Proceedings of the SIGCHI Conference on Human Factors in Computing Systems
CHI '85 Proceedings of the SIGCHI Conference on Human Factors in Computing Systems
KI '98 Proceedings of the 22nd Annual German Conference on Artificial Intelligence: Advances in Artificial Intelligence
Computational Linguistics
Speech and Language Processing (2nd Edition)
Speech and Language Processing (2nd Edition)
A Meta Modeling Framework for Domain Specific Process Management
COMPSAC '08 Proceedings of the 2008 32nd Annual IEEE International Computer Software and Applications Conference
Design challenges and misconceptions in named entity recognition
CoNLL '09 Proceedings of the Thirteenth Conference on Computational Natural Language Learning
The Stanford typed dependencies representation
CrossParser '08 Coling 2008: Proceedings of the workshop on Cross-Framework and Cross-Domain Parser Evaluation
Process model generation from natural language text
CAiSE'11 Proceedings of the 23rd international conference on Advanced information systems engineering
Foundations of Machine Learning
Foundations of Machine Learning
| Hi-index | 0.00 |
In this paper, we describe a novel approach of extracting models from natural language text sources. This requires linguistic analysis as well as techniques for interpreting and using the analysis results. Our linguistic analysis engine provides feature analysis for a rule-based model element detection. Furthermore, the presented approach enables users to generate domain- and application-specific model element detection rules based on natural language sample sentences. Detection rules also have to be connected to instantiation rules for the respective type of model element. This is done through a highly system-supported mapping step where users are able to choose elements from arbitrary meta models and to connect their properties with functions over natural language sentence parts. As both, the definition and application of detection rules is always a sensitive balancing act between precision and recall, these steps are highly interactive. That is why our current prototype also supports detection rule adaption and iterative rule set completion -- always to the level of current need.