We found a match
Your institution may have access to this item. Find your institution then sign in to continue.
- Title
Constraint information extraction for 3D geological modelling using a span-based joint entity and relation extraction model.
- Authors
Zhuang, Can; Liu, Chunhua; Zhu, Henghua; Ma, Yuhong; Shi, Guoping; Liu, Zhizheng; Liu, Bohan
- Abstract
Data sparsity has long been a problem in 3D geological modeling work. The geometric, topological, and attribute information of geological bodies in geological reports provide important constraint information during 3D geological modeling. However, manually extracting complex and diverse constraint knowledge from a large amount of textual data is a challenging and time-consuming task. The development of information extraction and text mining technology has made it possible to automatically extract textual constraint information. To this end, this study firstly summarized the textual description characteristics of geological body constraint information in geological reports, and used a span-based tagging scheme for data annotation; Secondly, a span-based joint entity and relation extraction framework was introduced to extract constraint information in geological 3D modeling, which improves the extraction capability of the geological modeling constraint information by obtaining deep semantic information of the characters through the BERT model, in addition, the model has the joint extraction capabilities of entity classification and relation classification on candidate entities; Finally, in the experiments study, a Chinese geological survey report was used as training data for evaluation, and we validated our method's effectiveness through comparison of our results to those of different models. We further compared and analyzed the impact of different parameters and span representations on our model's performance.
- Subjects
GEOLOGICAL modeling; DATA mining; LANGUAGE models; TEXT mining; GEOLOGICAL surveys; ARTIFICIAL joints
- Publication
Earth Science Informatics, 2024, Vol 17, Issue 2, p985
- ISSN
1865-0473
- Publication type
Article
- DOI
10.1007/s12145-024-01245-2