Parses for Gutenberg Children corpus

BERTRAM-weights-parses				- MST-parses using word-pair weights extracted from BERT-like Neural Net. Different experiments
capital-LGEnglish-noQuotes-manual 		- 200+ selected sentences manually parsed by linguist, using LG Parser version 5.5.1 tokenization
capital-random					- random parses generated from LG5.5.1/capital/parses
capital-sequential				- sequential parses generated from LG5.5.1/capital/parses
ChildrenGutenberg_dotless_annotated_MP005W2 	- Different MST-Parses of Adagram pre-disambiguated corpus with no dots (annotator params Min Prob=0.005, Window=2)
GC_lgAll 					- MST-Parses with LG-ANY-all mode (since 2019-04 - not in use because of tokenization inconsistent with LG-English)
GC_parses_win6_omdist 				- MST-Parses with clique-dist-win6, mst-dist modes (since 2019-04 - not in use because of tokenization inconsistent with LG-English)
GC-lg-cleaned					- MST-Parses of GC corpus tokenized by LG-English?.?.? (older than 5.5.1, tokenization difference less than 1% against 5.5.1)
LG 						- LG-English parses with some old LG Parser version (since 2019-04 - not in use) 
LG5.5.1 					- LG-English parses with LG Parser version 5.5.1
LG5.6.2 					- LG-English parses with LG Parser version 5.6.2
GC_LGEnglish_noQuotes_fullyParsed_psg		- Fully parsed sentences with no direct speech parsed with Anton's PSG parser 
GC_LGEnglish_noQuotes_fullyParsed_win6_omdist_parses - Fully parsed sentences with no direct speech parsed with MST-Parser R=6, Weight = 6/r, mst-dist = +1/r. Weights calculated only from lowercased version of http://langlearn.singularitynet.io/data/cleaned/English/Gutenberg-Children-Books/capital-LGEnglish-noQuotes-fullyParsed/GC_LGEnglish_noQuotes_fullyParsed.txt
GC_psg						- Parsed with Anton's PSG parser in non-incremental mode
GC_psg_inc					- Parsed with Anton's PSG parser in incremental mode
test 						- test corpora for references

For all MST-parsed files (also inside the folders above), the filename specifies the counting method used, as well as the mst-dist mode:
- For lg-ANY, the suffix "lg" is used, followed by the number of parses used (or All if all parses returned by LG-ANY were used). E.g. EnglishPOC_disamb-parses-lgAll.ull
- For clique methods, the suffix "win" is used, followed by the window size used. E.g. EnglishPOC_disamb-parses-win6.ull
- If distance weighting was used, the suffix "dist" is added at the end, prefixed by "o" if the weighting was used during the observe (counting) phase, or "m" if during mst-parsing. E.g. EnglishPOC_disamb-parses-win6-omdist.ull