Automated Data Mapping Based on FastText and LSTM for Business Systems

With the continuous development of information technology, massive information processing has become an important problem in business systems. However, themetadata information from different business systems lacks a unified and standardized description method. Mapping data by the manual way greatly reduces the efficiency. Therefore, an automated data mapping method is very necessary. In this paper, we regard data mapping as a text classification problem based on the following reasons: 1) the text classification technology has become more and moremature in the field of the natural language processing (NLP), which is very suitable for processing massive data; 2) a large number of heterogeneous mapping data can be treated as text. In order to implement automated datamapping, in this paper, we propose a classification model based on FastText and long-short term memory (LSTM) for data mapping in business systems. By observing the characteristics of mapping data in business systems, we firstly use FastText to learn word representation containing semantic information, and then adopt the LSTM model to extract features for text classification automatically. Experimental results show that the proposed method can automatically classify mapping data in business systems with common quality.