Commits

Show all
Author Commit Message Labels Comments Date
David McClosky
CHECKLIST.txt: Make testing fail fast
David McClosky
MODELS.rst: Minor updates, spelling correction
David McClosky
setup.py: PEP8-ification
David McClosky
python/bllipparser/RerankingParser.py: Fix issue #36 We now fall back to independent_tags() if the parse fails in the tag() method (but you can disable this and get an error instead if you pass allow_failures=True).
David McClosky
second-stage/programs/features/sp-data.h: Fix issue #35 Python reranker module can now be compiled with optimization! The fix is a simple one -- don't let nparses() get inlined if we're building SWIG extensions using a function attribute. From what I can tell, the noinline attribute works for both gcc and clang (not that clang can currently compile the reranker, but we don't want to block that). (Thanks for the tips, @headdenw!)
David McClosky
first-stage/TRAIN/InputTree.C: Fix issue #34 Add check for empty strings from readNext(). Test case added. Must. Fix. Formatting (sort of).
David McClosky
first-stage/TRAIN/InputTree.C: Minor comment update.
David McClosky
README.rst: Add Liang Huang's forest extensions
David McClosky
first-stage/TRAIN/InputTree.C: Fix issue #33 We now reset InputTree's internal state whenever we get an error. Added some tests for tree reading bad trees. Also replaced some magic numbers with their proper constants and a couple other cleanups.
David McClosky
first-stage/PARSE/SentRep.C: remove cruft
David McClosky
first-stage and python: Safer maximum sentence length checks Turns out max_sentence_length (MAXSENTLEN) is not really the maximum sentence length but two more than it (yikes). See issue #32. first-stage/PARSE/parseIt.C: fix check to avoid segfaults for sentence lengths one fewer than user-requested maximum length. python/bllipparser/RerankingParser.py: fixed similar check as above python/tests/test_reranking_parser.py: added long sentence tests
David McClosky
second-stage/programs/eval-weights: Fix compilation bitrot
David McClosky
second-stage/programs/features/merge-feature-counts.py: Add merging tool Tool for merging extracted features in "cvlm" format and their feature mappings.
David McClosky
python/bllipparser/RerankerFeatureCorpus.py: add FeatureMapping class read_cvlm_weights() and read_cvlm_feature_mapping() moved and renamed to be classmethods as part of FeatureMapping (a subclass of dict). Also changed cvlm_header_format() -> cvlm_format_header() to improve consistency.
David McClosky
Python READMEs: Link to python/examples/ directory
David McClosky
python/bllipparser/RerankingParser.py: add Tree.head(), Tree.dependencies() Fixes issue #30. Also some doc updates. python/examples/headfinding.py: demo code for extracting head finding info
David McClosky
first-stage/README.rst: fix RST formatting error
David McClosky
READMEs: linkify references to other README files
David McClosky
python/bllipparser/RerankerFeatureCorpus.py: add cvlm_header_format() Also docstring improvements.
David McClosky
python/bllipparser/RerankingParser.py: Reparse when n-best list is empty In parse_constrained() if we have span constraints and the n-best list is empty, we simulate a parse failure to force reparsing with relaxed span constraints. Thanks DK for finding this bug!
David McClosky
python/bllipparser/RerankingParser.py: parse_constrained() stringifies terms This is a (temporary) issue since Tree.label returns a DeprecatedGetter, not a string. We also allow for span constraints to be a single string instead of a list of strings. New test case added for this.
David McClosky
python/bllipparser/ParsingShell.py: more flake8-y
David McClosky
CHECKLIST.txt: add flake8 RerankerFeatureCorpus.py
David McClosky
python/bllipparser/RerankingParser.py: stringify filesystem paths The parser doesn't like Unicode strings for paths.
David McClosky
first-stage/TRAIN/InputTree.C: Allow ROOT as a legal toplevel type
David McClosky
second-stage/programs/wlle/cvlm-lbfgs.cc: Fix weights writing bug This fixes a segfault if your weights file is set to be in a directory that doesn't exist.
David McClosky
python/bllipparser/ModelFetcher.py: Add DK's WSJ-PTB3 unified parsing model Various docs updated to prefer this over my less standard WSJ models.
David McClosky
MODELS.rst: New docs on which parsing model to use README.rst, README-python.rst: point to MODELS.rst
David McClosky
second-stage/README-retrain.rst: $NBEST is really $NPARSES Thanks DK for catching this!
David McClosky
second-stage/README-retrain.rst: Update training estimate
  1. Prev
  2. Next