The present disclosure pertains to parsing addresses using a deep learning model. The deep learning model may be trained using a plurality of reference address records and a plurality of modifications to the plurality of reference address records. In some implementations, the model can be a multi-layer bi-directional long short-term memory recurrent neural network model. The output of the model may include indications of an address component type for each character of the address. The output can be parsed based on the indications of the address component type to obtain a plurality of labeled address components for the address. The labeled address components can be compared to reference address records to determine a matched address record identifying a geographic location.
展开▼