Text multi-label hierarchical classification method and system driven by feature words

• Main Authors: GAO JIAN, JIANG HANG, WANG CHENYU, LIN YUEFENG, LU JIDONG, NI MENGJUN, MIAO ZHONGCHEN, SHI GUANGWEI
• Format: Patent
• Language: Chinese; English
• Published: 22.12.2020
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; ELECTRIC DIGITAL DATA PROCESSING; PHYSICS

The invention discloses a text multi-label hierarchical classification method and system driven by feature words, and can solve the problem of text multi-label hierarchical classification by means offeature word driving under the condition of not providing annotation data and only needing to provide feature words related to labels. According to the technical scheme, a heterogeneous information network is used for learning word vectors, information except for texts is fully used, and the final technical effect can be improved. According to the method, a multi-label pseudo document generation technology is provided, and the technology is an important premise for the method to work. If the multi-label pseudo document generation technology is not provided, label data need to be provided, andexpensive labeling cost can be brought. According to the method, a confidence coefficient filtering mechanism is introduced in the self-training process, a novel confidence coefficient calculation method is designed, and the e