including 78 healthy controls and 78 AD patients. Bidirectional encoder representations from Transformer(BERTBase) embedding combined with Logistic Regression classifier was used and an accuracy of 0.88 is arrived, which was equivalent to the score of champion in the challenge and improved the base score by 0.27. Conclusion: Transfer learning method in this study improves AD prediction which can not only reduce the need for feature engineering, but also solve the problem of lacking sufficient large datasets sufficiently.