wroberts / nltk_tgrep Goto Github PK

Documented in Section 4.8 of the tgrep2 manual.

Probably the only obstacle to getting this done is to remove the Unicode string literals in /nltk_tgrep/tests/test_tgrep.py.

The node name /SBJ/ should match nodes names SBJ, SBJ1, and NP-SBJ. Currently, it doesn't match the last, because I'm using re.match. This should be changed to re.search.

Sections 4.6 and 4.7 of the tgrep2 manual.

1. How complicated is the distinction between back-links and cross-links?
2. Is it sufficient to greedily assign the first matching node to a given label, or would a proper implementation of linking potentially mean back-tracking during matching?

This is implemented in _tgrep_node_action, but is not defined in the search string grammar. Probably could be hacked by tweaking the regexp for node names, though perhaps this should be done properly, with a constant 'i@' prefix before, say node literal and node regexp values (so that one can't write i@*).

There is a commented-out test case for this.

Hi @wroberts, just wanted to say this tool is incredibly invaluable! Really a big time saver.

This question is more of a tgrep question, but I thought I might pose it here... I'm trying to develop a tregex for A with no children, or, in other words A as a leaf.

I've tried

"A < {}", "A !< *", and a few others...

Do you know how I can form this tregex or will I need to check the ParsedTree for subtrees of size 0?

Node name quotation permits node names to contain quotes (these must be escaped with a backslash).

Currently, the code doesn't check for this, or perform un-escaping. The current code calls .strip('"'), which is wrong.

Documented in section 4.9 of the tgrep2 manual.

• catching AttributeError in various places
• node name formats:
  • tgrep2 print command (' prefix to node name)
  • quoted node names
  • i@ case insensitive node names
• link relations:
  • <<, <<1

  • ,

  • <<'

  • '

  • undefined link relation (raises AssertionError right now)
• treepositions_no_leaves function
• tgrep_positions with search_leaves set to False

See, e.g., first coveralls build

Hallo,

Your readme example won't work in NLTK 3.0.0

tree = ParentedTree('(S (NP (DT the) (JJ big) (NN dog)) (VP bit) (NP (DT a) (NN cat)))')
ParentedTree: Expected a node value and child list

This works, however:

tree = ParentedTree.fromstring('(S (NP (DT the) (JJ big) (NN dog)) (VP bit) (NP (DT a) (NN cat)))')