Deep Transfer Learning Model for Semantic Address Matching

Address matching, which aims to match an input descriptive address with a standard in database, is key technology for achieving data spatialization. The construction of today’s smart cities depends heavily on the precise matching Chinese addresses. Existing methods that rely rules or text similarity struggle when dealing nonstandard data. Deep-learning-based often require extracting semantics embedded representation, not only complicates process, but also affects understanding semantics. Inspired by deep transfer learning, we introduce approach based pretraining fine-tuning model identify semantic similarities between various We first pretrain corpus enable (abbreviated as ASM) learn contexts unsupervised. then build labelled dataset using address-specific geographical feature, allowing problem be converted into binary classification prediction problem. Finally, fine-tune ASM and compare output several popular methods. results demonstrate our achieves best performance, precision, recall, F1 score above 0.98.