Why is syntax important?

• Foundation for semantic analysis

Why is syntax insufficient?

• syntax encodes the structure of language but doesn't directly address meaning
• syntax alone doesn't "grab" in an action to take in the world

Lexical semantics

• Vector representation that encodes information about the distribution of contexts a word appears in
• Words that appear in similar contexts have similar representations
• We can represent what individual words "mean" as a function of what other words they're related to

• "Grab" = execute GrabbingFunction() -> 코드화해서 의미 전달
• "the cup" = object ID 9AF1948A81CD22 -> representation

semantics

• Lexical semantics is concerned with representing the meaning of words
• Logical semantics is concerned with representing the meaning of sentences

Meaning representaion

• A meaning representation should be unambiugous; each statement in a meaning representation should be have one meaning

First-order logic (FOL)

• We want to representation every sentence as an umambiguous proposition in FOL

• How we map a natural language sentence to FOL is the task of semantic parsing; but we define the FOL relations and entities to be sensitive to what matters in our model

Relations

• N-art relations hold among FOL terms (constants, variables, functions) -> 더 복잡하게 가능하다.
  

Event semantics

[방법1]

[방법2]

[방법3]

shallow semantics

[뉴스 - twitter ... 에 따라 방법이 다를 수 있다.]

Thematic rolse

• Thematic roles capture the semantic commonality among arguments for different ralations

• John broke the window
• The window was broken by John
  -> 이 두 문장은 syntax로만 보면 완전히 다른 문장이다.

근데, Thematic role 까지 보면 같은 문장이다.

=> "창문이 깨졌다" 가 핵심이고 위의 features 을 이용해서 부가적인 정보를 더 붙임

• Thematic roles are very useful but difficlt to formally difine AGENT, THEME, etc.
• At the same time, they may be too coarse for som applications

Coarsening : Proto-roles

• 좀 더 '일반적'으로 role을 부여 (피상적 Level)
• Proto-roles = generalize thematic roles

2가지 Data

Propbank

• Sentences from the Penn Treebank annotated with proto-roles, along with lexical entries for each sense of a verb identifying the specific meaning of each proto-role that verb sense

• Verb-sepcific argument structures lets us map the commonalities among the different surface forms

FrameNet

• Propbank maps argument structure for individual verb senses

• FrameNet maps argument structure for frames, which are evoked by a lexical unit

• "A frame is a data-structure for representing a stereotyped situation" - Minsky 1975

Semantic Frame

=> Lexical units 를 이용해서 Destroy 에 Frame 을 씌울 수 있다.

=> 서로 같은 뜻이지만 문장 구조가 다른 두 문장에 Frame 을 씌워서 같은 문장으로 바라볼 수 있게 된다.
=> 같은 sell 과 bought 가 각 문장에서 같은 것을 의미한게 된다.