Short Text Enhanced Learning Analysis Method Of Online Learning Community

The study of learning analysis in online learning community is of great significance for understanding learning situation of learners and assisting teachers in making teaching decisions.In the early days of this field,learning behavior analysis was mainly based on statistical and behavioral data.In recent years,researchers have paid more and more attention to the impact of short text and other unstructured data on the learning process.With the continuous accumulation of short text data in the online learning community,how to effectively use the hidden information in these short text data has become a key issue to enhance the learning analysis effect.The learning analysis based on short text is influenced by many factors such as the quality of the text,accuracy of semantic understanding,and method of behavior analysis.It is also worth further research.Therefore,this paper uses deep learning and natural language processing techniques to propose a short text enhanced online learning community learning analysis research method.In this method,this paper designs a text error correction model based on the language model to improve the quality of text data in learning community;At the same time,according to the characteristics of the learning community,an improved text feature extractor is proposed to extract the hidden information of text;Finally,we fuse three kinds of information to build an end-to-end learning behavior analysis model.My main work of this article is as follows:1)In order to repair the errors in original text and obtain higher-quality text data,this paper combines a large pre-training model to design a text error correction model based on a hierarchical editing framework.This framework models source text based on BERT and obtains multiple semantic representations of the text;Then this framework uses semantic representation to calculate probability of error occurrence,thereby locating locations of error in the text;Finally,using hierarchical editing framework to obtain specific editing operations of each error location.Based on the CONLL-14 public data set,experiment shows that our framework has faster decoding speed and higher recall rate than the baseline model.2)In order to apply text data to the study of learning analysis,this paper proposes a short text-enhanced learning behavior analysis framework.This framework is divided into three steps:Firstly,determine the type of text features that need to be used;Secondly,based on the characteristics of the learning community,construct a text feature extractor based on natural language processing technology to obtain the corresponding text features;Finally,combine text features and other features to build an end-to-end behavior analysis model and carry out learning analysis research.Under this framework,a sentiment-enhanced performance prediction example is used to verify its effectiveness.The results show that this framework can effectively improve the accuracy of behavior analysis.