[addresses] Adding increasing null_phrase_probability for plain numerics in Spain so things like 2o B make it into the training data
This commit is contained in:
@@ -360,6 +360,7 @@ cross_streets:
|
|||||||
sample: true
|
sample: true
|
||||||
canonical_probability: 0.8
|
canonical_probability: 0.8
|
||||||
sample_probability: 0.2
|
sample_probability: 0.2
|
||||||
|
parentheses_probabililty: 0.5
|
||||||
|
|
||||||
po_boxes:
|
po_boxes:
|
||||||
apartado: &apartado
|
apartado: &apartado
|
||||||
@@ -728,16 +729,6 @@ units:
|
|||||||
canonical: salas
|
canonical: salas
|
||||||
numeric:
|
numeric:
|
||||||
direction: left
|
direction: left
|
||||||
directions:
|
|
||||||
alternatives:
|
|
||||||
- alternative: *derecha
|
|
||||||
probability: 0.45
|
|
||||||
- alternative: *izquierda
|
|
||||||
probability: 0.45
|
|
||||||
- alternative: *trasera
|
|
||||||
probability: 0.05
|
|
||||||
- alternative: *frente
|
|
||||||
probability: 0.05
|
|
||||||
|
|
||||||
alphanumeric: &unit_alphanumeric
|
alphanumeric: &unit_alphanumeric
|
||||||
default: *departamento
|
default: *departamento
|
||||||
@@ -1023,13 +1014,13 @@ countries:
|
|||||||
<<: *puerta
|
<<: *puerta
|
||||||
numeric:
|
numeric:
|
||||||
direction: left
|
direction: left
|
||||||
null_phrase_probability: 0.05
|
# If it's just puerta B, many times it's just e.g. 3o B for "tercero piso puerta B"
|
||||||
# Unrelated to others. If it's just puerta B, most of the time don't include puerta
|
null_phrase_probability: 0.15
|
||||||
ordinal:
|
ordinal:
|
||||||
direction: right
|
direction: right
|
||||||
gender: f
|
gender: f
|
||||||
direction_probability: 0.95
|
direction_probability: 0.95
|
||||||
null_phrase_probability: 0.8
|
null_phrase_probability: 0.8 # Let e.g. 5a be the entire unit string
|
||||||
# These sum to 1
|
# These sum to 1
|
||||||
numeric_probability: 0.25
|
numeric_probability: 0.25
|
||||||
ordinal_probability: 0.75
|
ordinal_probability: 0.75
|
||||||
|
|||||||
Reference in New Issue
Block a user