Short text sentiment analysis method based on sum product network depth autocoder

Abstract

The invention discloses a short text sentiment analysis method based on a sum product network depth autocoder. The method comprises the following steps of 1, preprocessing short text data; 2, utilizing a doc2vec model to train sentence vectors; 3, utilizing the sum product network depth coder to code the sentence vectors, and obtaining layered abstract characteristics of the sentence vectors; 4, utilizing a maximum product network depth decoder to decode the layered abstract characteristics, comparing the decoded characteristics with the primary sentence vector characteristics, calculating a reconstruction error, adjusting parameters of the sum product network depth autocoder to make the reconstruction error smallest, obtaining an optimal sum product network depth coder, and obtaining an optimal layered abstract characteristic by the optimal sum product network depth coder; 5, utilizing the optimal layered abstract characteristic to conduct online structure learning to generate a sum product network structure, using a small amount of short text data with tags to finely adjust a sum product network, using an online parameter learning algorithm to continuously adjust network parameters, inputting a test set, and obtaining sentiment classifications through the trained sum product network.

Description

technical field [0001] The present invention relates to sum product networks and short text sentiment analysis. Background technique [0002] With the continuous development of social networks, people are more willing to express their opinions through social networks, so understanding users' emotional tendencies through online comments of users has received widespread attention in the academic community. But the online comments of these users are short texts with few words. Short texts provide less information and are more colloquial, which increases the difficulty of sentiment analysis. [0003] Most of the current research based on supervised learning short text sentiment analysis has achieved good results, relying on a large amount of manually labeled data, making the system based on supervised learning need to pay a high labeling cost. On the contrary, unsupervised learning does not require manual labeling of data to train the model, which is a solution to reduce the c...

AI Technical Summary

Problems solved by technology

However, deep autoencoders also have many limitations: 1. Its structure is stacked by different types of modeling units. These modeling units have theoretical flaws, so the deep structure they build is not perfect.

2. Its unsupervised layer-by-layer greedy training only solves the local minimum problem to a certain extent, without global optimization. As the number of hidden layers increases, the gradient dilution becomes more serious, and its training is extremely cumbersome and requires a lot of skills

3. The model structure is often fixed, and it does not consider the relevance of the short text data itself, and cannot perfectly fit different types of short text data

Images