BoKenlm-syl-v0.4 - 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 (botok-rs SimpleTokenizer)
Training Corpus	bo_corpus.txt
Pruning	0 0 1
Tokens	166,004,503
Vocabulary Size	20,564

N-gram Statistics

Order	Count	D1	D2	D3+
1	20,564	0.5906	0.9834	1.4625
2	1,542,980	0.6391	1.0500	1.4271
3	5,447,977	0.7179	1.0874	1.4057
4	11,954,834	0.7930	1.1368	1.3910
5	16,598,281	0.7284	1.2407	1.4638

Memory Estimates

Type	MB	Details
probing	719	assuming -p 1.5
probing	827	assuming -r models -p 1.5
trie	321	without quantization
trie	170	assuming -q 8 -b 8 quantization
trie	285	assuming -a 22 array pointer compression
trie	133	assuming -a 22 -q 8 -b 8 array pointer compression and quantization

Training Resources

Metric	Value
Peak Virtual Memory	12,333 MB
Peak RSS	4,702 MB
Wall Time	92.8s
User Time	91.7s
System Time	25.1s

Usage

import kenlm

model = kenlm.Model("BoKenlm-syl-v0.4.arpa")

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

Files

• BoKenlm-syl-v0.4.arpa — ARPA format language model
• README.md — This model card

License

Apache 2.0