Abstract
We develop a finite-state transducer for translating unranked trees into general graphs. This work is motivated by recent progress in semantic parsing for natural language, where sentences are first mapped into tree-shaped syntactic representations, and then these trees are translated into graph semantic representations. We investigate formal properties of our tree-to-graph transducers and develop a polynomial time algorithm for translating a weighted language of input trees into a packed representation, from which best-score graphs can be efficiently recovered.
Highlights
In dependency semantic parsing, one is given a natural language sentence and has to output a directed graph representing an associated, mostlikely semantic analysis
Semantic parsing is currently receiving considerable attention, as attested by the number of approaches being proposed for its solution (Oepen et al, 2014, 2015) and by the variety of existing semantic representations and available datasets (Kuhlmann and Oepen, 2016)
A successful approach to dependency semantic parsing by Wang et al (2015b,a) first parses the input sentence into a dependency tree t, and applies a transition-based algorithm that translates t into a dependency graph in Abstract Meaning Representation (AMR), a popular semantic representation developed by Banarescu et al (2013)
Summary
One is given a natural language sentence and has to output a directed graph representing an associated, mostlikely semantic analysis. Since the graph construction mechanism we use is equivalent to hyperedge replacement, our notion of tree-to-graph transducers is essentially an unranked and extended generalization of theirs, except for the fact that ours cannot create multiple copies of unbounded material in the input This ability seems inappropriate for modeling natural language semantics. Groschwitz et al (2018) use a neural supertag parser to map a string into a dependency-style tree representation of the compositional structure of the corresponding AMR graph This tree is a term in a special algebra: its constants denote lexicalized AMR graph fragments, which are combined into larger and larger AMR graphs by two binary algebraic operations for graph combination. The tree-to-graph mapping is entirely deterministic, in contrast to our approach. Groschwitz et al (2018) provide an unsupervised alignment algorithm that extracts rules from semantic graph banks
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
