FMR: Functional Meaning Representation & Semantic Parsing Framework

Projects that uses FMR

mathsolver

codes: https://github.com/liuzl/mathsolver
demo: https://mathsolver.zliu.org/

What is semantic parsing?

Semantic parsing is the process of mapping a natural language sentence into an intermediate logical form which is a formal representation of its meaning.

The formal representation should be a detailed representation of the complete meaning of the natural language sentence in a fully formal language that:

Has a rich ontology of types, properties, and relations.
Supports automated reasoning or execution.

Representation languages

Early semantic parsers used highly domain-specific meaning representation languages, with later systems using more extensible languages like Prolog, lambda calculus, lambda dependancy-based compositional semantics (λ-DCS), SQL, Python, Java, and the Alexa Meaning Representation Language. Some work has used more exotic meaning representations, like query graphs or vector representations.

FMR, a formal meaning representation language

FMR stands for functional meaning representation
Context-Free Grammar for bridging NL and FMR
VIM Syntax highlighting for FMR grammar file

Tasks

Grammar checkers
Dialogue management
Question answering
Information extraction
Machine translation

What can FMR do, a glance overview

// semantic parsing
"五与5.8的和的平方的1.5次方与two的和减去261.712" =>
nf.math.sub(
  nf.math.sum(
    nf.math.pow(
      nf.math.pow(
        nf.math.sum(
          5,
          nf.math.to_number("5.8")
        ),
        2
      ),
      nf.math.to_number("1.5")
    ),
    2
  ),
  nf.math.to_number("261.712")
); // denotation: 1000

// slot filling
"从上海到天津的机票" => nf.flight("上海", "天津");
"到重庆，明天，从北京" => nf.flight("北京", "重庆");
"到上海去" => nf.flight(null, "上海");