Papers by Gregory Baker
The Construction and Evaluation of the LEAFTOP Dataset of Automatically Extracted Nouns in 1480 Languages (2022.lrec-1)
Copied to clipboard
| Challenge: | The Ethnologue reports the existence of 7,139 living languages. |
| Approach: | They propose to use probabilistic inference to identify likely translations in 1480 other languages . they use a Koine Greek New Testament as the source language and a custom scraper . |
| Outcome: | The proposed method provides lexiconaries with accuracy from 42% (Korafe) to 99% (Runyankole) it also provides language families where the technique works, and future improvements and extensions. |
Number Theory Meets Linguistics: Modelling Noun Pluralisation Across 1497 Languages Using 2-adic Metrics (2022.aacl-short)
Copied to clipboard
| Challenge: | Existing evidence to support p-adic neighbourhoods in languages is lacking . pmetrics are based on prime numbers and have infinitesimals to support calculus and triangle inequality to support geometry. |
| Approach: | They propose to use a linear regression problem to model pluralisation as a p-adic metric. |
| Outcome: | The proposed model outperforms Euclidean-space regressors on languages in Indo-European, Austronesian, Trans New-Guinea, Sino-Tibetan, Nilo-Saharan, Oto-Meanguean and Atlantic-Congo families. |