Transition-Based Deep Input Linearization
This work addresses error propagation in deep NLG for researchers and practitioners, though it is incremental as it builds on existing transition-based methods.
The paper tackled the problem of error propagation and limited information sharing in pipeline approaches for deep natural language generation by introducing a transition-based model that jointly performs linearization, function word prediction, and morphological generation, achieving the best results reported on a standard shared task.
Traditional methods for deep NLG adopt pipeline approaches comprising stages such as constructing syntactic input, predicting function words, linearizing the syntactic input and generating the surface forms. Though easier to visualize, pipeline approaches suffer from error propagation. In addition, information available across modules cannot be leveraged by all modules. We construct a transition-based model to jointly perform linearization, function word prediction and morphological generation, which considerably improves upon the accuracy compared to a pipelined baseline system. On a standard deep input linearization shared task, our system achieves the best results reported so far.