Remote supervision named entity recognition method for PU reinforcement learning

The invention relates to a remote supervision named entity recognition method for PU reinforcement learning, and belongs to the field of natural language processing and machine learning. The method mainly aims at solving the problems that a remote supervision sample of a Chinese named entity recognition task has noise marks, the model learning sample feature efficiency is low, and the training process lacks an effective monitoring mechanism. The method comprises the following steps: firstly, extracting a single sample sequence feature and multi-label scoring information from a text by using a BLSTM model; then, based on PU reinforcement learning, training a sample selector, and screening positive samples and negative samples from the tagged corpus and the remote supervised corpus; and sending the negative sample into a denoising reducer to obtain a reduced sample. A loss function for unbiased and consistent estimation of task loss is introduced, and a positive sample and a reduced sample are used for training