Interpreting Output Messages

Parse tree

A python utility is provided to unsquash the parse trees and dependencies output by stanford_parser_ros. This will create a tree structure:

from stanford_parser_msgs import unsquash_tree

[...]

def callback(msg): # msg type: stanford_parser_msgs/Parse
  [...]
  parse_tree = unsquash_tree(msg)
  [...]

Each node in the tree -- an instance of the class Tree -- has the following elements:

node.tag        # string
node.score      # double -- confidence rating of the Stanford Parser
node.word       # string
node.word_index # integer -- node.word = msg.words[node.word_index]
node.children   # listof(Tree)

The part-of-speech tags used in the Stanford Parser come from Penn Treebank II and can be found here.

Dependencies

Dependencies (a list located under msg.dependencies) have the following 3 elements:

• relation: Relation between the governor and dependent. A complete set of relations can be found in the Stanford Dependencies Manual.

• governor_index: The index of the governor in msg.words. In other words, the governor is msg.words[dependency.governor_index].

• dependent_index: The index of the dependent in msg.words. In other words, the dependent is msg.words[dependency.dependent_index].

For instance, the sentence "Give Bill the ball." has, among others, the dependency "dobj(give-2, ball-5)". For this dependency, relation is "dobj", governor_index is 2, and dependent_index is 5.