search

isi-vista / adam

Abduction to Demonstrate an Articulate Machine
MIT License
10 stars 4 forks source link

Introduce "I/me" and "You" to ObjectRecognizer when not using ObjectLearner #639

Open gabbard opened 4 years ago

gabbard commented 4 years ago

The SPEAKER/ADDRESSEE features should let you split I/me from you. I think we need to push the I/me distinction to phase 2. Little ids aren't good at it anyway.

gabbard commented 4 years ago

See https://github.com/isi-vista/adam/issues/640

gabbard commented 4 years ago

@LichMaster98 : is this complete?

lichtefeld commented 4 years ago

No, we can't currently recognize the object "I" using the ObjectRecognizer because the object "I/me" and "you" aren't in the list of objects in our OntologyNode found: https://github.com/isi-vista/adam/blob/master/adam/ontology/phase1_ontology.py#L449

Essentially we don't know how to create the pattern for "I" which is a problem related to: https://github.com/isi-vista/adam/issues/502 as "I/me" will also be recognized as "DAD" or "MOM' when this wouldn't be wrong just not the best answer.

paynesa commented 4 years ago

For M9 integrated learner with object recognizer, we don't recognize me or you currently -- working on debugging this.

lichtefeld commented 4 years ago

@paynesa I believe you fixed this issue?