Generating Shifting Sentiment for a Conversational Agent
Simon Whitehead, Lawrence Cavedon · 2010
We investigate techniques for generating alternative output sentences with varying sentiment, using (an approximation to) the Valentino method, based on SentiWordNet, of Guerini et al. We extend this method by filtering out unacceptable candidate sentences, using bigrams sourced from different corpora to determine whether lexical substitutions are appropriate in the given context. We also compare the generated candidates against human judgements of whether the desired sentiment shift has occurred: our results suggest limitations with the overall knowledge-based approach, and we propose potential directions for improvement. 1