Learning to parse from a semantic objective: It works. Is it syntax?

Recent work on reinforcement learning and other gradient estimators for latent tree learning has made it possible to train neural networks that learn to both parse a sentence and use the resulting parse to interpret the sentence, all without exposure to ground-truth parse trees at training time. Surprisingly, these models often perform better at sentence understanding tasks than models that use parse trees from conventional parsers. This paper aims to investigate what these latent tree learning models learn. We replicate two such models in a shared codebase and find that (i) they do outperform baselines on sentence classification, but that (ii) their parsing strategies are not especially consistent across random restarts, (iii) the parses they produce tend to be shallower than PTB parses, and (iv) these do not resemble those of PTB or of any other recognizable semantic or syntactic grammar formalism.