Access
- You can download Welsh word embeddings here. They are are available to developers under the FreeBSD licence.
- You can browse the Welsh word embedding space here.
- Welsh word embeddings come together with a game of words, which you can use to practice your Welsh.
- If you are a developer, you can evaluate your Welsh word embeddings using the following benchmarks.
| Test | Based on | Format | Explanation | Download |
|---|---|---|---|---|
| Relatedness | WordSimilarity-353 | word1,word2,score, e.g. llyfr,papur,7.46 |
word pair with their relatedness score [0..10] | wordsimilarity-353_cy |
| Similarity | SimLex-999 | word1,word2,POS,score, e.g. bendigedig,ardderchog,A,8.63 |
word pair with their similarity score [0..10] | simlex-999_cy |
| Association | USF | word1,word2,POS,...,score, e.g. miniog,pwl,A,...,1 |
word pair with their free association score {0,1} | simlex-999_cy |
| Synonymy | TOEFL | word1,word2,word3,word3,word5, e.g. hogyn,bachgen,menyw,gwraig,myfyriwr |
word with 4 choices for its synonym | synonyms_cy |
| Analogy | gensim | word1 word2 word3 word4, e.g. cymru cymreig serbia serbaidd |
two word pairs with analogous relationship | analogies_cy |
| Categorisation | WordNet classes | word,Category, e.g. olewydd,Ffrwyth |
214 words grouped into 13 categories | categories_cy |
