Entity relationship extraction method based on improved deep residual network and attention mechanism

Abstract

The invention discloses an entity relationship extraction method based on an improved deep residual network and a relationship word attention mechanism, which comprises the following steps: convertingEnglish into a word vector by processing a remote supervision entity relationship data set; obtaining relation words, and calculating cosine similarity to extract relation word weights of sentences;respectively introducing a segmentation mechanism and a gating mechanism to form a segmented residual error network with a gate module, and segmenting the characteristic channel by adopting the segmented gating residual error network; splicing the average pooling layer on the basis of the maximum pooling layer at the rear section of the residual network to form a double-pooling layer, wherein thedouble-pooling layer is used for reserving sentence structure features; obtaining final sentence features of the model; and performing relationship classification on the obtained sentence features toobtain an output entity relationship. According to the method, the deep residual error network is applied to a remote supervision data set of relationship extraction, the noise reduction capability ofthe model is effectively improved by deepening the number of network layers, and meanwhile, a segmented gating residual error module is provided, so that the performance of the deep residual error network is effectively improved.

Description

technical field [0001] The invention belongs to the technical field of natural language processing, and in particular relates to an entity relationship extraction method based on an improved deep residual network and a relationship word attention mechanism. Background technique [0002] Nowadays, deep learning models have been widely used in various fields, but what has been bothering many domestic and foreign scholars is that the commonly used deep learning methods require a large number of training data sets to train the model in order to achieve better fitting results. The construction of conventional training data sets requires a lot of manual labeling work, which will increase research costs and time expenditures. In addition, data labeling work in certain professional fields (such as medicine, agriculture, finance, military, etc.) also requires labelers to have certain professional domain knowledge, which makes the progress of labeling work very difficult. Therefore, ...

AI Technical Summary

Problems solved by technology

Furthermore, the characteristics of the deep residual network make the structural difference of each layer of the network very small, which directly leads to the advantages and disadvantages of the information between the feature channels. Once the network is deepened, it means that the redundant feature channels will become more and more more

Images