Fusion model-based Chinese address identification method
The invention discloses a Chinese address recognition method based on a fusion model, which comprises the following steps: firstly, collecting unstructured text data including Chinese addresses, preprocessing the unstructured text data to form a data set, then constructing a plurality of recognition...
Saved in:
Main Authors | , |
---|---|
Format | Patent |
Language | Chinese English |
Published |
20.10.2023
|
Subjects | |
Online Access | Get full text |
Cover
Loading…
Summary: | The invention discloses a Chinese address recognition method based on a fusion model, which comprises the following steps: firstly, collecting unstructured text data including Chinese addresses, preprocessing the unstructured text data to form a data set, then constructing a plurality of recognition models, training the data set, and respectively performing Chinese address recognition on the input text data based on the trained recognition models; and then constructing a text scoring system, evaluating output results of the plurality of models based on text scores, and finally fusing recognition results based on evaluation results to obtain a final Chinese address recognition result. According to the method, a plurality of optimized models are constructed through machine learning and deep learning to deal with the complexity and diversity of Chinese addresses, and the models can process common mark noise in the addresses, such as punctuation marks, omitted words and wrong segmented words, so that the recognit |
---|---|
Bibliography: | Application Number: CN202311176478 |