Text coding model training method, information retrieval method and equipment

Abstract

The embodiment of the invention discloses a text coding model training method, an information retrieval method and equipment, and belongs to the technical field of machine learning. The method comprises the following steps: inputting sample texts in a text relation network into a text coding model to obtain sample feature vectors corresponding to the sample texts; determining model loss based on the sample feature vector and the objective function; performing iterative training on the text coding model based on the model loss; in response to the text retrieval operation, obtaining retrieval information based on the text retrieval operation; inputting the retrieval information into the text coding model to obtain a retrieval information feature vector corresponding to the retrieval information; determining a target text from a text library based on the retrieval information feature vector; and displaying the target text through the retrieval result display interface. Modeling is performed based on the network relationship of the sample text, and the model can obtain relatively accurate vector representation by capturing semantic information of the text itself under the conditions of sparse network edges and relatively much noise of the text relationship network.

Description

technical field [0001] The embodiments of the present application relate to the technical field of machine learning, and in particular to a text coding model training method, information retrieval method and equipment. Background technique [0002] Information retrieval is a frequently used operation in daily life, such as paper retrieval, news retrieval, and medical consultation retrieval. The user enters keywords or key sentences in the search box, and the terminal retrieves content related to the keywords or key sentences from the document library according to the document search rules, and displays the search results for the user to view. [0003] Related technologies usually encode the text content input by the user into a continuous vector, and then use the model to calculate the similarity between it and the vector representation of each document in the document library, and then determine the retrieval result based on the vector distance. For the training process of...

AI Technical Summary

Problems solved by technology

[0004] However, the above-mentioned contrastive learning pays more attention to the relationship between each node in the text network. When the edges of the text network are relatively sparse and the edges are noisy, the performance of the model will decrease; and this method requires the vector inner product between positive samples to be as large as possible. Large, the vector inner product between negative samples is as small as possible, if the negative samples cannot be properly selected, it will have a great negative impact on the performance of the model

Images