Text sentiment classification method, storage medium and computer

• Main Authors: HE TIANCHANG, XU CHENGGONG, ZENG WEIBO, LYU JUNYU, LIN XIANGXIAN, NI SHILONG, ZHENG YAOSONG, SU JIANGWEN
• Format: Patent
• Language: Chinese; English
• Published: 16.01.2018
• Subjects: CALCULATING; COMPUTING; COUNTING; ELECTRIC DIGITAL DATA PROCESSING; PHYSICS

Provided are a text sentiment classification method, a storage medium and a computer. The text sentiment classification method comprises the following steps that a sentiment dictionary is establishedfor input texts, wherein the sentiment dictionary establishing step includes that word properties are selected and expressed, bottom-layer feature vectors are extracted, middle-layer features are extracted, the sentiment dictionary is combined to acquire word vectors of training samples, and middle-layer feature vectors are obtained after the word vectors of the training samples are pooled; weighting fusion is conducted on the bottom-layer feature vectors and the middle-layer feature vectors to obtain fused feature vectors, and classification results are calculated based on a bottom-layer feature vector classification model, a middle-layer feature vector classification model and a fused feature vector classification model. The problem that sentiment classification in the prior art is not efficient and stable enoug