README.md

metadata

language: mni
language_name: Manipuri
language_family: tibetoburman_other
tags:
  - wikilangs
  - nlp
  - tokenizer
  - embeddings
  - n-gram
  - markov
  - wikipedia
  - feature-extraction
  - sentence-similarity
  - tokenization
  - n-grams
  - markov-chain
  - text-mining
  - fasttext
  - babelvec
  - vocabulous
  - vocabulary
  - monolingual
  - family-tibetoburman_other
license: mit
library_name: wikilangs
pipeline_tag: text-generation
datasets:
  - omarkamali/wikipedia-monthly
dataset_info:
  name: wikipedia-monthly
  description: Monthly snapshots of Wikipedia articles across 300+ languages
metrics:
  - name: best_compression_ratio
    type: compression
    value: 4.321
  - name: best_isotropy
    type: isotropy
    value: 0.6424
  - name: vocabulary_size
    type: vocab
    value: 0
generated: 2026-01-10T00:00:00.000Z

Manipuri - Wikilangs Models

Comprehensive Research Report & Full Ablation Study

This repository contains NLP models trained and evaluated by Wikilangs, specifically on Manipuri Wikipedia data. We analyze tokenizers, n-gram models, Markov chains, vocabulary statistics, and word embeddings.

📋 Repository Contents

Models & Assets

• Tokenizers (8k, 16k, 32k, 64k)
• N-gram models (2, 3, 4, 5-gram)
• Markov chains (context of 1, 2, 3, 4 and 5)
• Subword N-gram and Markov chains
• Embeddings in various sizes and dimensions (aligned and unaligned)
• Language Vocabulary
• Language Statistics

Analysis and Evaluation

• 1. Tokenizer Evaluation
• 2. N-gram Model Evaluation
• 3. Markov Chain Evaluation
• 4. Vocabulary Analysis
• 5. Word Embeddings Evaluation
• 6. Morphological Analysis (Experimental)
• 7. Summary & Recommendations
• Metrics Glossary
• Visualizations Index

---

1. Tokenizer Evaluation

Results

Vocab Size	Compression	Avg Token Len	UNK Rate	Total Tokens
8k	3.393x	3.40	0.2560%	174,579
16k	3.741x	3.75	0.2824%	158,309
32k	4.017x	4.02	0.3031%	147,458
64k	4.321x 🏆	4.33	0.3261%	137,077

Tokenization Examples

Below are sample sentences tokenized with each vocabulary size:

Sample 1: ꯂꯥꯢꯔꯤꯛ ꯄꯔꯤꯡ (ꯕꯨꯛ ꯁꯤꯔꯤꯁ) ꯑꯁꯤ ꯑꯉꯥꯡꯁꯤꯡꯒꯤ ꯂꯥꯢꯔꯤꯛꯁꯤꯡꯒꯤ ꯃꯅꯨꯡꯗꯒꯤ ꯑꯃꯅꯤ ꯫ ꯃꯁꯤꯁꯨ ꯌꯦꯡꯕꯤꯌꯨ ꯃ...

Vocab	Tokens	Count
8k	▁ꯂꯥꯢꯔꯤꯛ ▁ꯄꯔꯤꯡ ▁( ꯕꯨꯛ ▁ꯁꯤꯔꯤꯁ ) ▁ꯑꯁꯤ ▁ꯑꯉꯥꯡꯁꯤꯡꯒꯤ ▁ꯂꯥꯢꯔꯤꯛꯁꯤꯡꯒꯤ ▁ꯃꯅꯨꯡꯗꯒꯤ ... (+7 more)	17
16k	▁ꯂꯥꯢꯔꯤꯛ ▁ꯄꯔꯤꯡ ▁( ꯕꯨꯛ ▁ꯁꯤꯔꯤꯁ ) ▁ꯑꯁꯤ ▁ꯑꯉꯥꯡꯁꯤꯡꯒꯤ ▁ꯂꯥꯢꯔꯤꯛꯁꯤꯡꯒꯤ ▁ꯃꯅꯨꯡꯗꯒꯤ ... (+7 more)	17
32k	▁ꯂꯥꯢꯔꯤꯛ ▁ꯄꯔꯤꯡ ▁( ꯕꯨꯛ ▁ꯁꯤꯔꯤꯁ ) ▁ꯑꯁꯤ ▁ꯑꯉꯥꯡꯁꯤꯡꯒꯤ ▁ꯂꯥꯢꯔꯤꯛꯁꯤꯡꯒꯤ ▁ꯃꯅꯨꯡꯗꯒꯤ ... (+7 more)	17
64k	▁ꯂꯥꯢꯔꯤꯛ ▁ꯄꯔꯤꯡ ▁( ꯕꯨꯛ ▁ꯁꯤꯔꯤꯁ ) ▁ꯑꯁꯤ ▁ꯑꯉꯥꯡꯁꯤꯡꯒꯤ ▁ꯂꯥꯢꯔꯤꯛꯁꯤꯡꯒꯤ ▁ꯃꯅꯨꯡꯗꯒꯤ ... (+7 more)	17

Sample 2: ꯃꯁꯤꯡ ꯑꯁꯤ ꯗꯒꯤ ꯍꯦꯟꯅ ꯆꯥꯎꯕ ꯃꯁꯤꯡ ꯑꯃꯅꯤ꯫ ꯃꯁꯤꯗꯒꯤ ꯃꯁꯤꯡ ꯱ ꯍꯦꯟꯅ ꯆꯥꯎꯕ ꯃꯁꯤꯡ ꯗꯤ ꯅꯤ꯫ ꯃꯇꯦꯡ ꯂꯧꯔꯛꯐ...

Vocab	Tokens	Count
8k	▁ꯃꯁꯤꯡ ▁ꯑꯁꯤ ▁ꯗꯒꯤ ▁ꯍꯦꯟꯅ ▁ꯆꯥꯎꯕ ▁ꯃꯁꯤꯡ ▁ꯑꯃꯅꯤ꯫ ▁ꯃꯁꯤꯗꯒꯤ ▁ꯃꯁꯤꯡ ▁꯱ ... (+11 more)	21
16k	▁ꯃꯁꯤꯡ ▁ꯑꯁꯤ ▁ꯗꯒꯤ ▁ꯍꯦꯟꯅ ▁ꯆꯥꯎꯕ ▁ꯃꯁꯤꯡ ▁ꯑꯃꯅꯤ꯫ ▁ꯃꯁꯤꯗꯒꯤ ▁ꯃꯁꯤꯡ ▁꯱ ... (+11 more)	21
32k	▁ꯃꯁꯤꯡ ▁ꯑꯁꯤ ▁ꯗꯒꯤ ▁ꯍꯦꯟꯅ ▁ꯆꯥꯎꯕ ▁ꯃꯁꯤꯡ ▁ꯑꯃꯅꯤ꯫ ▁ꯃꯁꯤꯗꯒꯤ ▁ꯃꯁꯤꯡ ▁꯱ ... (+11 more)	21
64k	▁ꯃꯁꯤꯡ ▁ꯑꯁꯤ ▁ꯗꯒꯤ ▁ꯍꯦꯟꯅ ▁ꯆꯥꯎꯕ ▁ꯃꯁꯤꯡ ▁ꯑꯃꯅꯤ꯫ ▁ꯃꯁꯤꯗꯒꯤ ▁ꯃꯁꯤꯡ ▁꯱ ... (+11 more)	21

Sample 3: ꯑꯂꯤ ꯐꯥꯓꯜ ꯑꯁꯤ ꯏꯟꯗꯤꯌꯥꯒꯤ ꯍꯤꯟꯗꯤ ꯂꯣꯟꯒꯤ ꯕꯣꯜꯂꯤꯋꯨꯗ (ꯍꯤꯟꯗꯤ ꯃꯃꯤ ꯀꯨꯝꯃꯩ)ꯒꯤ ꯁꯛꯇꯝ ꯂꯥꯡꯕ ꯁꯤꯟꯂꯣꯢ ...

Vocab	Tokens	Count
8k	▁ꯑ ꯂꯤ ▁ꯐꯥ ꯓ ꯜ ▁ꯑꯁꯤ ▁ꯏꯟꯗꯤꯌꯥꯒꯤ ▁ꯍꯤꯟꯗꯤ ▁ꯂꯣꯟꯒꯤ ▁ꯕꯣꯜꯂꯤꯋꯨꯗ ... (+18 more)	28
16k	▁ꯑꯂꯤ ▁ꯐꯥ ꯓ ꯜ ▁ꯑꯁꯤ ▁ꯏꯟꯗꯤꯌꯥꯒꯤ ▁ꯍꯤꯟꯗꯤ ▁ꯂꯣꯟꯒꯤ ▁ꯕꯣꯜꯂꯤꯋꯨꯗ ▁( ... (+17 more)	27
32k	▁ꯑꯂꯤ ▁ꯐꯥ ꯓ ꯜ ▁ꯑꯁꯤ ▁ꯏꯟꯗꯤꯌꯥꯒꯤ ▁ꯍꯤꯟꯗꯤ ▁ꯂꯣꯟꯒꯤ ▁ꯕꯣꯜꯂꯤꯋꯨꯗ ▁( ... (+17 more)	27
64k	▁ꯑꯂꯤ ▁ꯐꯥꯓꯜ ▁ꯑꯁꯤ ▁ꯏꯟꯗꯤꯌꯥꯒꯤ ▁ꯍꯤꯟꯗꯤ ▁ꯂꯣꯟꯒꯤ ▁ꯕꯣꯜꯂꯤꯋꯨꯗ ▁( ꯍꯤꯟꯗꯤ ▁ꯃꯃꯤ ... (+15 more)	25

Key Findings

• Best Compression: 64k achieves 4.321x compression
• Lowest UNK Rate: 8k with 0.2560% unknown tokens
• Trade-off: Larger vocabularies improve compression but increase model size
• Recommendation: 32k vocabulary provides optimal balance for production use

---

2. N-gram Model Evaluation

Results

N-gram	Variant	Perplexity	Entropy	Unique N-grams	Top-100 Coverage	Top-1000 Coverage
2-gram	Word	1,742	10.77	10,310	35.2%	72.4%
2-gram	Subword	1,226 🏆	10.26	15,112	41.9%	79.3%
3-gram	Word	1,239	10.28	9,547	36.7%	82.6%
3-gram	Subword	7,080	12.79	65,488	23.1%	50.6%
4-gram	Word	1,700	10.73	18,256	32.5%	80.1%
4-gram	Subword	22,325	14.45	195,890	16.2%	37.5%
5-gram	Word	1,478	10.53	14,280	31.6%	83.2%
5-gram	Subword	35,425	15.11	266,503	14.0%	33.1%

Top 5 N-grams by Size

2-grams (Word):

Rank	N-gram	Count
1	ꯃꯇꯦꯡ ꯂꯧꯔꯛꯐꯝ	9,094
2	ꯃꯁꯤꯡ ꯑꯁꯤ	6,658
3	ꯆꯥꯎꯕ ꯃꯁꯤꯡ	5,847
4	ꯍꯦꯟꯅ ꯆꯥꯎꯕ	5,424
5	ꯃꯁꯤꯁꯨ ꯌꯦꯡꯕꯤꯌꯨ	4,237

3-grams (Word):

Rank	N-gram	Count
1	ꯍꯦꯟꯅ ꯆꯥꯎꯕ ꯃꯁꯤꯡ	5,408
2	ꯗꯒꯤ ꯍꯦꯟꯅ ꯆꯥꯎꯕ	4,147
3	ꯌꯦꯡꯕꯤꯌꯨ ꯃꯇꯦꯡ ꯂꯧꯔꯛꯐꯝ	3,732
4	ꯃꯁꯤꯁꯨ ꯌꯦꯡꯕꯤꯌꯨ ꯃꯇꯦꯡ	3,724
5	ꯂꯧꯔꯛꯐꯝ ꯗꯒꯤ ꯍꯦꯟꯅ	2,348

4-grams (Word):

Rank	N-gram	Count
1	ꯗꯒꯤ ꯍꯦꯟꯅ ꯆꯥꯎꯕ ꯃꯁꯤꯡ	4,146
2	ꯃꯁꯤꯁꯨ ꯌꯦꯡꯕꯤꯌꯨ ꯃꯇꯦꯡ ꯂꯧꯔꯛꯐꯝ	3,723
3	ꯂꯧꯔꯛꯐꯝ ꯗꯒꯤ ꯍꯦꯟꯅ ꯆꯥꯎꯕ	2,348
4	ꯃꯇꯦꯡ ꯂꯧꯔꯛꯐꯝ ꯗꯒꯤ ꯍꯦꯟꯅ	2,348
5	ꯍꯦꯟꯅ ꯆꯥꯎꯕ ꯃꯁꯤꯡ ꯑꯃꯅꯤ	1,590

5-grams (Word):

Rank	N-gram	Count
1	ꯂꯧꯔꯛꯐꯝ ꯗꯒꯤ ꯍꯦꯟꯅ ꯆꯥꯎꯕ ꯃꯁꯤꯡ	2,348
2	ꯃꯇꯦꯡ ꯂꯧꯔꯛꯐꯝ ꯗꯒꯤ ꯍꯦꯟꯅ ꯆꯥꯎꯕ	2,348
3	ꯗꯒꯤ ꯍꯦꯟꯅ ꯆꯥꯎꯕ ꯃꯁꯤꯡ ꯑꯃꯅꯤ	1,585
4	ꯑꯁꯤ ꯗꯒꯤ ꯍꯦꯟꯅ ꯆꯥꯎꯕ ꯃꯁꯤꯡ	1,585
5	ꯃꯁꯤꯡ ꯑꯁꯤ ꯗꯒꯤ ꯍꯦꯟꯅ ꯆꯥꯎꯕ	1,582

2-grams (Subword):

Rank	N-gram	Count
1	_ ꯃ	104,012
2	_ ꯑ	89,663
3	ꯡ _	60,916
4	ꯒꯤ _	53,149
5	ꯁꯤ ꯡ	47,278

3-grams (Subword):

Rank	N-gram	Count
1	_ ꯑ ꯃ	33,170
2	_ ꯃ ꯁꯤ	28,206
3	ꯁꯤ ꯡ _	26,827
4	_ ꯑ ꯁꯤ	22,140
5	ꯃ ꯁꯤ ꯡ	19,797

4-grams (Subword):

Rank	N-gram	Count
1	_ ꯑ ꯁꯤ _	18,348
2	_ ꯃ ꯁꯤ ꯡ	16,785
3	ꯃ ꯁꯤ ꯡ _	16,509
4	ꯁꯤ ꯡ _ ꯑ	14,961
5	_ ꯑ ꯃ ꯅꯤ	11,342

5-grams (Subword):

Rank	N-gram	Count
1	_ ꯃ ꯁꯤ ꯡ _	13,768
2	ꯃ ꯁꯤ ꯡ _ ꯑ	11,210
3	ꯑ ꯃ ꯁꯨ ꯡ _	10,162
4	_ ꯑ ꯃ ꯁꯨ ꯡ	10,153
5	_ ꯃ ꯇꯦ ꯡ _	9,972

Key Findings

• Best Perplexity: 2-gram (subword) with 1,226
• Entropy Trend: Decreases with larger n-grams (more predictable)
• Coverage: Top-1000 patterns cover ~33% of corpus
• Recommendation: 4-gram or 5-gram for best predictive performance

---

3. Markov Chain Evaluation

Results

Context	Variant	Avg Entropy	Perplexity	Branching Factor	Unique Contexts	Predictability
1	Word	0.7036	1.629	3.89	86,410	29.6%
1	Subword	1.2295	2.345	13.46	2,717	0.0%
2	Word	0.1754	1.129	1.33	335,779	82.5%
2	Subword	0.8032	1.745	4.49	36,564	19.7%
3	Word	0.0432	1.030	1.06	446,350	95.7%
3	Subword	0.5398	1.454	2.69	164,127	46.0%
4	Word	0.0126 🏆	1.009	1.02	471,596	98.7%
4	Subword	0.3671	1.290	1.81	440,820	63.3%

Generated Text Samples (Word-based)

Below are text samples generated from each word-based Markov chain model:

Context Size 1:

1. ꯑꯁꯤ ꯕꯤ ꯗꯤ ꯑꯦꯂꯦꯝꯕꯔꯇꯀꯤ ꯑꯦꯟꯁꯥꯏꯛꯂꯣꯄꯤꯗꯤ ꯑꯃꯁꯨꯡ ꯂꯂꯣꯟ ꯏꯇꯤꯛ ꯇꯧꯅꯕꯥ ꯅꯥꯠꯀꯤ ꯑꯣꯏꯕ ꯋꯥꯊꯣꯛ ꯂꯝꯖꯤꯡ ꯂꯝꯇꯥꯛꯄ ꯑꯗꯨꯗꯤ ꯑꯍꯨꯝꯁꯨꯕ
2. ꯃꯁꯤꯡ ꯍꯥꯏꯄꯇꯤ ꯆꯩꯁꯤꯡ ꯃꯁꯤꯡ ꯑꯃꯅꯦ ꯃꯊꯪꯀꯤ ꯃꯁꯤꯡ ꯸꯹꯹ ꯐꯥꯎꯕ ꯂꯩꯕ ꯆꯩꯁꯤꯡ ꯑꯁꯤ ꯃꯤꯇꯩꯀꯤ ꯆꯩꯁꯤꯡ ꯏꯌꯦꯛꯅ ꯁꯦꯝꯄꯅꯦ
3. ꯑꯃꯅꯤ ꯃꯁꯤꯗꯒꯤ ꯃꯁꯤꯡ ꯍꯥꯏꯄꯇꯤ ꯆꯩꯁꯤꯡ ꯏꯌꯦꯛꯅ ꯁꯦꯝꯄꯅꯤ ꯃꯁꯤꯡ ꯸꯰꯰ ꯃꯇꯦꯡ ꯂꯧꯔꯛꯐꯝ ꯃꯄꯥꯟꯒ ꯁꯝꯅꯕꯁꯤꯡ out in ukrainian

Context Size 2:

1. ꯃꯇꯦꯡ ꯂꯧꯔꯛꯐꯝ ꯃꯃꯤ ꯀꯨꯝꯃꯩ ꯁꯛꯇꯝ ꯂꯥꯡꯕꯤ ꯑꯣꯢꯈꯤ ꯃꯇꯦꯡ ꯂꯧꯔꯛꯐꯝ ꯏꯁꯩꯁꯤꯡ
2. ꯃꯁꯤꯡ ꯑꯁꯤ ꯗꯒꯤ ꯍꯦꯟꯅ ꯆꯥꯎꯕ ꯃꯁꯤꯡ ꯑꯃꯅꯤ ꯃꯁꯤꯡ ꯑꯁꯤ ꯗꯒꯤ ꯍꯦꯟꯅ ꯆꯥꯎꯕ ꯃꯁꯤꯡ ꯗꯤ ꯅꯤ ꯃꯇꯦꯡ ꯂꯧꯔꯛꯐꯝ
3. ꯍꯦꯟꯅ ꯆꯥꯎꯕ ꯃꯁꯤꯡ ꯗꯤ ꯅꯤ ꯆꯩꯁꯤꯡꯂꯣꯟ ꯆꯩꯁꯤꯡꯂꯣꯟ ꯗ ꯌꯦꯡꯂꯗꯤ ꯃꯁꯤꯡ ꯑꯁꯤ ꯃꯤꯇꯩꯀꯤ ꯆꯩꯁꯤꯡ ꯏꯌꯦꯛꯅ ꯁꯦꯝꯄꯅꯦ ꯃꯁꯤꯡ ꯑꯁꯤ

Context Size 3:

1. ꯗꯒꯤ ꯍꯦꯟꯅ ꯆꯥꯎꯕ ꯃꯁꯤꯡ ꯗꯤ ꯃꯇꯦꯡ ꯂꯧꯔꯛꯐꯝ ꯗꯒꯤ ꯍꯦꯟꯅ ꯆꯥꯎꯕ ꯃꯁꯤꯡ ꯑꯃꯅꯤ ꯀꯥꯈꯟ ꯴ ꯂꯩꯕ ꯃꯁꯤꯡ ꯑꯃꯅꯤ ꯃꯁꯤꯁꯨ
2. ꯃꯁꯤꯁꯨ ꯌꯦꯡꯕꯤꯌꯨ ꯃꯇꯦꯡ ꯂꯧꯔꯛꯐꯝ ꯂꯥꯢꯔꯤꯛꯁꯤꯡ khamlangba erengba puwaree neinarol by yaima lamgdum kakching ha...
3. ꯌꯦꯡꯕꯤꯌꯨ ꯃꯇꯦꯡ ꯂꯧꯔꯛꯐꯝ ꯄꯅꯥꯁꯤꯡ ꯄꯅꯥ

Context Size 4:

1. ꯃꯁꯤꯁꯨ ꯌꯦꯡꯕꯤꯌꯨ ꯃꯇꯦꯡ ꯂꯧꯔꯛꯐꯝ ꯃꯌꯥꯝꯒꯤ ꯑꯁꯨꯞꯄ ꯆꯠꯅ ꯀꯥꯡꯂꯣꯟꯁꯤꯡ
2. ꯃꯇꯦꯡ ꯂꯧꯔꯛꯐꯝ ꯗꯒꯤ ꯍꯦꯟꯅ ꯆꯥꯎꯕ ꯃꯁꯤꯡ ꯑꯃꯅꯤ ꯃꯁꯤꯗꯒꯤ ꯃꯁꯤꯡ ꯱ ꯍꯦꯟꯅ ꯆꯥꯎꯕ ꯃꯁꯤꯡ ꯗꯤ ꯅꯤ ꯆꯩꯁꯤꯡꯂꯣꯟ ꯆꯩꯁꯤꯡꯂꯣꯟ ꯗ ꯌꯦꯡꯂꯗꯤ
3. ꯂꯧꯔꯛꯐꯝ ꯗꯒꯤ ꯍꯦꯟꯅ ꯆꯥꯎꯕ ꯃꯁꯤꯡ ꯑꯃꯅꯤ ꯃꯁꯤꯗꯒꯤ ꯃꯁꯤꯡ ꯱ ꯍꯦꯟꯅ ꯆꯥꯎꯕ ꯃꯁꯤꯡ ꯗꯤ ꯅꯤ ꯆꯩꯁꯤꯡꯂꯣꯟ ꯆꯩꯁꯤꯡꯂꯣꯟ ꯗ ꯌꯦꯡꯂꯗꯤ ꯃꯁꯤꯡ

Generated Text Samples (Subword-based)

Below are text samples generated from each subword-based Markov chain model:

Context Size 1:

1. _꯱_ꯂꯧꯁꯤꯅ_ꯑꯃꯁꯤꯡꯄꯤ_ꯅꯣꯐ_
2. ꯃꯆꯠꯇꯦꯝꯊꯣꯛꯀꯤ_ꯃꯗꯤꯂꯣꯟ_ꯃꯛ
3. ꯡ_ꯃꯥ_ꯐꯣꯟ_ꯆꯥꯎꯆꯦꯟꯅꯔꯤꯕ_ꯆꯥ

Context Size 2:

1. _ꯃꯌꯦꯛꯇꯤ_"le_phe_ger
2. _ꯑꯁꯤ_ꯑꯣꯗꯤꯁꯤꯡꯒꯤ_ꯍꯧꯔꯀꯈꯤ_꯫_
3. ꯡ_ꯑꯆꯧꯄ_ꯀꯪꯂꯩꯄꯥꯛ_ꯑꯃꯃꯁꯤꯁꯨ_

Context Size 3:

1. _ꯑꯃꯁꯨ_ꯌꯦꯡꯕꯤꯌꯨ_ꯃꯇꯦꯡ_ꯂꯧꯔꯛꯐ
2. _ꯃꯁꯤꯁꯨ_ꯌꯦꯡꯉꯨ_꯴꯰_(ꯆꯥꯍꯨꯝ_(
3. ꯁꯤꯡ_ꯑꯃꯗꯒꯤ_ꯃꯍꯨꯠꯇ,_ꯍꯦꯗꯤꯁꯇ

Context Size 4:

1. _ꯑꯁꯤ_ꯑꯆꯧꯕ_ꯑꯆꯥꯄꯣꯠ_ꯑꯃꯗꯥ_ꯑꯣꯄꯦ
2. ꯃꯁꯤꯡ_ꯑꯣꯏꯅ_ꯊꯝꯕ_ꯌꯥꯏ,_ꯃꯅꯥ-
3. ꯁꯤꯡ_ꯑꯁꯤꯁꯨ_ꯌꯥꯝꯅ_ꯁꯨꯔꯨꯡꯁꯤꯡꯗ_ꯅ

Key Findings

• Best Predictability: Context-4 (word) with 98.7% predictability
• Branching Factor: Decreases with context size (more deterministic)
• Memory Trade-off: Larger contexts require more storage (440,820 contexts)
• Recommendation: Context-3 or Context-4 for text generation

---

4. Vocabulary Analysis

Statistics

Metric	Value
Vocabulary Size	35,928
Total Tokens	676,105
Mean Frequency	18.82
Median Frequency	3
Frequency Std Dev	209.83

Most Common Words

Rank	Word	Frequency
1	ꯃꯁꯤꯡ	18,949
2	ꯑꯁꯤ	18,391
3	ꯑꯃꯅꯤ	11,341
4	ꯑꯃꯁꯨꯡ	10,185
5	ꯃꯇꯦꯡ	10,150
6	ꯂꯧꯔꯛꯐꯝ	9,121
7	ꯍꯦꯟꯅ	6,892
8	ꯆꯥꯎꯕ	6,104
9	ꯃꯁꯤꯁꯨ	5,695
10	ꯌꯦꯡꯕꯤꯌꯨ	4,565

Least Common Words (from vocabulary)

Rank	Word	Frequency
1	ꯖꯡꯁꯟ	2
2	ꯁ꯭ꯀꯣꯜꯀꯣꯞ	2
3	ꯃꯥꯏꯀ꯭ꯔꯣꯋꯦꯚ	2
4	ꯁ꯭ꯋꯔꯠ	2
5	ꯃꯤꯆꯤꯌꯁ	2
6	ꯅꯥꯀꯃꯨꯔꯥ	2
7	ꯁꯔꯀ꯭ꯌꯨꯠ	2
8	ꯇꯟꯅꯦꯂꯤꯡ	2
9	ꯁꯥꯏꯀꯃꯣꯔ	2
10	ꯗꯦꯕꯥꯏꯁ	2

Zipf's Law Analysis

Metric	Value
Zipf Coefficient	1.0805
R² (Goodness of Fit)	0.996289
Adherence Quality	excellent

Coverage Analysis

Top N Words	Coverage
Top 100	35.5%
Top 1,000	65.4%
Top 5,000	82.5%
Top 10,000	88.9%

Key Findings

• Zipf Compliance: R²=0.9963 indicates excellent adherence to Zipf's law
• High Frequency Dominance: Top 100 words cover 35.5% of corpus
• Long Tail: 25,928 words needed for remaining 11.1% coverage

---

5. Word Embeddings Evaluation

5.1 Cross-Lingual Alignment

5.2 Model Comparison

Model	Dimension	Isotropy	Semantic Density	Alignment R@1	Alignment R@10
mono_32d	32	0.6424	0.3709	N/A	N/A
mono_64d	64	0.3014	0.3657	N/A	N/A
mono_128d	128	0.0542	0.3495	N/A	N/A
aligned_32d	32	0.6424 🏆	0.3667	0.0080	0.0540
aligned_64d	64	0.3014	0.3759	0.0060	0.0480
aligned_128d	128	0.0542	0.3530	0.0040	0.0620

Key Findings

• Best Isotropy: aligned_32d with 0.6424 (more uniform distribution)
• Semantic Density: Average pairwise similarity of 0.3636. Lower values indicate better semantic separation.
• Alignment Quality: Aligned models achieve up to 0.8% R@1 in cross-lingual retrieval.
• Recommendation: 128d aligned for best cross-lingual performance

---

6. Morphological Analysis (Experimental)

This section presents an automated morphological analysis derived from the statistical divergence between word-level and subword-level models. By analyzing where subword predictability spikes and where word-level coverage fails, we can infer linguistic structures without supervised data.

6.1 Productivity & Complexity

Metric	Value	Interpretation	Recommendation
Productivity Index	5.000	High morphological productivity	Reliable analysis
Idiomaticity Gap	1.511	High formulaic/idiomatic content	-

6.2 Affix Inventory (Productive Units)

These are the most productive prefixes and suffixes identified by sampling the vocabulary for global substitutability patterns. A unit is considered an affix if stripping it leaves a valid stem that appears in other contexts.

Productive Prefixes

Prefix	Examples
-ꯑ	ꯑꯜꯒꯥ, ꯑꯍꯨꯃꯁꯨꯕꯥ, ꯑꯣꯢꯕꯤ
-ꯃ	ꯃꯄꯨꯔꯣꯏꯕꯅ, ꯃꯤꯠꯅꯥ, ꯃꯔꯨꯞꯁꯤꯡꯅ
-ꯁ	ꯁꯥꯗ꯭ꯗꯥꯝꯕꯨ, ꯁ꯭ꯇꯥꯝꯞ, ꯁ꯭ꯇꯥꯐ
-ꯂ	ꯂꯩꯔꯝꯕ, ꯂꯤꯊ꯭ꯋꯥꯅꯤꯌꯥꯟ, ꯂꯥꯟꯃꯤꯒꯤ
-ꯀ	ꯀꯧꯁꯤꯀꯥ, ꯀꯥꯡꯕ, ꯀꯅꯕꯤꯒꯅꯤ
-ꯄ	ꯄꯨꯔꯥꯛꯂꯦ, ꯄꯤꯖꯕ, ꯄꯦꯒꯦꯠ
-ꯍ	ꯍꯧꯖꯤꯛꯁꯨ, ꯍꯨꯝꯐꯨꯇꯔꯥꯍꯨꯝꯊꯣꯏ, ꯍꯣꯡꯗꯣꯛꯇꯨꯅ
-ꯇ	ꯇꯥꯛꯅꯕꯒꯤ, ꯇꯃꯊꯤꯕꯥ, ꯇꯪꯈꯥꯏꯗꯥ

Productive Suffixes

Suffix	Examples
-ꯡ	ꯈꯣꯡ, ꯆꯤꯊꯤꯁꯤꯡ, ꯇꯃꯦꯡꯂꯣꯡ
-ꯟ	ꯂꯤꯊ꯭ꯋꯥꯅꯤꯌꯥꯟ, ꯌꯥꯡꯈꯩꯃꯥꯄꯟ, ꯑꯣꯐꯤꯌꯣꯟ
-ꯅ	ꯃꯄꯨꯔꯣꯏꯕꯅ, ꯂꯩꯇꯕꯅ, ꯃꯔꯨꯞꯁꯤꯡꯅ
-ꯗ	꯹ꯗ, ꯋꯥꯔꯜꯗꯋꯥꯏꯗ, ꯃꯤꯇꯩꯂꯣꯟꯗ
-ꯕ	ꯂꯩꯔꯝꯕ, ꯎꯞꯂꯕ, ꯄꯤꯖꯕ
-s	epirus, chris, andreas
-ꯝ	ꯂꯝꯇꯥꯛꯂꯣꯢꯁꯛꯇꯝ, ꯉꯥꯉꯝ, ꯁꯤꯟꯂꯝ
-ꯛ	ꯑꯥꯏꯀꯣꯅꯣꯒ꯭ꯔꯥꯐꯤꯛ, ꯊꯕꯛꯁꯤꯡꯒꯤꯗꯃꯛ, ꯇꯧꯕꯒꯤꯗꯃꯛ

6.3 Bound Stems (Lexical Roots)

Bound stems are high-frequency subword units that are semantically cohesive but rarely appear as standalone words. These often correspond to the 'core' of a word that requires inflection or derivation to be valid.

Stem	Cohesion	Substitutability	Examples
ther	2.36x	17 contexts	uther, other, there
tion	2.34x	15 contexts	nation, action, motion
atio	2.35x	10 contexts	ratio, nation, nations
ꯍꯜꯂꯛ	1.80x	19 contexts	ꯍꯜꯂꯛꯄ, ꯍꯜꯂꯛꯏ, ꯍꯜꯂꯛꯄꯒ
ꯝꯅꯃꯛ	1.89x	12 contexts	ꯄꯨꯝꯅꯃꯛ, ꯄꯨꯝꯅꯃꯛꯅ, ꯄꯨꯝꯅꯃꯛꯇ
ꯔꯛꯐꯝ	1.60x	11 contexts	ꯂꯔꯛꯐꯝ, ꯂꯧꯔꯛꯐꯝ, ꯍꯧꯔꯛꯐꯝ

6.4 Affix Compatibility (Co-occurrence)

This table shows which prefixes and suffixes most frequently co-occur on the same stems, revealing the 'stacking' rules of the language's morphology.

Prefix	Suffix	Frequency	Examples
-ꯂ	-ꯕ	27 words	ꯂꯝꯀꯣꯏꯕ, ꯂꯥꯢꯔꯦꯝꯕ
-ꯁ	-ꯡ	26 words	ꯁꯔꯨꯀꯁꯤꯡ, ꯁꯥꯆꯨꯡ
-ꯑ	-ꯗ	26 words	ꯑꯗꯨꯋꯥꯏꯗ, ꯑꯥꯏꯂꯦꯟꯗ
-ꯂ	-ꯡ	24 words	ꯂꯥꯡꯂꯤꯕꯁꯤꯡ, ꯂꯦꯝꯍꯧꯕꯁꯤꯡ
-ꯑ	-ꯟ	21 words	ꯑꯦꯁ꯭ꯇꯣꯟ, ꯑꯦꯗꯃꯟ
-ꯄ	-ꯡ	21 words	ꯄꯟꯅꯨꯡ, ꯄꯔꯦꯡ
-ꯃ	-ꯡ	19 words	ꯃꯆꯥꯅꯨꯄꯥꯁꯤꯡ, ꯃꯌꯨꯡ
-ꯑ	-ꯡ	19 words	ꯑꯦꯜꯕꯝꯁꯤꯡ, ꯑꯍꯂꯁꯤꯡ
-ꯑ	-ꯁ	19 words	ꯑꯥꯔꯀꯥꯟꯁꯥꯁ, ꯑꯣꯒꯁ
-ꯑ	-ꯛ	18 words	ꯑꯁꯨꯛ, ꯑꯗꯨꯈꯛꯇꯃꯛ

6.5 Recursive Morpheme Segmentation

Using Recursive Hierarchical Substitutability, we decompose complex words into their constituent morphemes. This approach handles nested affixes (e.g., prefix-prefix-root-suffix).

Word	Suggested Split	Confidence	Stem
ꯇꯤꯎꯁꯅꯒꯤꯗꯃꯛ	ꯇꯤꯎꯁꯅꯒꯤ-ꯗ-ꯃꯛ	7.5	ꯗ
ꯄꯨꯊꯣꯛꯄꯒꯤꯗꯃꯛ	ꯄꯨꯊꯣꯛꯄꯒꯤ-ꯗ-ꯃꯛ	7.5	ꯗ
ꯏꯊꯤꯌꯣꯄꯤꯌꯥꯟ	ꯏꯊꯤꯌꯣꯄꯤꯌꯥ-ꯟ	4.5	ꯏꯊꯤꯌꯣꯄꯤꯌꯥ
ꯄꯨꯟꯁꯤꯟꯈꯤꯕ	ꯄꯨꯟꯁꯤꯟꯈꯤ-ꯕ	4.5	ꯄꯨꯟꯁꯤꯟꯈꯤ
ꯚꯤꯗꯤꯑꯣꯁꯤꯡꯗ	ꯚꯤꯗꯤꯑꯣꯁꯤꯡ-ꯗ	4.5	ꯚꯤꯗꯤꯑꯣꯁꯤꯡ
ꯍꯦꯟꯒꯠꯂꯛꯂꯤꯕ	ꯍꯦꯟꯒꯠꯂꯛꯂꯤ-ꯕ	4.5	ꯍꯦꯟꯒꯠꯂꯛꯂꯤ
ꯐꯨꯡꯒꯥꯋꯥꯔꯤꯗ	ꯐꯨꯡꯒꯥꯋꯥꯔꯤ-ꯗ	4.5	ꯐꯨꯡꯒꯥꯋꯥꯔꯤ
ꯍꯣꯁ꯭ꯄꯤꯇꯥꯜꯗ	ꯍꯣꯁ꯭ꯄꯤꯇꯥꯜ-ꯗ	4.5	ꯍꯣꯁ꯭ꯄꯤꯇꯥꯜ
ꯄꯨꯔꯣꯍꯤꯠꯁꯤꯡꯅ	ꯄꯨꯔꯣꯍꯤꯠꯁꯤꯡ-ꯅ	4.5	ꯄꯨꯔꯣꯍꯤꯠꯁꯤꯡ
ꯌꯨꯅꯤꯚꯔꯁꯤꯇꯤꯗ	ꯌꯨꯅꯤꯚꯔꯁꯤꯇꯤ-ꯗ	4.5	ꯌꯨꯅꯤꯚꯔꯁꯤꯇꯤ
ꯅꯤꯟꯍꯨꯔꯁꯥꯒꯅ	ꯅꯤꯟꯍꯨꯔꯁꯥꯒ-ꯅ	4.5	ꯅꯤꯟꯍꯨꯔꯁꯥꯒ
ꯅꯨꯃꯤꯗꯥꯡꯋꯥꯏꯔꯝꯗ	ꯅꯨꯃꯤꯗꯥꯡꯋꯥꯏꯔꯝ-ꯗ	4.5	ꯅꯨꯃꯤꯗꯥꯡꯋꯥꯏꯔꯝ
relationships	relationship-s	4.5	relationship
ꯍꯣꯔꯥꯏꯖꯣꯟꯁ	ꯍ-ꯣꯔꯥꯏꯖꯣꯟ-ꯁ	3.0	ꯣꯔꯥꯏꯖꯣꯟ
ꯁꯃꯨꯗ꯭ꯔꯁꯤꯡꯗꯥ	ꯁ-ꯃ-ꯨꯗ꯭ꯔꯁꯤꯡꯗꯥ	3.0	ꯨꯗ꯭ꯔꯁꯤꯡꯗꯥ

6.6 Linguistic Interpretation

Automated Insight: The language Manipuri shows high morphological productivity. The subword models are significantly more efficient than word models, suggesting a rich system of affixation or compounding.

Note on Idiomaticity: The high Idiomaticity Gap suggests a large number of frequent multi-word expressions or formulaic sequences that are statistically distinct from their component parts.

---

7. Summary & Recommendations

Production Recommendations

Component	Recommended	Rationale
Tokenizer	64k BPE	Best compression (4.32x)
N-gram	2-gram	Lowest perplexity (1,226)
Markov	Context-4	Highest predictability (98.7%)
Embeddings	100d	Balanced semantic capture and isotropy

---

Appendix: Metrics Glossary & Interpretation Guide

This section provides definitions, intuitions, and guidance for interpreting the metrics used throughout this report.

Tokenizer Metrics

Compression Ratio

Definition: The ratio of characters to tokens (chars/token). Measures how efficiently the tokenizer represents text.

Intuition: Higher compression means fewer tokens needed to represent the same text, reducing sequence lengths for downstream models. A 3x compression means ~3 characters per token on average.

What to seek: Higher is generally better for efficiency, but extremely high compression may indicate overly aggressive merging that loses morphological information.

Average Token Length (Fertility)

Definition: Mean number of characters per token produced by the tokenizer.

Intuition: Reflects the granularity of tokenization. Longer tokens capture more context but may struggle with rare words; shorter tokens are more flexible but increase sequence length.

What to seek: Balance between 2-5 characters for most languages. Arabic/morphologically-rich languages may benefit from slightly longer tokens.

Unknown Token Rate (OOV Rate)

Definition: Percentage of tokens that map to the unknown/UNK token, indicating words the tokenizer cannot represent.

Intuition: Lower OOV means better vocabulary coverage. High OOV indicates the tokenizer encounters many unseen character sequences.

What to seek: Below 1% is excellent; below 5% is acceptable. BPE tokenizers typically achieve very low OOV due to subword fallback.

N-gram Model Metrics

Perplexity

Definition: Measures how "surprised" the model is by test data. Mathematically: 2^(cross-entropy). Lower values indicate better prediction.

Intuition: If perplexity is 100, the model is as uncertain as if choosing uniformly among 100 options at each step. A perplexity of 10 means effectively choosing among 10 equally likely options.

What to seek: Lower is better. Perplexity decreases with larger n-grams (more context). Values vary widely by language and corpus size.

Entropy

Definition: Average information content (in bits) needed to encode the next token given the context. Related to perplexity: perplexity = 2^entropy.

Intuition: High entropy means high uncertainty/randomness; low entropy means predictable patterns. Natural language typically has entropy between 1-4 bits per character.

What to seek: Lower entropy indicates more predictable text patterns. Entropy should decrease as n-gram size increases.

Coverage (Top-K)

Definition: Percentage of corpus occurrences explained by the top K most frequent n-grams.

Intuition: High coverage with few patterns indicates repetitive/formulaic text; low coverage suggests diverse vocabulary usage.

What to seek: Depends on use case. For language modeling, moderate coverage (40-60% with top-1000) is typical for natural text.

Markov Chain Metrics

Average Entropy

Definition: Mean entropy across all contexts, measuring average uncertainty in next-word prediction.

Intuition: Lower entropy means the model is more confident about what comes next. Context-1 has high entropy (many possible next words); Context-4 has low entropy (few likely continuations).

What to seek: Decreasing entropy with larger context sizes. Very low entropy (<0.1) indicates highly deterministic transitions.

Branching Factor

Definition: Average number of unique next tokens observed for each context.

Intuition: High branching = many possible continuations (flexible but uncertain); low branching = few options (predictable but potentially repetitive).

What to seek: Branching factor should decrease with context size. Values near 1.0 indicate nearly deterministic chains.

Predictability

Definition: Derived metric: (1 - normalized_entropy) × 100%. Indicates how deterministic the model's predictions are.

Intuition: 100% predictability means the next word is always certain; 0% means completely random. Real text falls between these extremes.

What to seek: Higher predictability for text generation quality, but too high (>98%) may produce repetitive output.

Vocabulary & Zipf's Law Metrics

Zipf's Coefficient

Definition: The slope of the log-log plot of word frequency vs. rank. Zipf's law predicts this should be approximately -1.

Intuition: A coefficient near -1 indicates the corpus follows natural language patterns where a few words are very common and most words are rare.

What to seek: Values between -0.8 and -1.2 indicate healthy natural language distribution. Deviations may suggest domain-specific or artificial text.

R² (Coefficient of Determination)

Definition: Measures how well the linear fit explains the frequency-rank relationship. Ranges from 0 to 1.

Intuition: R² near 1.0 means the data closely follows Zipf's law; lower values indicate deviation from expected word frequency patterns.

What to seek: R² > 0.95 is excellent; > 0.99 indicates near-perfect Zipf adherence typical of large natural corpora.

Vocabulary Coverage

Definition: Cumulative percentage of corpus tokens accounted for by the top N words.

Intuition: Shows how concentrated word usage is. If top-100 words cover 50% of text, the corpus relies heavily on common words.

What to seek: Top-100 covering 30-50% is typical. Higher coverage indicates more repetitive text; lower suggests richer vocabulary.

Word Embedding Metrics

Isotropy

Definition: Measures how uniformly distributed vectors are in the embedding space. Computed as the ratio of minimum to maximum singular values.

Intuition: High isotropy (near 1.0) means vectors spread evenly in all directions; low isotropy means vectors cluster in certain directions, reducing expressiveness.

What to seek: Higher isotropy generally indicates better-quality embeddings. Values > 0.1 are reasonable; > 0.3 is good. Lower-dimensional embeddings tend to have higher isotropy.

Average Norm

Definition: Mean magnitude (L2 norm) of word vectors in the embedding space.

Intuition: Indicates the typical "length" of vectors. Consistent norms suggest stable training; high variance may indicate some words are undertrained.

What to seek: Relatively consistent norms across models. The absolute value matters less than consistency (low std deviation).

Cosine Similarity

Definition: Measures angular similarity between vectors, ranging from -1 (opposite) to 1 (identical direction).

Intuition: Words with similar meanings should have high cosine similarity. This is the standard metric for semantic relatedness in embeddings.

What to seek: Semantically related words should score > 0.5; unrelated words should be near 0. Synonyms often score > 0.7.

t-SNE Visualization

Definition: t-Distributed Stochastic Neighbor Embedding - a dimensionality reduction technique that preserves local structure for visualization.

Intuition: Clusters in t-SNE plots indicate groups of semantically related words. Spread indicates vocabulary diversity; tight clusters suggest semantic coherence.

What to seek: Meaningful clusters (e.g., numbers together, verbs together). Avoid over-interpreting distances - t-SNE preserves local, not global, structure.

General Interpretation Guidelines

1. Compare within model families: Metrics are most meaningful when comparing models of the same type (e.g., 8k vs 64k tokenizer).
2. Consider trade-offs: Better performance on one metric often comes at the cost of another (e.g., compression vs. OOV rate).
3. Context matters: Optimal values depend on downstream tasks. Text generation may prioritize different metrics than classification.
4. Corpus influence: All metrics are influenced by corpus characteristics. Wikipedia text differs from social media or literature.
5. Language-specific patterns: Morphologically rich languages (like Arabic) may show different optimal ranges than analytic languages.

Visualizations Index

Visualization	Description
Tokenizer Compression	Compression ratios by vocabulary size
Tokenizer Fertility	Average token length by vocabulary
Tokenizer OOV	Unknown token rates
Tokenizer Total Tokens	Total tokens by vocabulary
N-gram Perplexity	Perplexity by n-gram size
N-gram Entropy	Entropy by n-gram size
N-gram Coverage	Top pattern coverage
N-gram Unique	Unique n-gram counts
Markov Entropy	Entropy by context size
Markov Branching	Branching factor by context
Markov Contexts	Unique context counts
Zipf's Law	Frequency-rank distribution with fit
Vocab Frequency	Word frequency distribution
Top 20 Words	Most frequent words
Vocab Coverage	Cumulative coverage curve
Embedding Isotropy	Vector space uniformity
Embedding Norms	Vector magnitude distribution
Embedding Similarity	Word similarity heatmap
Nearest Neighbors	Similar words for key terms
t-SNE Words	2D word embedding visualization
t-SNE Sentences	2D sentence embedding visualization
Position Encoding	Encoding method comparison
Model Sizes	Storage requirements
Performance Dashboard	Comprehensive performance overview

---

About This Project

Data Source

Models trained on wikipedia-monthly - a monthly snapshot of Wikipedia articles across 300+ languages.

Project

A project by Wikilangs - Open-source NLP models for every Wikipedia language.

Maintainer

Omar Kamali - Omneity Labs

Citation

If you use these models in your research, please cite:

@misc{wikilangs2025,
  author = {Kamali, Omar},
  title = {Wikilangs: Open NLP Models for Wikipedia Languages},
  year = {2025},
  doi = {10.5281/zenodo.18073153},
  publisher = {Zenodo},
  url = {https://huggingface.co/wikilangs}
  institution = {Omneity Labs}
}

License

MIT License - Free for academic and commercial use.

Links

• 🌐 Website: wikilangs.org
• 🤗 Models: huggingface.co/wikilangs
• 📊 Data: wikipedia-monthly
• 👤 Author: Omar Kamali
• 🤝 Sponsor: Featherless AI

---

Generated by Wikilangs Models Pipeline

Report Date: 2026-01-10 12:16:30