Spatial relation extraction method and device based on semantic dependency

Abstract

The invention discloses a spatial relationship extraction method and device based on semantic dependency, and the method comprises the steps: constructing a semantic dependency recognition model, extracting a spatial relationship tuple from an unstructured text, firstly, defining semantic dependency types, namely a trigger word-containing role type and a trigger word-free role type; then inputting a to-be-recognized text and pre-labeled space elements into a semantic dependency recognition model based on the deep self-attention network, and obtaining all semantic dependency recognition results among the space elements in combination with the defined semantic dependency type; and finally, combining the obtained semantic dependency recognition results, and outputting a complete spatial relationship tuple. According to the method, a spatial relationship extraction problem is converted into a semantic dependency recognition problem, and various spatial relationships including a spatial relationship containing a trigger word and a spatial relationship not containing the trigger word can be processed at the same time; according to the method, the spatial semantic information in the text can be effectively expressed, the semantic dependence among the spatial elements is extracted, manual feature design is not needed, the generalization performance is high, and the accuracy rate is high.

Description

Technical field [0001] The present invention belongs to the technical field of natural language processing, and related to information extraction techniques, a spatial relationship extraction method and apparatus based on semantic dependence. Background technique [0002] With the rapid development of the Internet information industry and the arrival of network big data ages, network data has grown rapidly, and people expect to quickly and efficiently excavate useful information from massive data. [0003] The text contains rich spatial information. The meaning of spatial information is relatively wide, generally reflects information about the spatial distribution characteristics of objective objects, such as its own position, spatial structure, and association with other objects on spatial distribution. Understanding spatial information in natural language can provide underlying support for application systems in different fields, for example: for question and answer systems, ca...

AI Technical Summary

Problems solved by technology

[0004] At present, a lot of work has been carried out in the field of spatial relationship extraction. Researchers at the Institute of Human Language Technology at the University of Texas proposed a sieve-based extraction method, which decomposes spatial relationships into multiple sub-relations, and uses a combination of syntax trees A series of support vector machine classifiers of features classify sub-relationships, but training classifiers requires manual design of features, which is inefficient and poor in scalability; researchers at the University of Lisbon tried to use a sequence labeling model based on convolutional neural networks. The fixed text first identifies the trigger word, also known as the spatial relationship word, and then uses the trigger word and the text as input to extract other roles of the spatial relationship. The main problem of this method is that it cannot handle the spatial relationship without the trigger word, that is, the implicit type of spatial relationship; the patent application "A Method for Recognition and Extraction of Water Conservancy Spatial Relationship Words" (public number CN110532553A) proposes a method for identifying spatial relationship words in the field of water conservancy based on weak supervision, and mines spatial relationship patterns through seed sets , and then extract the spatial relationship tuples, this method needs to design a large number of lexical grammar rules and features in the process of processing; patent application "method for extracting spatial relationship of geographical location points, method and device for training extraction model" (public number CN111737383A) , proposed a method for extracting the spatial relationship of geographical location points, which is modeled as a sequence labeling problem. For a given location, its spatial location information is output. This method also cannot handle the relationship without trigger words.

Images