Simple Convolutional Neural Networks with Linguistically-Annotated Input for Answer Selection in Question Answering
Year: 2018
Type: dissertation
Abstract: With the advent of deep learning methods, researchers have been increasingly preferring \ndeep learning methods over decades-old feature-engineering-inspired work in Natural Language Processing (NLP). The research community has been moving away from otherwise dominant feature engineering approaches; rather, is gravitating towards more complicated \nneural architectures. Highly competitive tools like part-of-speech taggers that exhibit \nhuman-like accuracy are traded off for complex networks, with the hope that the neural \nnetwork will learn the features needed. In fact, there have been efforts to do NLP "from \nscratch" with neural networks that altogether eschew featuring engineering based tools \n(Collobert et al, 2011). In our research, we modify the input that is fed to neural networks \nby annotating the input with linguistic information: POS tags, Named Entity Recognition \noutput, linguistic relations, etc. With just the addition of these linguistic features on a \nsimple Siamese convolutional neural network, we are able to achieve state-of-the-art results. \nWe argue that this strikes a better balance between feature vs. network engineering.
Source: UWSpace (University of Waterloo)
Author: Royal Sequiera
Language: English
Topic: Topic Modeling
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W2885344072