Importance of Search and Evaluation Strategies in Neural Dialogue Modeling
This work addresses the challenge of generating more engaging and diverse responses in dialogue systems, though it is incremental as it builds on existing search methods.
The study tackled the problem of improving neural dialogue models by comparing search strategies, finding that better algorithms like iterative beam search lead to higher-rated conversations, with results showing increased diversity and improved human ratings.
We investigate the impact of search strategies in neural dialogue modeling. We first compare two standard search algorithms, greedy and beam search, as well as our newly proposed iterative beam search which produces a more diverse set of candidate responses. We evaluate these strategies in realistic full conversations with humans and propose a model-based Bayesian calibration to address annotator bias. These conversations are analyzed using two automatic metrics: log-probabilities assigned by the model and utterance diversity. Our experiments reveal that better search algorithms lead to higher rated conversations. However, finding the optimal selection mechanism to choose from a more diverse set of candidates is still an open question.