Semantic Hypergraphs

• url: https://arxiv.org/abs/1908.10784
• author: @camille-roth @telmo-menezes
• #related
  • Computational Social Science
  • Syntactic Annotation
  • Universal Dependencies v1

Highlights

• symbolic methods possess several proper and important features, namely that they can offer human-readable knowledge representations ofknowledge, aswell as language under- standing through formal and inspectable rule-based logical inference.
• Initial approaches to NLP were of a symbolic nature, based on rules written by hand, or in algorithms akin to the ones that are used for pro- gramming language interpreters and compilers, such as recursive descent parsers. It became apparent that the diversity of grammatical constructs that can be found in natural language is too large to be tackled in such a way.
  • provokedThoughts
    • Maybe, but with gd (Private) I've gotten 95% of what I need already
• SHs can represent semantic information that is lost in the graphic representation, for example the ability to express n-ary relationships, propositions about propositions and constructive definitions of concepts.
• Argumentation Mining
• hyperedges may be ordered [23] and recursive [32]. Ordering entails that the position in which a vertex participates in the hyperedge is relevant (as is the case with directed graphs). Recursivity means that hyperedges can participate as vertices in other hyperedges.
• Metatrees and metagraphs are themselves too low a level to be suitable as a programming language which human beings would want to use on a daily basis. Instead, they have more in common with compiler intermediate languages, which are term rewriting systems used by machines (compilers) to perform transformations on data.

Comparable

• Cyc
• Atomspace / OpenCog
• Abstract Meaning Representation
• OpenNLP
• Conceptnet

---

Backlinks

• N Ary Relationships