LAGr: Label Aligned Graphs for Better Systematic Generalization in Semantic

Machine Learning Probabilistic Graphical Models
#1

This paper proposes a probabilistic model for unaligned semantic parsing. The backbone is a vanilla tagger/seq2seq to predict node labels given token sequence. Then a biaffine-like module is applied to nodes to predict edge labels. This backbone can be trained directly on aligned graphs. On unaligned graphs, they propose to factorize the probability over every possible alignments. One alignment can be viewed as a permutation over tokens. Then, the probability of nodes and edges can be obtained by marginalizing the permutations. Since there are factorial number of permutations, this method is intractable. So, they approximate this probability using hard Expectation Maximization, they only consider the maximum-a-posteriori and optimize it instead. To infer this posteriori, they drop edge labels and only consider edge nodes. To further improve this approximation, they generate a shortlist of candidate alignments and train on them in fully aligned way.

Comments

  • Nice math and approximation.
  • Not sure if this method can align AMR nodes correctly but I’m curious.
Rating
  • 5: Transformative: This paper is likely to change our field. It should be considered for a best paper award.
  • 4.5: Exciting: It changed my thinking on this topic. I would fight for it to be accepted.
  • 4: Strong: I learned a lot from it. I would like to see it accepted.
  • 3.5: Leaning positive: It can be accepted more or less in its current form. However, the work it describes is not particularly exciting and/or inspiring, so it will not be a big loss if people don’t see it in this conference.
  • 3: Ambivalent: It has merits (e.g., it reports state-of-the-art results, the idea is nice), but there are key weaknesses (e.g., I didn’t learn much from it, evaluation is not convincing, it describes incremental work). I believe it can significantly benefit from another round of revision, but I won’t object to accepting it if my co-reviewers are willing to champion it.
  • 2.5: Leaning negative: I am leaning towards rejection, but I can be persuaded if my co-reviewers think otherwise.
  • 2: Mediocre: I would rather not see it in the conference.
  • 1.5: Weak: I am pretty confident that it should be rejected.
  • 1: Poor: I would fight to have it rejected.

0 voters