Mongolian text sentiment analysis method based on multi-size CNN and LSTM model

Abstract

The invention discloses a Mongolian text sentiment analysis method based on a multi-size CNN and an model. The method comprises the steps of: preprocessing a Chinese and Mongolian sentiment text corpus; converting the words obtained through preprocessing into dynamic word vectors; connecting a multi-size CNN network and an mLSTM network in parallel to form a Mongolian text sentiment analysis model; splicing features extracted by the two to serve as emotion features finally extracted by the model; using a transfer learning strategy to take a large-scale Chinese emotion text corpus as a training set, transferring a neural network parameter weight obtained by training to a Mongolian text emotion analysis model as an initial parameter, and training by using the preprocessed Mongolian emotion text corpus to obtain a Mongolian text emotion analysis model based on a multi-size CNN and an LSTM model; and comparing and evaluating the analysis result of the model and the analysis result of a single network analysis method according to the accuracy rate, the recall rate and the F1 value, and therefore, the purpose of improving the Mongolian text sentiment analysis performance is achieved.

Description

technical field [0001] The invention belongs to the technical field of artificial intelligence, and in particular relates to a Mongolian text sentiment analysis method based on multi-size CNN and LSTM models. Background technique [0002] With the rapid development of Internet technology, more and more people begin to share their feelings, views and opinions by expressing various opinions on platforms such as Weibo, forums, film and television websites, and shopping websites. And the content published by users may contain different emotional colors: positive or negative; support or opposition. The core of sentiment analysis is to divide the sentiment expressed in a piece of text into positive and negative categories. Its research value has been fully reflected in product reviews and recommendations, public opinion monitoring, and information prediction. [0003] With the rise of artificial intelligence, deep learning methods have received widespread attention. Because of th...

AI Technical Summary

Problems solved by technology

However, there are relatively few Mongolian corpus resources that can be collected, so it cannot meet the requirements of deep neural network model training

Third, the current single neural network model does not have good real-time performance when solving text sentiment analysis, and cannot extract local and global information of the text at the same time, resulting in poor classification results

Images