BERT-BiLSTM-CRF Based Named Entity Recognition Method for Controlled Speech
Zixuan Wang, Yidi Wang, Xuan Wang, Qinyue He · 2023
As research on automatic speech recognition technology in the controlled speech field continues to deepen, a variety of controlled speech recognition systems have emerged. However, due to varying levels of quality, it is essential to evaluate the effectiveness of these systems. A crucial factor in assessing controlled speech recognition systems is their accuracy in recognizing keywords. As such, it is of significant value to investigate techniques to extract essential keywords from recognized controlled speech texts. To this end, this paper proposes a BERT-BiLSTM-CRF based model for named entity recognition research. Furthermore, this model’s performance was tested on the CLUENER2020 dataset, providing feasible and promising results. Given the limitations of available data and research for entity recognition in the controlled speech field, this paper also designed a named entity recognition dataset for this field using the BIO annotation method. This achievement not only expands the research scope of named entity recognition but also provides critical groundwork for future advancements in the field of controlled speech recognition.