FAQ system sequencing method, device and system based on hybrid strategy

Abstract

The invention provides an FAQ system sequencing method, device and system based on a hybrid strategy. The method comprises a step 10 of model training, namely a step of inputting a first training set into a text matching model to obtain a trained text matching model, and inputting a second training set into a pairwise sequencing model to obtain a trained pairwise sequencing model; a step 20 of coarse screening, namely a step of calculating the similarity between the questions input by a user and candidates in the standard candidate library through the trained text matching model, sequencing according to scores, and selecting the first m candidates; a step 30 of fine sequencing, namely a step of scoring the m candidates through the trained pair sequencing model, and returning top-k candidates according to score sequencing; wherein k <= m. The text matching model pays attention to the correlation between single sentence pairs, and the recalled candidate range can be narrowed through coarse screening of the text matching model. The Pairwise ranking model considers the relative position among the candidates, and learning of the partial order relation between the candidates is more sufficient. The combination of the two types of models can improve the sequencing performance of an existing FAQ system based on a single text matching strategy.

Description

technical field [0001] One or more embodiments of the present invention relate to the technical field of information retrieval, and in particular relate to a sorting method, device and system for an FAQ (Frequently Asked Questions, frequently asked questions) system based on mixed strategies. Background technique [0002] This section is intended to provide a background or context for implementations of the invention that are recited in the claims. The descriptions herein may include concepts that could be explored, but not necessarily concepts that have been previously thought of or explored. Therefore, unless otherwise indicated herein, what is described in this section is not prior art to the description and claims in this application and is not admitted to be prior art by inclusion in this section. Intelligent question answering is a field of artificial intelligence. With the rapid growth of Internet business, the market demand for intelligent question answering is incr...

AI Technical Summary

Problems solved by technology

[0007] One or more embodiments of this specification describe the FAQ system sorting method, device and system based on the mixed strategy, using the mixed sorting strategy of text matching model coarse screening and pairwise sorting model fine sorting, which combines the advantages of the two types of models and solves the problem of In the existing technology, there is a text matching model that can only judge whether a query is related to a certain candidate, but cannot model the partial order relationship between multiple candidates corresponding to a query.

Images