+*At the other end of the spectrum, in the traditional semanticparsing literature, the problem is framedas predicting compositional semantic representations. These are mainly geared towards question-answering rather than task completion, and are usually directly executed against a knowledge base.Some of the standard datasets in this area include GeoQuery [24] and WebQuestions [2].Within both of these areas, neural approaches have supplanted previous feature-engineering basedapproaches in recent years [10, 14]. In the context of tree-structured semantic parsing, some otherinteresting approaches include Seq2Tree [7] which modifiesthe standard Seq2Seq decoder to betteroutput trees; SCANNER [5, 4] which extends the RNNG formulation specifically for semantic pars-ing such that the output is no longer coupled with the input; and TRANX [23] and Abstract SyntaxNetwork [19] which generate code along a programming language schema. For graph-structured se-mantic parsing [1, 11], SLING [21] produces graph-structured parses by modeling semantic parsingas a neural transition parsing problem with a more expressive transition tag set. While graph struc-tures can provide more detailed semantics, they are more difficult to parse and can be an overkill forunderstanding task oriented utterances.* *[Improving Semantic Parsing for Task Oriented Dialog](https://arxiv.org/pdf/1902.06000.pdf)*
0 commit comments