Chinese Medical Named Entity Recognition Based on Parameter Transfer Learning

To reduce the dependence on the labeled data in the target domain of Chinese medical named entity recognition task, we studied the application of parameter transfer learning in Chinese medical named entity recognition. The method firstly combines the data of different domains with the target domains data for word embedding training, so as to achieve semantic information sharing at the representation layer. Secondly, the internal encoding layer parameters of the source domain model are transferred to the target domain model by training the source domain model. Finally, the parameter transfer is combined with the constructed weak label dataset to solve the inconsistent distribution of labels in target and source domains, which means the parameter transfer in decoding layer is achieved. The bottom-up parameter transfer of the neural network is achieved through the above steps. Experiment shows that the proposed method can successfully improve the recognition performance of the target domain model on small samples, and reduce the dependence of the target domain labeling data.