Relation extraction-oriented sentence structure information acquisition method

Abstract

The invention discloses a relation extraction-oriented sentence structure information acquisition method. The method comprises the following steps of: 1, extracting a relationship mention statement which comprises two entities and has a known entity semantic relationship category from a data set; 2, separating and marking the entities in the relationship mention statement extracted in the step 1 by using entity markers and a separator; 3, performing vector mapping on a text based on a pre-training word vector lookup table or a random word vector lookup table; 4, performing convolution operation on a vector matrix representing the text through a neural network to extract sentence structure features; 5, performing maximum pooling operation on a result which is obtained after convolution, andfurther obtaining abstract features; and 6, carrying out full connection, and predicting a classification result through a Softmax layer. According to the method, the sentence entities are marked andseparated before the convolutional neural network, so that semantic information of content of each part can be better obtained, sentence structure features taking the entities as centers are obtained, relationship extraction is carried out, and relatively good performance can be achieved.

Description

Technical field [0001] The invention relates to a method for processing input data into a neural network, in particular to a method for acquiring sentence structure information oriented to relation extraction, and belongs to the technical field of natural language processing. Background technique [0002] With the rapid spread of computers worldwide and the rapid development of Internet technology, various data such as video, audio, pictures, text, etc. have proliferated, and a large amount of information has appeared in front of users in the form of electronic digitization. In order to cope with the severe challenges brought by the information explosion, professional automation tools are urgently needed to extract the truly valuable information from the massive data, and information extraction emerges at the historic moment. Information extraction technology is a widely used information processing technology in the field of natural language processing, and relation extraction is...

AI Technical Summary

Problems solved by technology

Entity semantic relationship extraction is performed by entity tags in the sentence, so that the neural network can obtain the relative position information and semantic connection information between the vocabulary and the entity pair in the sentence other than the entity, so as to obtain the structural information of the sentence centered on the two entities. And to a certain extent, it avoids the feature sparsity problem caused by traditional machine learning methods, thereby improving the performance of relationship extraction, and effectively solving the problem of not being able to make good use of sentence structure information.

Images