A Neural Network Approach to Classifying Generic Expressions
Abstract
Generic expressions make statements about nonspecific entities. Characterizing generic expressions make broad statements about classes of entities, without belonging to any temporal structure, while habitual generic expressions refer to regularly occurring events and actions. Non-generic expressions can also be classified as characterizing or habitual, complicating the task of differentiating between generic and non-generic expressions. Correctly differentiating generic expressions from non-generic expressions plays a role in tasks such as information extraction and knowledge base population, as this distinction determines the type of information that can be interpreted from a given expression.
The goal of this research is to develop a neural network that classifies generic expressions. Previous machine learning approaches to classifying generic expressions required precise feature extraction prior to the training process, and did not fully utilize semantic information in learning to classify generic expressions. A neural network approach will allow the system to learn from a wide variety of grammatical and semantic features, and adjust its internal model to take advantage of features that strongly identify an expression as generic or non-generic.
Citation
De Guzman, Carlo Jacob (2018). A Neural Network Approach to Classifying Generic Expressions. Undergraduate Research Scholars Program. Available electronically from https : / /hdl .handle .net /1969 .1 /164531.