Amit Moryossef
Amit Moryossef
It would be interesting to compare the performance of the two approaches on different languages. We could use word lists (https://github.com/michmech/lemmatization-lists) which are context-free lemmas for words or we can...
Our current text-to-gloss translation can lead to ambiguity in the generated glosses. We propose enhancing our system by incorporating gloss sense disambiguation to better capture the intended meaning of the...
We need to improve the fluency and naturalness of the pose sequences generated by our system. Currently, our approach generates a sequence of poses that represent a sign language sentence,...
We need to improve our handling of unknown gloss variations. We propose developing a system that can modify a known gloss to generate the desired variation. ### Numbers We need...
We need a strategy for handling glosses that do not exist in our lexicon. One proposed solution is to leverage another lexicon with a written representation of the gloss (HamNoSys...
We should add resources from JW, like the bible.
Example code for loading the dataset into memory, including poses: https://github.com/sign-language-processing/transcription/blob/main/pose_to_segments/src/evaluate_mediapi.py#L69 - This should be able to load the face of mediapipe as well. - This should be able to...
- BSL: https://bslsignbank.ucl.ac.uk/dictionary/ - ASL: https://aslsignbank.haskins.yale.edu/signs/show_all/ - NGT: https://signbank.cls.ru.nl/signs/show_all/ - Finnish: https://signbank.csc.fi/dictionary/ - Auslan: https://auslan.org.au/dictionary/search/?query=A&category= - NZSL: https://www.nzsl.nz/ - Polish: https://www.slownikpjm.uw.edu.pl/en - SGSL: https://blogs.ntu.edu.sg/sgslsignbank/signs/ - Asian: http://cslds.org/asiansignbank/
WikiData includes some data in SignWriting, for example https://www.wikidata.org/wiki/Special:EntityData/Q14759.json Under "ase" it shows ```json "ase": { "language": "ase", "value": "M528x523S14c02497x497S14c0a472x500S2e85e483x478 M525x535S2e748483x510S10011501x466S2e704510x500S10019476x475 M551x515S1dc50504x485S1dc58474x485S26512449x501S26506536x501" }, ``` All the data with SignWriting: https://w.wiki/6LDX...
sentences are avaialble! https://github.com/how2sign/how2sign-data/issues/1#issuecomment-1248072979 should be able to load sentence level data