Amir Zeldes

There should be a button on spans whose parent is a span to collapse one level as a convenience method to reduce redundant spans.

Add support for schemas

1

Currently, the schema relation type is not supported. The declaration must be distinguished on import, which currently fails for entries like: ``` ``` This should be added and visualized as...

The setting `max_suffix_length` in config.ini isn't actually being used here: https://github.com/amir-zeldes/xrenner/blob/master/xrenner/modules/xrenner_marker.py#L631-L648 This should be fixed (read feature from lex.filters["max_suffix_length"] and apply to substring window size)

Allow pluggable genre classifier (e.g. Naive Bayes), dump genre during training data generation and allow as feature for classification. Requires #88 to be able to read whole document before processing...

Needed for Chinese non-closed list definite tokens, consisting of demonstrative + counter, but always labeled 'det': 这个 (only with det!) 各 那个

Chinese non named entities

1

- [ ] Assign major entity types to top Chinese nouns (500/500) - [x] Also disambiguate 600 NORP items

Some languages order names: First-Last (Chinese, Japanese etc.) There should be a setting in config.ini to configure this and xrenner_marker.py should take this into consideration.

For non-Arabic numeral script languages, the numbers.tab file is a partial solution for cardinality detection. There should be some pluggable mechanism to e.g. parse Chinese numbers into int/float

See auto version of GUM_fiction_oversite

Preconj children should only be attached to coordinate markables: * Individual markables without `preconj`: both [A] and [B] * Coordinate markable with `preconj`: [both A and B]