Cross-domain big data named entity recognition based on transfer learning

The automatic acquisition of important target entities across domains poses a series of challenges. This paper proposes a cross-domain large-scale named entity recognition model based on transfer learning, which integrates a lite BERT, Bi-Gated recurrent unit, and conditional random field into one model. Due to the limited availability of publicly accessible data and the multitude of entity categories, this study initially utilizes ALBERT for generating distributed character vectors. Subsequently, the BiGRU model is employed to capture the contextualization of text sequences. Finally, constraints are added to obtain specific labeling results. Experimental results demonstrate that the proposed transfer learning-based cross-domain large-scale named entity recognition model achieves an improvement in F1 score, thus providing support for the artificial intelligence-enabled cross-domain endeavors.