Text entity relationship classification method based on small sample learning

Abstract

The invention discloses a text entity relationship classification method based on small sample learning. The text entity relationship classification method comprises the following steps: 1) extracting semantic features of instance vectors in a data set by adopting a convolutional neural network as an instance encoder; 2) in a small sample learning scene, endowing each instance with a weight by designing a prototype-level attention mechanism module, and representing a prototype of each relationship in a weighted summation manner; and 3) in a small sample learning scene, replacing a new metric function. Extracting a feature coefficient in a support set vector by using convolution operation through a distance level attention mechanism module, and calculating a distance between each relation prototype and a query instance in a support set by using a product of a Manhattan distance formula and the feature coefficient as a new metric function; and 4) realizing small sample relation classification by utilizing a softmax function.

Description

technical field [0001] The invention relates to a text entity relationship classification method based on small sample learning, and belongs to the technical field of text data recognition. Background technique [0002] As one of the important subtasks of knowledge extraction, relation classification is receiving more and more attention. For unstructured text data, the task of relation classification is to extract the semantic relationship between two or more entities from the text. At present, in relation classification problems, most of the mature technologies achieve excellent experimental results by improving traditional neural network models (such as recurrent neural networks, convolutional neural networks, etc.). However, the datasets selected in the experiment are simple short sentences with predefined categories, and the sample distribution of each relationship is relatively uniform. However, in practical applications, there are often challenges such as small data ...

AI Technical Summary

Problems solved by technology

Due to the small amount of data in the few-shot learning scenario, when an instance is far away from other instances in the mapping space, it will lead to a huge bias in the average computed prototype.

Once such noise data exists in large quantities, it will have a great impact on the final relationship classification effect; at the same time, for a relationship feature vector, only a part of the dimension has a clear distinction on the final classification result, once extracted from the support set The instance vector has the problem of sparse features, which will cause a large deviation to the final classification result

Images