Incomplete statement rewriting method and device, incomplete statement model training method and device and related equipment

Abstract

The invention relates to the artificial intelligence technology, and provides an incomplete statement rewriting and model training method and device and related equipment, and the method comprises the steps: carrying out the word segmentation of a multi-round dialogue text, inputting the segmented words into a trained statement rewriting model, and converting each segmented word into a vector representation through the trained statement rewriting model; on the basis of a self-attention mechanism of a trained statement rewriting model, capturing correlation among the segmented words according to vector representation, and obtaining a plurality of target self-attention matrixes; adjusting the target self-attention matrix to obtain a corresponding target word segmentation relation matrix; based on a semantic segmentation mechanism of a trained statement rewriting model, obtaining a target word editing matrix according to all the target word segmentation relation matrixes; and based on the target word editing matrix, executing editing operation of the corresponding editing type on the incomplete statement to obtain a rewritten statement. According to the method, the problems of omission and complementation and anaphora resolution in incomplete utterance rewriting are simplified and solved by utilizing the self-attention matrix.

Description

technical field

[0001] The present application relates to the technical field of artificial intelligence, and in particular, to a method, apparatus and related equipment for rewriting and model training of incomplete sentences. Background technique

[0002] In recent years, due to the huge potential commercial value of incomplete sentence rewriting in the field of dialogue robots, it has attracted more and more attention from academia and industry. The goal of incomplete utterance rewriting is to solve the problems of denotative resolution and omission completion in multiple rounds of dialogue, so that sentences can fully express their meanings even when they are out of context. The methods in the prior art usually convert the incomplete sentence rewriting task into a machine translation task or a text generation task, and adopt a generative model with a copy mechanism, and require some additional auxiliary skills such as adding part-of-speech tagging, copying mechanism, two...

Examples

Embodiment Construction

[0028] In order to make the purposes, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments It is a part of the embodiments of this application, but not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work shall fall within the protection scope of the present application. It should be understood that the specific embodiments described herein are only used to explain the present application, but not to limit the present application.

[0029] The method for rewriting incomplete sentences provided in this application can be applied to computer devices such as servers or terminal devices, for example, per...

technical field

[0001] The present application relates to the technical field of artificial intelligence, and in particular, to a method, apparatus and related equipment for rewriting and model training of incomplete sentences. Background technique

[0002] In recent years, due to the huge potential commercial value of incomplete sentence rewriting in the field of dialogue robots, it has attracted more and more attention from academia and industry. The goal of incomplete utterance rewriting is to solve the problems of denotative resolution and omission completion in multiple rounds of dialogue, so that sentences can fully express their meanings even when they are out of context. The methods in the prior art usually convert the incomplete sentence rewriting task into a machine translation task or a text generation task, and adopt a generative model with a copy mechanism, and require some additional auxiliary skills such as adding part-of-speech tagging, copying mechanism, two...

Examples

Embodiment Construction

[0028] In order to make the purposes, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments It is a part of the embodiments of this application, but not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work shall fall within the protection scope of the present application. It should be understood that the specific embodiments described herein are only used to explain the present application, but not to limit the present application.

[0029] The method for rewriting incomplete sentences provided in this application can be applied to computer devices such as servers or terminal devices, for example, per...

Images