Neural machine translation decoding acceleration method based on discrete attention mechanism

Abstract

The invention discloses a neural machine translation decoding acceleration method based on a discrete attention mechanism, and the method comprises the steps: constructing a neural machine translationmodel for training a parallel corpus and based on an attention mechanism; generating a machine translation word list through the parallel corpus; carrying out the further training to obtain a model parameter after training convergence, and enabling the model parameter to serve as a baseline system; converting partial parameters of the attention mechanism in the network into integer values by scaling the model parameter values, and mapping floating-point numbers into the integer section of the integer values; replacing the normalized part which cannot be subjected to integer calculation with alinear structure which is beneficial to integer calculation; controlling whether all numerical values participate in operation or not through a threshold value before the linear structure; and adjusting the selected threshold value on 1%-2% of data randomly extracted from the training data, so that a better translation result can be achieved. According to the method, the real-time corresponding speed can be improved on the basis of the latest implementation of rapid reasoning and on the premise that the model performance is hardly reduced.

Description

technical field [0001] The invention relates to a neural machine translation decoding acceleration technology, in particular to a neural machine translation decoding acceleration method based on a discrete attention mechanism. Background technique [0002] Machine translation (Machine Translation) is the use of computer programs to translate a natural language into another natural language, which belongs to the category of computational linguistics. In 1949, Warren Weaver published a memorandum titled "Translation", which marked that machine translation based on modern computers officially entered the stage of history. Machine translation not only involves human cognition of its own language and way of thinking, but also involves many fields such as artificial intelligence, information theory, knowledge engineering, and software engineering. It is a discipline that intersects multiple technologies in depth. In the past ten years, the research and industrialization of machin...

AI Technical Summary

Problems solved by technology

[0009] In view of the fact that the machine translation method in the prior art relies too much on single-precision and double-precision floating point and cannot take advantage of low-precision numerical calculations, the technical problem to be solved by the present invention is to provide a neural machine translation decoding based on a discrete attention mechanism The acceleration method makes full use of the natural advantages of low computational complexity of fixed-point numbers, and can improve the real-time response speed on the basis of the latest implementation of fast reasoning with almost no decline in model performance.

Images