Feature naming conventions for positional information

mr-martian / rebabel-format

Python library for interacting with reBabel data files

MIT License

1 stars 4 forks source link

Feature naming conventions for positional information #4

Open mr-martian opened 2 months ago

mr-martian commented 2 months ago

Feature names have been a bit ad-hoc up to this point and we have acquired some inconsistencies.

Format	Time-aligned	Non-time-aligned
ELAN	`alignment:startime`, `alignment:endtime`	`alignment:index`
CoNLL-U	N/A	`meta:index` for words, nothing for sentences
FLExText	N/A?	`meta:index`
Macula	N/A	None
SFM	N/A	`meta:index` on words and morphemes, nothing on sentences
Text-Fabric	N/A	None

pandersity commented 1 month ago

FLEx does support time alignment, but only if it's been imported from ELAN. In other words, you can't input time alignment via FLEx's UI, but its data model does support it, and the fields BeginTimeOffset and EndTimeOffset can appear in .flextext files that are exported from FLEx.

Flex Data Model: https://github.com/sillsdev/liblcm/blob/master/src/SIL.LCModel/MasterLCModel.xml