LAUSR

The automatic generation of Chinese enterprise abbreviations is a task that extracts enterprise abbreviations to represent the enterprise full name. Traditional methods do not divide abbreviations in detail, which leads to a poor generation effect of irregular Chinese enterprise abbreviations generation, and the best selection method of candidate abbreviations among traditional methods is still coarse-grained relationship modeling. To solve the problem of irregular abbreviation generation and abbreviation screening, this paper proposes a two-channel Chinese enterprise automatic abbreviation generation method. First, in the two-channel method, the enterprise component channel outputs regular candidate abbreviations, and the single-character channel outputs irregular candidate abbreviations to improve the processing effect of the method on irregular abbreviations. Then we design a Bayesian filtering model based on the position relationship of abbreviations in enterprise components to improve the final effect of the automatic generation of Chinese enterprise abbreviations. The results show that our effect is the best in the data performance of Chinese enterprises.