BoKenlm-syl - Tibetan KenLM Language Model

A KenLM n-gram language model trained on Tibetan text, tokenized with syllable tokenizer.

Model Details

Parameter	Value
Model Type	Modified Kneser-Ney 5-gram
Tokenizer	Tibetan syllable-based (split on tseg ་ / shad །)
Training Corpus	bo_corpus.txt
Pruning	0 0 1
Tokens	62,199,201
Vocabulary Size	101,462

N-gram Statistics

Order	Count	D1	D2	D3+
1	101,462	0.7273	0.9932	1.2615
2	1,656,776	0.7098	1.0359	1.3537
3	3,201,003	0.7505	1.0850	1.3624
4	5,576,668	0.8147	1.1485	1.3783
5	6,215,679	0.7784	1.2430	1.4828

Memory Estimates

Type	MB	Details
probing	348	assuming -p 1.5
probing	408	assuming -r models -p 1.5
trie	165	without quantization
trie	90	assuming -q 8 -b 8 quantization
trie	146	assuming -a 22 array pointer compression
trie	71	assuming -a 22 -q 8 -b 8 array pointer compression and quantization

Training Resources

Metric	Value
Peak Virtual Memory	12,333 MB
Peak RSS	2,981 MB
Wall Time	46.4s
User Time	45.4s
System Time	15.7s

Usage

import kenlm

model = kenlm.Model("BoKenlm-syl.arpa")

# Score a tokenized sentence
score = model.score("བོད་ སྐད་ ཀྱི་ ཚིག་ གྲུབ་ འདི་ ཡིན།")
print(score)

Files

• BoKenlm-syl.arpa — ARPA format language model
• README.md — This model card

License

Apache 2.0