Hoorcollege 4: Semantic graphs

Question 1

Directed graphs

Answer 1

G = (N, E)
can also be labelled / rooted graphs

Nodes are circles. Edge (a, b) is an arrow from node a to node b)

Question 2

Trees as graphs

Answer 2

A tree is a rooted graph in which, for every node n, there is exactly one directed path from the root to n.

• Path is a sequence of nodes in which each adjacent pair (a, b) is connected by an edge; i.e. (a,b) ∈ E or (b,a) ∈ E.
• Directed path is a sequence of nodes in which each adjacent pair is in E

Question 3

semantic representations often have ways of handling these “extra” dependencies

Answer 3

• Indexed traces and other empty categories in syntax trees
  ti -> referent is the same as higher in DP
• Variables in logical expressions
  λx : dance(x) ∧ sing(x)
• Graphs

Because relate to more than one other part of the sentence

Question 4

More graph concepts

Answer 4

• Undirected graph: Graph in which edges are binary sets, not order pairs
• (un) direced cycle: (un) directed path from a node back to itself
• Reentrancy: A node with more than one incoming edge
• DAG: Directed graph with no directed cyles

Question 5

Bilexical Semantic dependencies

Answer 5

3 frameworks:
* DM, PAS, PSD
* Semantic representation design
Should functional words be in the graph?
What about semantic scope?
Which kinds of coreference should create reentrancies?

Question 6

Abstract meaning representation (AMR)

Answer 6

Abstract away from the sentence
* Edges: propbank (Argi) similar idea to framenet
* Verb: stem + number (e.g pilot-01)
* Provides numbers for arguments Arg0, Arg01)