Position embedding interpretation method and device, computer equipment and storage medium

Abstract

The invention discloses a position embedding interpretation method and device, computer equipment and a storage medium. The method comprises the following steps: acquiring an original data set, and preprocessing the original data set; acquiring a plurality of positions corresponding to the plurality of pieces of position data; according to a plurality of position context sequences in the preprocessed original data set, performing learning by using a Skip-gram model to obtain a plurality of position embedding vectors; obtaining a plurality of position categories corresponding to the plurality of pieces of position category data; performing learning by using a Skip-gram model to obtain a plurality of position category embedding vectors, and enabling the plurality of position embedding vectors and the plurality of position category embedding vectors to be located in an original vector space; and according to a predetermined rule, converting each position embedding vector from an original vector space to a semantic vector space to obtain a position semantic representation corresponding to each position embedding vector. According to the method, more semantic information can be learned, so that the vector of each dimension has interpretability.

Description

technical field [0001] The embodiments of the present invention relate to the technical field of machine learning, and in particular, to a position embedding interpretation method, device, computer equipment and storage medium. Background technique [0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art. [0003] The learning of embedding models is mainly divided into two categories: the first is inspired by the success of word embedding models, and the word vector (Word2vec) framework can be used to learn position embedding (ie position embedding vector) from check-in data. Check-in sequences can be modeled and the influence of linear context is captured to learn location embeddings, which can be used for personalized place recommendations. In addition to the sequential mode, the follow-up begins to consider dynamic user preferences and the time factor of learning position re...

AI Technical Summary

Problems solved by technology

At the same time, the interpretability of positional embeddings is also ignored

[0005] To sum up, the position embedding models in related technologies ignore the interpretability of the model, so that each dimension of the learned vector has no specific meaning and cannot meet the needs of business scenarios

Images