Slot position prediction method

Abstract

The embodiment of the invention discloses a slot position prediction method. The slot prediction method comprises the steps of obtaining a second preset number of target slot models under a target intention of a to-be-predicted statement based on initial slot models under a plurality of initial intentions and a first preset number of statement samples; performing slot prediction on the to-be-detected statement based on the second preset number of target slot models to obtain a second preset number of prediction results; and determining slot position information of a target slot position of the to-be-predicted statement based on the weight of the target slot position model and a corresponding prediction result. According to the method, on the basis of slot models under different intentions, the semantic richness of the pre-training vector can be mined and improved by utilizing similar association between the intentions and combining small sample data under the target intention, so that the model training efficiency and the data utilization rate are improved.

Description

technical field [0001] The embodiment of the present invention relates to a slot prediction method. Background technique [0002] In human-computer intelligent interaction, slot prediction is a very important technology, which aims to identify the user's more fine-grained demand purposes under the coarse-grained dialogue intention classification (such as the destination under the travel intention, the play intention under the title, etc.). At present, the commonly used method in the industry is to first use the pre-training model (such as BERT, etc.) to extract word pre-training vectors for one purpose, and then use the recurrent neural network (RNN) algorithm to extract different types of slots from the sentence; Intention, reuse the above training process to independently train its slot prediction model. [0003] At present, the transformation of the technical process in the industry mainly lies in upgrading the pre-training model (for example, replacing the BERT model w...

AI Technical Summary

Problems solved by technology

However, the shortcomings of this improvement method are mainly manifested in the following two aspects: (1) The slot training under multiple intentions is independent of each other, and a lot of work is repeated; (2) When the data cannot be supplemented, the model effect is greatly reduced. Small

Images