A Novel Named Entity Recognition Mehod for Bus Route Identification in Social Media

With the rapid development of Internet, many people post their comments on smart bus services in social media. Some travelers' comments may contain bus route information. If bus route information can be identified, it is beneficial for regulatory personnel to troubleshoot all buses on the route, thereby improving the level of intelligent bus service. Therefore, the paper established a newly named entity recognition method to recognize bus route information. Firstly, the Bidirectional Encoder Representations from Transformer (BERT) model was used to encode the texts into vector representations. Then, multi Bidirectional Gated Recurrent Unit (MultiBiGRU) layers were used to improve the accuracy of feature extraction. Finally, the Conditional Random Fields (CRF) layer was used to calculate the global optimal result and the identification results of bus route information was obtained. The test results illustrated that the F1 value in the test set of the method reached 0.9389, and the recognition accuracy was higher than models such as CRF and BiGRU-CRF.