基于多头注意力图卷积网络的社交媒体文本摘要提取方法OA

With the rise of social media,large volumes of textual data have emerged,containing rich information such as user opinions and sentiments,as well as structured social interaction information like retweets and replies.These interactions not only reflect the relationships and behaviors among users but also play a vital role in revealing event dynamics and information dissemination paths,making them crucial for text summarization.However,traditional techniques primarily focus on processing textual content while neglecting social interaction information,which may result in inaccurate summaries or omission of critical information.To address the key challenges in social media text summarization,this paper proposes a graph convolutional network model enhanced with a multi-head attention mechanism.The proposed method first encodes tweets using a pre-trained BERT model to generate initial textual feature representations.Subsequently,a multi-head attention mechanism is employed to evaluate the semantic associations between tweets,and a graph convolutional network is applied to further enhance the representation of tweet relationships,thereby capturing more comprehensive contextual information and semantic relations.Finally,based on the tweet feature,the significance scores of each tweet are evaluated,enabling the efficient and accurate extraction of representative text summaries.This method effectively integrates the semantic information and social signals of tweets to enhance the accuracy and effectiveness of summary generation.Experimental results demonstrate that it outperforms traditional methods,particularly in social media environments,where it better captures relationships between tweets,thereby improving summary quality.This research offers new insights into the study of social media text summarization and provides theoretical support for processing complex social media data.