Research on Argument Text Clustering Method Based on Autoencoder

With the rapid development and wide application of the Internet, people can express their opinions on social platforms anytime and anywhere through a terminal. However, in online forums, seminar systems, and argument systems, although the topics discussed may be different, the platform will put together irrelevant texts based on the time of submission of the text. In this way, not only the relevance between the texts is reduced, but also the difficulty for users to read is increased. This paper proposes a model that combines BiLSTM and autoencoder. The model uses BiLSTM to further extract text features, and then uses a clustering algorithm to supervise and update the weights in the autoencoder network. Its principle is to first calculate the centroid of the K-Means algorithm, and then feed the obtained centroid to the network model to update the weights of the model, and finally calculate a more suitable centroid based on the network model to make the clustering algorithm get more accurate results. Experiments show that the network model is significantly better than the baseline method on the two data sets.