tsolakghukasyan/thesis-resources

Word embeddings

Word embeddings, generated for the tokens in ArmTDP v2.3 train, test sets:

• fastText [dim=200,minn=1,maxn=3]
• no-fastText [dim=200,minn=1,maxn=3]
• so-fastText [dim=200,minn=2,maxn=4]
• average-of-BPEmb [dim=50,vs=50k]
• average-of-BPE-custom [dim=50,vs=25k]

COMBO models

Model files, obtained by training COMBO on ArmTDP v2.3 train set, using embeddings above:

• with fastText
• with no-fastText
• with so-fastText
• with average-of-BPEmb
• with average-of-BPE-custom