Deep Contrast Learning Approach for Address Semantic Matching

Address is a structured description used to identify specific place or point of interest, and it provides an effective way locate people objects. The standardization Chinese name address occupies important position in the construction smart city. Traditional specification technology often adopts methods based on text similarity rule bases, which cannot handle complex, missing, redundant information well. This paper transforms task into calculating pairs, proposes contrast learning matching model attention-Bi-LSTM-CNN network (ABLC). First all, ABLC use Trie syntax tree algorithm extract elements. Next, basic idea learning, hybrid neural applied learn semantic address. Finally, Manhattan distance calculated as two addresses. Experiments self-constructed dataset with data augmentation demonstrate that proposed has better stability performance compared other baselines.