Identify the Difficulty of the Word ERNIE Through the Model Based on BP Neural Network

The experiment requires the development of a model to solve keywords by difficulty categories, as well as identifying the attributes associated with each category, being able to identify the difficulty of the word ERNIE through the model and validating the correctness of the classification model. We first chose to use the expert decision method to define the relationship between the combined number of answers and the difficulty level of the word, then classified the difficulty by building a gradient boosting tree and used the model built to calculate the feature importance and then identify the attributes associated with each classification, using Spearman correlation analysis to evaluate the difficulty of ERNIE to obtain the difficulty level of the word and to validate the classification. The correctness of the model is verified.