Research on the Construction of New Media Social Culture Based on Long Short-term Memory

The maturity and popularization of computer-aided technology is a prerequisite for the construction of new media social culture, but the purpose of the construction of new media social culture is not to realize the application and popularization of computer-aided technology. This paper attempts to analyze the cultural characteristics of new media from the perspective of humanities, and carry out the research on the social and cultural construction of new media based on computer-aided DL(Deep learning) technology, so as to grasp the pulse of the current social and cultural development of new media and explore a road of social and cultural construction of new media suitable for the characteristics of the times. Therefore, based on LSTM (long short-term memory) in DL and attention mechanism, this paper proposes a hierarchical attention network to realize text classification. At the same time, two levels of attention mechanism are introduced to obtain the best representation of the text. The results show that the micro-average F1 value of this model on English data set is 0.769, which is 3.532% higher than that of LSTM model, and has a certain improvement compared with that without introducing topic information. The effectiveness of two models, which combine attention mechanism and conditional coding, to introduce topic target information is verified. © 2023 CAD Solutions, LLC, http://www.cad-journal.net.