Address matching algorithm based on deep learning model

Abstract

The invention relates to an address matching algorithm based on a deep learning model. The address matching algorithm comprises the following steps: firstly, carrying out word segmentation on an address in a corpus by utilizing a Jieba Chinese word segmentation library; carrying out address word vector training by utilizing a word vector (Word2vec) model; and finally, carrying out address text semantic similarity calculation by utilizing an enhanced sequence inference model (ESIM), and outputting a matching result. The method is different from a traditional address matching algorithm which focuses on directly performing similarity calculation and text matching by utilizing literal overlapping of a matching address, and the algorithm focuses on researching the semantic similarity degree ofan address text and completing a matching task on the basis of the semantic similarity degree. A deep learning algorithm suitable for current massive multi-source heterogeneous address data matching tasks is provided.

Description

technical field [0001] The invention relates to the field of computer deep learning, in particular to a deep learning method for address matching. Background technique [0002] With the rapid development of information technology, the spatio-temporal data generated by various industries such as medical care, communication, and logistics are increasing day by day. According to statistics, more than 80% of human activities and urban information are related to geographic spatial location, and its main link is address information. Therefore, addresses are playing an increasingly important role in people's lives. Existing industry data (such as medical care, public security, etc.) usually store spatial location attributes in the form of address text. To realize centralized management, analysis and information sharing in geographic space, it must first be converted into spatial data . To complete this process, it is necessary to find and obtain the geographic coordinates corresp...

AI Technical Summary

Problems solved by technology

Due to the complexity of the expression of Chinese addresses, most addresses only have a certain spatial orientation in semantics, but lack a standard address structure

Traditional address matching methods mainly focus on the matching relationship between words in address texts, and cannot accurately identify the same pointing relationship of different addresses in different expressions: for example, "2502, Lane 1, Longtengge, Fuyong" and "Baishixia Community, Fuyong Street, Baoan No. 2, Lane 1, Longteng Pavilion, Defeng Road" refers to the same geographical location, but there is not much overlap in literal expression, and it is difficult to accurately identify the relevant relationship by directly matching words

Therefore, in this context, the traditional address matching method is no longer applicable to match multi-source heterogeneous massive address data

Images