lfoppiano / grobid-quantities / 32652882124

39%

push

github

Guidelines and documentation updates  (#195)

* doc(guidelines): the @unit attribute takes the lexicon notation (#39)

The question of which form the @unit attribute should take - `min` or `minute`,
`a` or `yr` for the year - was never answered, so the corpus has both.

Rules that the lexicon decides: the first notation of the entry when the unit
has one, the singular lemma otherwise. A unit missing from the lexicon is added
there rather than given an invented notation in the annotation.

This is a proposal: it is consistent with what most of the corpus already does,
but it is annotation policy and should be confirmed before the corpus is
normalised against it.

* doc(guidelines): rates of counted objects annotate the unit only (#46)

`600 h −1` for a zenithal hourly rate raised the question of whether `meteor/h`
is a unit and, if so, of what type.

Rules that it is not: the unit is the reciprocal time, annotated as FREQUENCY
with `unit="h^-1"`, and the counted object belongs to the quantified object. No
unit type has to be invented for a counted object.

This is a proposal and should be confirmed before the corpus follows it.

* doc(guidelines): resolutions annotate the unit, not the pixel (#43)

`40 mas per pixel` raised the question of whether `mas/pixel` is a unit and
whether a RESOLUTION type is needed.

Rules that the pixel is a counted object rather than a unit: only `mas` is
annotated, as ANGLE, and `per pixel` belongs to the quantified object. No new
unit type is needed, and the ruling matches the one for hourly rates.

This is a proposal and should be confirmed before the corpus follows it.

* doc(guidelines): currencies are annotated, with UNKNOWN as a type (#59)

Whether `150$` or `3 €` should be annotated at all was left open, so amounts of
money are absent from the corpus.

Rules that they are annotated, with `UNKNOWN` as the unit type until the
vocabulary has one - which the guidelines already allow in case of doubt.
Annotating them now... (continued)

1019 of 3002 branches covered (33.94%)

Branch coverage included in aggregate %.

2670 of 6463 relevant lines covered (41.31%)

1.65 hits per line