Advancing Identification of Causal Relations in Natural Language
Open AccessAs the amount of available digital text has been growing exponentially, the development of tools and algorithms for processing this often-unstructured natural language is critical to providing scalable solutions for various applications. These algorithms are often classified under the broad field of Natural Language Processing (NLP). One major component of NLP is the identification of the semantic relationships within various constituents. The tools that can be used to identify causal relations are the ones that are relatively less advanced, despite their promise and potential to unlock great value across various applications.In this thesis, we investigate three different applications where the identification of causal relations can solve major problems and, in turn, provide insight into where and how causal relation identification models can be improved to increase their performance in these specific application areas and beyond. By introducing new frameworks and building novel tools, we first demonstrate the utility of causal relations in the following areas, (1) methods for discerning between satire and false news, (2) analyzing and predicting the spread of various types of information online, and (3) developing computational tools for characterizing the capability of a given text in creating a coherent mental representation for readers. Following these demonstrations, we initially highlight the need to build more robust models for accurate identification of causal relations in natural language, following which we present a framework for building new commonsense knowledge-augmented models to identify causal relations and explanations in the text.In summary, in this thesis, we initially demonstrate how enhanced causal relation models and tools can enable the automation of advanced tasks within three impactful application areas. These demonstrations, in turn, provide insight into how novel frameworks and tools can be developed to increase the power and efficacy of causal relation identification algorithms in these applications and beyond. Throughout the thesis, and in each chapter, we also comment on how our research can be used to guide the field through future developments of causal relation classification models and algorithms.
- All rights reserved
Notice to Authors
If you are the author of this work and you have any questions about the information on this page, please use the Contact form to get in touch with us.
| Thumbnail | Title | Date Uploaded | Visibility | Actions |
|---|---|---|---|---|
|
|
Hosseini_gwu_0075A_16177.pdf | 2022-10-04 | Open Access |
|