Feature-enriched character-level convolutions for text regression

Abstract

We present a new model for text regression that seamlessly combine engineered features and character-level information through deep parallel convolution stacks, multi-layer perceptrons and multitask learning. We use these models to create the SHEF/CNN systems for the sentence-level Quality Estimation task of WMT 2017 and Emotion Intensity Analysis task of WASSA 2017. Our experiments reveal that combining character-level clues and engineered features offers noticeable performance improvements over using only one of these sources of information in isolation.