AUTHOR=Liu Ning , Yuan Zhenming , Chen Yan , Liu Chuan , Wang Lingxing 

TITLE=Learning implicit sentiments in Alzheimer's disease recognition with contextual attention features

JOURNAL=Frontiers in Aging Neuroscience

VOLUME=Volume 15 - 2023

YEAR=2023

URL=https://www.frontiersin.org/journals/aging-neuroscience/articles/10.3389/fnagi.2023.1122799

DOI=10.3389/fnagi.2023.1122799

ISSN=1663-4365

ABSTRACT=Background: Alzheimer’s disease(AD) is difficult to be diagnosed based on language because of the implicit emotion of transcripts, AD recognition is a supervised fuzzy implicit emotion classification with document level. Recent neural network-based approaches has paid no attention to the implicit sentiment entailed in AD transcripts.
Method: In order to detect deep semantic information, a two-level attention mechanism is proposed, enabling it can attend different important contents as constructing the document representation. To be specific, the document vector is built by aggregating important words to sentence vectors and important sentences into document vectors progressively.
Results: Experiment results show that our method achieves a best 88.28% accuracy  on annotated Pitt corpora, which validates its effective of learning implicit sentiment representation for our model.  
Conclusion: The proposed model can select informative words and sentences qualitatively by attention layers, and this method gives a good inspiration for AD diagnosis based on implicit sentiment transcripts.