Existing benchmarks used to evaluate the performance of end-to-end neural dialog systems lack a key component: natural variation present in human conversations. Most datasets are constructed through crowdsourcing, where the crowd workers follow a fixed template of instructions while enacting the role of a user/agent. This results in straight-forward, somewhat routine, and mostly trouble-free conversations, as crowd workers do not think to represent the full range of actions that occur naturally with real users. In this work, we investigate the impact of naturalistic variation on two goal-oriented datasets: bAbI dialog task and Stanford Multi-Domain Dataset (SMD). We also propose new and more effective testbeds for both datasets, by introducing naturalistic variation by the user. We observe that there is a significant drop in performance (more than 60% in Ent. F1 on SMD and 85% in per-dialog accuracy on bAbI task) of recent state-of-the-art end-to-end neural methods such as BossNet and GLMP on both datasets.

We present a new method for aligning sentences with their translations in a parallel bilingual corpus. Previous approaches have generally been based either on sentence length or word correspondences. Sentence-length-based methods are relatively fast and fairly accurate. Word-correspondence-based methods are generally more accurate but much slower, and usually depend on cognates or a bilingual lexicon. Our method adapts and combines these approaches, achieving high accuracy at a modest computational cost, and requiring no knowledge of the languages or the corpus beyond division into words and sentences.

We develop an improved form of left-corner chart parsing for large context-free grammars, introducing improvements that result in significant speed-ups more compared to previously-known variants of left corner parsing. We also compare our method to several other major parsing approaches, and find that our improved left-corner parsing method outperforms each of these across a range of grammars. Finally, we also describe a new technique for minimizing the extra information needed to efficiently recover parses from the data structures built in the course of parsing.