融合粗糙数据推理的卷积记忆网络文本情感分析

为解决现有情感分类算法在特征提取中缺乏对语义关联规则的运用，以及在分词后产生大量与情感预测无关的词，导致挖掘出的特征不具代表性的问题。提出一种融合粗糙数据推理的卷积记忆网络情感分析模型。通过上下文信息使用粗糙数据推理获得文本的情感词集Word2Vec词向量表示，并融合FastText词向量来改进特征向量嵌入层。其次使用卷积神经网络（CNN）拼接双向长短期记忆网络（BiLSTM）提取更深层次的情感特征。最后加入Attention机制计算权重，筛选显著重要特征。通过多组对比实验表明该模型具有较高的准确率和F1值，有效提升了情感分类的预测能力。

Convolutional Memory Network Text Sentiment Analysis Based on Rough Data Inference

Current sentiment classification algorithms lack the use of semantic association rules in feature extraction and produce a large number of words irrelevant to sentiment prediction after word segmentation, resulting in unrepresentative features mined. A convolution memory network emotion analysis model based on rough data reasoning is proposed. Firstly, according to the context information, rough data reasoning was used to obtain the word2vec word vector representation of the emotional word set of the text, and the fasttext word vector was fused to improve the feature vector embedding layer.Attention mechanism was added to calculate the weight. Filter for salient and important features. Multi-group comparison experiments show that this model has high accuracy and F1 value, and effectively improves the prediction ability of emotion classification.