Comparison to Stanford Scene Graph Parser

[berthyf96]

Thanks for the great implementation. Does this function the exact same way as the Stanford Scene Graph Parser? If not, how is it different from the Stanford parser?

[vacancy]

No. They are not identical. This repo tries to replicate only part of the results of the rule-based scene graph parser by Schuster et al, with a focus on the empirical performance.

It DOES cover all rules implemented in the Stanford Parser, with minor heuristic optimizations and extra rules for handling

• compound nouns
• phrasal prepositions such as "in front of"
• phrasal verbs
• detection of scene-level nouns (e.g., airport)

However, it does NOT cover some features of the Stanford Parser, including:

• quantificational modifiers such as "a lot of"
• pronoun resolution
• plural nouns (see Sec. 3.1.3 of Schuster et al)

[berthyf96]

Thanks for the quick and detailed response!