Crowdsourcing text integration method based on multi-stage transfer learning strategy integration

Abstract

The invention provides a crowdsourcing text integration method based on multi-stage transfer learning strategy synthesis. The crowdsourcing text integration method specifically comprises the following steps: 1, constructing a transfer generation type crowdsourcing text integration model TTGCIF; 2, obtaining semantic prototypes of the source domain text data set and the target domain text data set; 3, performing word embedding processing on the semantic prototype; 4, performing data distribution alignment according to the maximum mean value difference; 5, performing semantic prototype transduction model training on the TTGCIF; 6, processing the source domain text data set into a training task set; 7, inputting the training task set into the TTGCIF to carry out field fast adaptation model training; and 8, inputting a part of the target domain text data set into the TTGCIF to carry out model fine tuning training. Through the process, text integration is realized. According to the method, the requirement for data labels in a traditional method can be abandoned, waste of manpower and material resources is reduced, and crowdsourcing text integration in a data scarcity scene is greatly promoted.

Description

technical field [0001] The invention belongs to the technical field of natural language processing. Background technique [0002] A large number of supervised trained models in the field of natural language processing require text with labeled values ​​for training. However, there are few sources of labeled text in related fields. Except for some standard datasets, if new datasets are needed for training, we can only rely on experts to manually generate text label values, which requires a lot of manpower and material resources. In the crowdsourcing environment, a large amount of idle manpower is used to artificially generate labeled values ​​for unlabeled text, and a large amount of text with labeled values ​​can be obtained for training at a very small cost. Relying on the crowdsourcing model to obtain labeled text data has become the main way to obtain training data and labeled values ​​in the field of machine learning. [0003] The crowdsourcing model is a model in whic...

AI Technical Summary

Problems solved by technology

For example, existing methods based on deep learning almost all use supervised training models, which require a large amount of labeled text for model training; In addition, related methods do not make full use of the information that existing data in related fields can provide to improve the effect of crowdsourced text integration

Images