- 期刊
Classification of Adverse Drug Reactions in Social Media
摘要
There are a large number of tweets with adverse drug reactions in social media, and the semantics of different levels of tweets are different. This article proposes a deep learning network method based on supervised learning, starting from different levels of tweets, Find the special semantics contained in different levels, then merge them, and use the BiLSTM method to build a classification model to find tweets related to adverse drug reactions. This article starts from the three levels of sentences; the first level: the character level, using the charCNN model to extract features at the character level; the second level of words, using the BERT model to extract features at this level Vector, third level: sentence level, use the syntactic dependency toolkit developed by Stanford to extract the relationship pairs between words in the sentence and vectorize them; after obtaining the three levels of feature vectors, use the BiLSTM model Build the classifier and finally complete the model.