-
公开(公告)号：EP3848855A1
公开(公告)日：2021-07-14
申请号：EP18934368.4
申请日：2018-09-19
发明人： ZHANG, Qing , YANG, Wei , XIAO, Yifan , ZHANG, Lianghe , JUI, Shangling
摘要： Embodiments of this application provide an intention identification model learning method and an apparatus, and a device, and relates to the field of communications technologies, to help improve accuracy of the intention identification model in a human-machine dialog system, improve accuracy of task execution in the human-machine dialog system, and improve user experience. The method includes: receiving, by a server, positive data (S101) that corresponds to a first skill and that is entered by a skill developer; generating, by the server based on the positive data that corresponds to the first skill, negative data (S102) that corresponds to the first skill; determining, by the server, a second skill (S103) similar to the first skill; obtaining, by the server, data (S104) that corresponds to each second skill; generating, by the server, a second base model (S105) based on the data that corresponds to the second skill and a first base model stored on the server; and performing learning (S 106), by the server, based on the second base model and the positive data and the negative data that correspond to the first skill, and generating an intention identification model.
-
公开(公告)号：EP4227858A1
公开(公告)日：2023-08-16
申请号：EP21891127.9
申请日：2021-11-10
发明人： XIAO, Yifan , ZHANG, Jian , ZHONG, Zhao
IPC分类号： G06N3/08
摘要： A neural network structure determining method is disclosed. The method includes: obtaining a to-be-trained initial neural network, where the initial neural network includes M first blocks block and a second block, the second block is connected to each first block, and each first block corresponds to one trainable target weight; performing model training on the initial neural network, to obtain M updated target weights; and updating a connection relationship between the second block and the M first blocks in the initial neural network based on the M updated target weights, to obtain a first neural network. In this application, in a process of searching for a connection relationship between blocks in the initial neural network, a trainable target weight is added to a connection between blocks, an updated target weight is used as a basis for determining importance of the connection relationship between the blocks, and the connection relationship between the blocks is selected and removed based on the updated target weight, to search for a topology structure of the neural network.
-
搜索结果
2 条记录 (耗时 0.007 秒)
