wikilangs/blk

Pa'o Karen - Wikilangs Models

Comprehensive Research Report & Full Ablation Study

This repository contains NLP models trained and evaluated by Wikilangs, specifically on Pa'o Karen 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	4.022x	4.02	0.0580%	1,056,850
16k	4.430x	4.43	0.0639%	959,541
32k	4.613x	4.61	0.0665%	921,415
64k	4.848x 🏆	4.85	0.0699%	876,870

Tokenization Examples

Below are sample sentences tokenized with each vocabulary size:

Sample 1: မျန်မာခမ်းထီကိုယို တွိုင်ꩻဒေႏသတန် အဝ်ႏ ( ၇ )တွိုင်ꩻ နဝ်ꩻသွူ ။

Vocab	Tokens	Count
8k	▁မျန်မာခမ်းထီ ကိုယို ▁တွိုင်ꩻဒေႏသတန် ▁အဝ်ႏ ▁( ▁၇ ▁) တွိုင်ꩻ ▁နဝ်ꩻ သွူ ... (+1 more)	11
16k	▁မျန်မာခမ်းထီ ကိုယို ▁တွိုင်ꩻဒေႏသတန် ▁အဝ်ႏ ▁( ▁၇ ▁) တွိုင်ꩻ ▁နဝ်ꩻသွူ ▁။	10
32k	▁မျန်မာခမ်းထီ ကိုယို ▁တွိုင်ꩻဒေႏသတန် ▁အဝ်ႏ ▁( ▁၇ ▁) တွိုင်ꩻ ▁နဝ်ꩻသွူ ▁။	10
64k	▁မျန်မာခမ်းထီ ကိုယို ▁တွိုင်ꩻဒေႏသတန် ▁အဝ်ႏ ▁( ▁၇ ▁) တွိုင်ꩻ ▁နဝ်ꩻသွူ ▁။	10

Sample 2: ဝေင်ꩻနောင်ꩻတရားယိုနဝ်ꩻ အဝ်ႏဒျာႏ မျန်မာခမ်းထီ ၊ ဖြဝ်ꩻခမ်းနယ်ႏအခဝ်နဝ်၊ တောင်ႏကီꩻခရ...

Vocab	Tokens	Count
8k	▁ဝေင်ꩻန ောင်ꩻ တရား ယိုနဝ်ꩻ ▁အဝ်ႏဒျာႏ ▁မျန်မာခမ်းထီ ▁၊ ▁ဖြဝ်ꩻခမ်းနယ်ႏ အခဝ်နဝ်၊ ▁တောင်ႏကီꩻခရဲင်ႏ ... (+8 more)	18
16k	▁ဝေင်ꩻန ောင်ꩻ တရား ယိုနဝ်ꩻ ▁အဝ်ႏဒျာႏ ▁မျန်မာခမ်းထီ ▁၊ ▁ဖြဝ်ꩻခမ်းနယ်ႏ အခဝ်နဝ်၊ ▁တောင်ႏကီꩻခရဲင်ႏ ... (+8 more)	18
32k	▁ဝေင်ꩻန ောင်ꩻ တရား ယိုနဝ်ꩻ ▁အဝ်ႏဒျာႏ ▁မျန်မာခမ်းထီ ▁၊ ▁ဖြဝ်ꩻခမ်းနယ်ႏ အခဝ်နဝ်၊ ▁တောင်ႏကီꩻခရဲင်ႏ ... (+8 more)	18
64k	▁ဝေင်ꩻနောင်ꩻ တရားယိုနဝ်ꩻ ▁အဝ်ႏဒျာႏ ▁မျန်မာခမ်းထီ ▁၊ ▁ဖြဝ်ꩻခမ်းနယ်ႏ အခဝ်နဝ်၊ ▁တောင်ႏကီꩻခရဲင်ႏ ▁၊ ▁ဝေင်ꩻနယ်ႏပ ... (+6 more)	16

Sample 3: အမုဲင် ခမ်းထီ ကသှိုပ်စဒါႏ ငဝ်းလဝ်းနီꩻ ၃၅လာအို ၉၄ ထူႏတောမ်

Vocab	Tokens	Count
8k	▁အမုဲင် ▁ခမ်းထီ ▁က သ ှို ပ် စဒါႏ ▁ငဝ်း လ ဝ်း ... (+6 more)	16
16k	▁အမုဲင် ▁ခမ်းထီ ▁ကသ ှိုပ် စဒါႏ ▁ငဝ်း လဝ်း နီꩻ ▁၃၅ လာအို ... (+3 more)	13
32k	▁အမုဲင် ▁ခမ်းထီ ▁ကသှိုပ်စဒါႏ ▁ငဝ်းလဝ်းနီꩻ ▁၃၅လာအို ▁၉ ၄ ▁ထူႏတောမ်	8
64k	▁အမုဲင် ▁ခမ်းထီ ▁ကသှိုပ်စဒါႏ ▁ငဝ်းလဝ်းနီꩻ ▁၃၅လာအို ▁၉၄ ▁ထူႏတောမ်	7

Key Findings

• Best Compression: 64k achieves 4.848x compression
• Lowest UNK Rate: 8k with 0.0580% 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	2,539	11.31	4,306	21.2%	57.9%
2-gram	Subword	1,398 🏆	10.45	24,285	42.8%	77.0%
3-gram	Word	3,862	11.92	6,537	18.8%	47.3%
3-gram	Subword	11,299	13.46	129,572	19.0%	45.1%
4-gram	Word	16,871	14.04	23,296	9.0%	22.0%
4-gram	Subword	54,089	15.72	405,489	10.1%	25.8%
5-gram	Word	15,317	13.90	19,946	8.7%	21.0%
5-gram	Subword	138,288	17.08	617,898	5.8%	16.6%

Top 5 N-grams by Size

2-grams (Word):

Rank	N-gram	Count
1	နဝ်ꩻ အဝ်ႏဒျာႏ	719
2	အဝ်ႏဒျာႏ မျန်မာခမ်းထီ	691
3	ခရိစ်နေင်ႏ ဗာႏ	403
4	ဗာႏ စာႏရင်ꩻအလꩻ	320
5	မျန်မာခမ်းထီ အခဝ်ထာႏဝ	295

3-grams (Word):

Rank	N-gram	Count
1	နဝ်ꩻ အဝ်ႏဒျာႏ မျန်မာခမ်းထီ	624
2	အဝ်ႏဒျာႏ မျန်မာခမ်းထီ အခဝ်ထာႏဝ	295
3	ခရိစ်နေင်ႏ ဗာႏ စာႏရင်ꩻအလꩻ	261
4	ဗာႏ စာႏရင်ꩻအလꩻ ဝေင်ꩻကိုနဝ်ꩻ	161
5	ထာꩻထွာဖုံႏ လွူးဖွာꩻသားဖုံႏ သီမားသားဖုံႏ	153

4-grams (Word):

Rank	N-gram	Count
1	နဝ်ꩻ အဝ်ႏဒျာႏ မျန်မာခမ်းထီ အခဝ်ထာႏဝ	282
2	ခရိစ်နေင်ႏ ဗာႏ စာႏရင်ꩻအလꩻ ဝေင်ꩻကိုနဝ်ꩻ	161
3	လွူးဖွာꩻသားဖုံႏ သီမားသားဖုံႏ မွူးနီꩻအုံပဆားနီꩻဖုံႏတောမ်ႏ အထွတ်အမျတ်မွူးနီꩻဖုံႏ	153
4	သီမားသားဖုံႏ မွူးနီꩻအုံပဆားနီꩻဖုံႏတောမ်ႏ အထွတ်အမျတ်မွူးနီꩻဖုံႏ အာႏကွိုꩻ	153
5	ထာꩻထွာဖုံႏ လွူးဖွာꩻသားဖုံႏ သီမားသားဖုံႏ မွူးနီꩻအုံပဆားနီꩻဖုံႏတောမ်ႏ	153

5-grams (Word):

Rank	N-gram	Count
1	လွူးဖွာꩻသားဖုံႏ သီမားသားဖုံႏ မွူးနီꩻအုံပဆားနီꩻဖုံႏတောမ်ႏ အထွတ်အမျတ်မွူးနီꩻဖုံႏ အာႏကွိုꩻ	153
2	ထာꩻထွာဖုံႏ လွူးဖွာꩻသားဖုံႏ သီမားသားဖုံႏ မွူးနီꩻအုံပဆားနီꩻဖုံႏတောမ်ႏ အထွတ်အမျတ်မွူးနီꩻဖုံႏ	153
3	သွူ ထာꩻထွာဖုံႏ လွူးဖွာꩻသားဖုံႏ သီမားသားဖုံႏ မွူးနီꩻအုံပဆားနီꩻဖုံႏတောမ်ႏ	151
4	ခရိစ်နေင်ႏ ဗာႏ စာႏရင်ꩻအလꩻ ဝေင်ꩻကိုနဝ်ꩻ လိုꩻဖြာꩻခြွဉ်းအဝ်ႏ	131
5	အဝ်ႏသော့ꩻနဝ်ꩻသွူ ခရိစ်နေင်ႏ ဗာႏ စာႏရင်ꩻအလꩻ ဝေင်ꩻကိုနဝ်ꩻ	111

2-grams (Subword):

Rank	N-gram	Count
1	ာ ႏ	142,384
2	၊ _	135,380
3	ꩻ _	126,353
4	ဝ် ꩻ	102,695
5	င် ꩻ	96,805

3-grams (Subword):

Rank	N-gram	Count
1	န ဝ် ꩻ	77,014
2	ဝ် ꩻ _	57,567
3	ꩻ ၊ _	31,811
4	သွူ ။ _	31,570
5	ႏ ၊ _	30,928

4-grams (Subword):

Rank	N-gram	Count
1	န ဝ် ꩻ _	45,450
2	နေ ာ ဝ် ꩻ	23,553
3	ꩻ သွူ ။ _	18,993
4	ꩻ န ဝ် ꩻ	18,023
5	ႏ န ဝ် ꩻ	17,057

5-grams (Subword):

Rank	N-gram	Count
1	ဝ် ꩻ သွူ ။ _	15,761
2	ꩻ န ဝ် ꩻ _	12,522
3	နေ ာ ဝ် ꩻ _	11,865
4	ႏ န ဝ် ꩻ _	10,503
5	န ဝ် ꩻ သွူ ။	10,311

Key Findings

• Best Perplexity: 2-gram (subword) with 1,398
• Entropy Trend: Decreases with larger n-grams (more predictable)
• Coverage: Top-1000 patterns cover ~17% 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	0.2308	1.173	1.60	381,069	76.9%
1	Subword	1.2202	2.330	20.98	2,909	0.0%
2	Word	0.0412	1.029	1.06	609,269	95.9%
2	Subword	0.7534	1.686	5.49	61,020	24.7%
3	Word	0.0155	1.011	1.02	645,305	98.5%
3	Subword	0.4733	1.388	2.77	335,231	52.7%
4	Word	0.0088 🏆	1.006	1.01	656,933	99.1%
4	Subword	0.3156	1.245	1.90	930,014	68.4%

Generated Text Samples (Word-based)

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

Context Size 1:

1. ၂ ဖြုံႏလဲ့ အဝ်ႏသွူ ခမ်းတွူးကောင်ꩻယို အမိဉ်ꩻနဝ်ꩻ ဖန်ဖေႏ စဲဉ်ႏဖေႏဒျာႏလွဉ်းလွဉ်းသွူ ယိုလွုမ်ꩻမကာႏ ဗွေႏဗ...
2. ၃ ပွုမ်ႏယိုသွူ က အဟံ ခွေနဝ်ꩻ ကောလက္ခံႏသား ၂ ၃ ပေါႏပါႏဠိဒျာႏနဝ်ꩻ သော့ꩻတောဝ်းအမုဲင် ဟော်ꩻဖတ်ဗော့ꩻ ပါႏဠ...
3. ၁ ခြပ် စီ သွံဆီသူ တနတ်တလီꩻ air combat information management unit mimu ဝေင်ꩻနယ်ႏရွုမ်ꩻဖုံႏနဝ်ꩻ အဝ်ႏဒ...

Context Size 2:

1. နဝ်ꩻ အဝ်ႏဒျာႏ မျန်မာခမ်းထီ ဖြဝ်ꩻခမ်းနယ်ႏအခဝ်ကွဉ်ႏ မွိုင်ꩻတုံခရဲင်ႏ ဝေင်ꩻနယ်ႏမွိုင်ꩻတုံကို ကပါဒါႏ ဝေင...
2. အဝ်ႏဒျာႏ မျန်မာခမ်းထီ အခဝ်ထာႏဝ ပဂိုꩻတွိုင်ꩻဒေႏသတန် အခဝ်ကွဉ်ႏထင်ꩻ တောင်ႏအူခရဲင်ႏ ဝေင်ꩻနယ်ႏဖျူးကို ကပါ...
3. ခရိစ်နေင်ႏ ဗာႏ စဲ့ꩻအစိုႏရစိုးကို ကဗွောင်လွေꩻဒါႏ ခမ်းလင်လစ်ꩻခမ်းတောမ်ႏ ဖြေꩻစာကွန်ႏ လွယ်စယ်ခမ်းကူဂဲတ်လ...

Context Size 3:

1. နဝ်ꩻ အဝ်ႏဒျာႏ မျန်မာခမ်းထီ အခဝ်ကွဉ်ႏထင်ꩻ ဖြဝ်ꩻခမ်းနယ်ႏ အခဝ်ထင်ꩻ ဟိုပန်ခရဲင်ႏ ဝနမ်းပဲင်ႏအိုပ်ချုတ်ခွင...
2. အဝ်ႏဒျာႏ မျန်မာခမ်းထီ အခဝ်ထာႏဝ နေပီဒေါ်ခမ်းခြွဉ်းဗူႏဟံႏနယ်ႏ လယ်ဝွေးခရဲင်ႏကို ကအဝ်ႏပါသော့ꩻဒါႏ ဝေင်ꩻနယ...
3. ခရိစ်နေင်ႏ ဗာႏ စာႏရင်ꩻအလꩻ ဝေင်ꩻကိုနဝ်ꩻ လိုꩻဖြာꩻအဝ်ႏ ၁၁ ၃၀၅ ဖြာꩻသွူ အဝ်ႏဒျာႏ ထာဝယ် မေက် ကာꩻတဖူꩻတန်လော...

Context Size 4:

1. နဝ်ꩻ အဝ်ႏဒျာႏ မျန်မာခမ်းထီ အခဝ်ထာႏဝ ဧရာႏဝတီႏတွိုင်ꩻဒေႏသတန် မအူပိဉ်ခရဲင်ႏ ကို ကပါဒါႏ ဝေင်ꩻနယ်ႏတဖြုံႏဒ...
2. ခရိစ်နေင်ႏ ဗာႏ စာႏရင်ꩻအလꩻ ဝေင်ꩻကိုနဝ်ꩻ လိုꩻဖြာꩻအဝ်ႏ ၁၀ ၄၄၃ ဖြာꩻသွူ အဉ်းမယို ကရီးခါနဝ်ꩻ ထွာဒျာႏ ဒုံအဉ...
3. ထာꩻထွာဖုံႏ လွူးဖွာꩻသားဖုံႏ သီမားသားဖုံႏ မွူးနီꩻအုံပဆားနီꩻဖုံႏတောမ်ႏ အထွတ်အမျတ်မွူးနီꩻဖုံႏ အာႏကွိုꩻ

Generated Text Samples (Subword-based)

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

Context Size 1:

1. _တသီႏပေႏစမုံးဝါꩻစွဉ်းထဲ
2. ꩻနဝ်ႏပုဂ္ဂိုလ်ႏ_ဇာဝ်ꩻသား
3. ႏအခြာဏ်ႏတောယိုတဲ့_ဟောႏရာ

Context Size 2:

1. ာႏနဝ်ꩻ_အောဝ်ꩻသွူကျောင်ႏဒျာ
2. ၊_တွိုက်_ကြွဲႏ_ဖန်_သွူ။_ဓမ္မပ
3. ꩻ_ခင်ႏငံႏ_မန်း"ကို_ကတဲမ်

Context Size 3:

1. နဝ်ꩻသွူ။_နီကွဉ်ကꩻမွိုန်း။_၂။
2. ဝ်ꩻ_အံႏဖြာꩻနောဝ်ꩻ_ပအိုဝ်ႏယို
3. ꩻ၊_မဲ့သျင်ႏကျင်ꩻ။_နီလိတ်_အဝ်

Context Size 4:

1. နဝ်ꩻ_ဟဲ့ꩻဗာႏသꩻ_ကွား_ကွန်ပေ
2. နောဝ်ꩻ_ဘဝပေါင်ꩻ_ရွဉ်ခန်ဗီႏ_
3. ꩻသွူ။_ပအိုဝ်ႏစွိုးခွိုꩻသီး_သွိုန်ႏသ

Key Findings

• Best Predictability: Context-4 (word) with 99.1% predictability
• Branching Factor: Decreases with context size (more deterministic)
• Memory Trade-off: Larger contexts require more storage (930,014 contexts)
• Recommendation: Context-3 or Context-4 for text generation

---

4. Vocabulary Analysis

Statistics

Metric	Value
Vocabulary Size	67,819
Total Tokens	396,228
Mean Frequency	5.84
Median Frequency	2
Frequency Std Dev	39.85

Most Common Words

Rank	Word	Frequency
1	၂	3,796
2	၃	3,380
3	၁	3,330
4	အာႏကွိုꩻ	3,141
5	နဝ်ꩻ	2,717
6	၄	2,608
7	၅	2,058
8	ထွာဒျာႏ	1,623
9	၆	1,585
10	အဝ်ႏဒျာႏ	1,494

Least Common Words (from vocabulary)

Rank	Word	Frequency
1	တထာနမ်းနောဝ်ꩻ	2
2	တထာဖြွီꩻဖုံႏ	2
3	antihistamine	2
4	ပထမခွိုꩻ	2
5	ဒုတိယခွိုꩻ	2
6	histamine	2
7	တနယ်ႏလိုမ်းဆဲင်ႏရာꩻ	2
8	အခြေပြုမူလတန်ꩻ	2
9	ပထမကြီးတန်ꩻတွမ်ႏ	2
10	ရန်ႏကုန်ႏတုံး	2

Zipf's Law Analysis

Metric	Value
Zipf Coefficient	0.7916
R² (Goodness of Fit)	0.998007
Adherence Quality	excellent

Coverage Analysis

Top N Words	Coverage
Top 100	17.9%
Top 1,000	34.4%
Top 5,000	51.9%
Top 10,000	61.5%

Key Findings

• Zipf Compliance: R²=0.9980 indicates excellent adherence to Zipf's law
• High Frequency Dominance: Top 100 words cover 17.9% of corpus
• Long Tail: 57,819 words needed for remaining 38.5% 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	0.8632 🏆	0.3270	N/A	N/A
mono_64d	64	0.8595	0.2722	N/A	N/A
mono_128d	128	0.6854	0.2261	N/A	N/A
aligned_32d	32	0.8632	0.3317	0.0135	0.1716
aligned_64d	64	0.8595	0.2717	0.0745	0.2844
aligned_128d	128	0.6854	0.2281	0.1625	0.3386

Key Findings

• Best Isotropy: mono_32d with 0.8632 (more uniform distribution)
• Semantic Density: Average pairwise similarity of 0.2762. Lower values indicate better semantic separation.
• Alignment Quality: Aligned models achieve up to 16.3% 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	0.267	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
-လိ	လိက်ပအိုဝ်ႏ, လိုꩻစားယိုဖုံႏနဝ်ꩻ, လိတ်လုံးကို
-လို	လိုꩻစားယိုဖုံႏနဝ်ꩻ, လိုꩻခိုဖုံႏလဲ့, လိုꩻမွိုက်နဝ်ꩻ

Productive Suffixes

Suffix	Examples
-ꩻ	ဝန်ႏကိုတွော့ꩻ, ရခဲင်ႏခွန်ဟော်ခံꩻ, သဗ္ဗညုဘုရာꩻ
-ႏ	လဲဉ်အံႏ, အလင်္ကာႏ, မာꩻသွော့ကုသိုလ်ႏ
-်ꩻ	လာအိုခမ်းထီနဝ်ꩻ, တသီႏအံႏနယ်ꩻနဝ်ꩻ, ယိုသွံပါꩻထွာနဝ်ꩻ
-း	အောဝ်ႏဟမ်ႏခမ်ႏဖာႏလောင်း, လူထုအလောင်း, ဉာဏ်ႏတောႏဆꩻချာလွဉ်းလွဉ်း
-ဝ်ꩻ	လာအိုခမ်းထီနဝ်ꩻ, တသီႏအံႏနယ်ꩻနဝ်ꩻ, ယိုသွံပါꩻထွာနဝ်ꩻ
-်း	အောဝ်ႏဟမ်ႏခမ်ႏဖာႏလောင်း, လူထုအလောင်း, ဉာဏ်ႏတောႏဆꩻချာလွဉ်းလွဉ်း
-နဝ်ꩻ	လာအိုခမ်းထီနဝ်ꩻ, တသီႏအံႏနယ်ꩻနဝ်ꩻ, ယိုသွံပါꩻထွာနဝ်ꩻ
-ာႏ	အလင်္ကာႏ, ဖန်ဆင်ꩻမာꩻခါꩻဒျာႏ, ကိုꩻကွယ်ႏသားအာဗာႏ

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.

No significant bound stems detected.

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
-လိ	-ꩻ	83 words	လိုꩻမဉ်ꩻ, လိုꩻနမ်းအကိုအထန်ႏနီဖဲ့ꩻ
-လိ	-ႏ	64 words	လိုꩻစွဲဉ်ႏ, လိုꩻမုရေꩻအစွိုꩻအဗူႏဖုံႏ
-လိ	-်ꩻ	61 words	လိုꩻမဉ်ꩻ, လိုꩻယုက်နဝ်ꩻ
-လိ	-ဝ်ꩻ	45 words	လိုꩻယုက်နဝ်ꩻ, လိတ်မွူးပအိုဝ်ႏယိုခါနဝ်ꩻ
-လိ	-နဝ်ꩻ	37 words	လိုꩻယုက်နဝ်ꩻ, လိတ်မွူးပအိုဝ်ႏယိုခါနဝ်ꩻ
-လိ	-း	36 words	လိုꩻခမ်း, လိုႏတဝ်း
-လိ	-်ႏ	23 words	လိုꩻစွဲဉ်ႏ, လိုꩻသွုန်ႏထီဓာတ်တွမ်ႏ
-လိ	-်း	19 words	လိုꩻခမ်း, လိုႏတဝ်း
-လိ	-ာႏ	15 words	လိုꩻမျိုꩻတွမ်ႏခမ်းထီအတာႏ, လိတ်လုဲင်ꩻတွမ်ႏအနုပညာႏ
-လိ	-ွူ	5 words	လိုꩻမဉ်အံႏနွောင်ꩻနိစ်စက်ဒါႏဝင်ꩻဖုံႏနဝ်ꩻသွူ, လိုႏမာꩻထူႏလွလဲဉ်းဒျာႏနောဝ်ꩻသွူ

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
ကွဲညညနဝ်ꩻ	ကွဲညည-နဝ်ꩻ	4.5	ကွဲညည
သꩻကိုနဝ်ꩻ	သꩻကို-နဝ်ꩻ	4.5	သꩻကို
လိုꩻယင်ဟန်ႏနဝ်ꩻ	လို-ꩻယင်ဟန-်ႏ-နဝ်ꩻ	4.5	ꩻယင်ဟန
နင်ꩻသုမနာနဝ်ꩻ	နင်ꩻသုမနာ-နဝ်ꩻ	4.5	နင်ꩻသုမနာ
ပုဏ္ဏာꩻနဝ်ꩻ	ပုဏ္ဏာꩻ-နဝ်ꩻ	4.5	ပုဏ္ဏာꩻ
နာꩻတဲ့နဝ်ꩻ	နာꩻတဲ့-နဝ်ꩻ	4.5	နာꩻတဲ့
ခန္ဓာႏတန်ယိုနဝ်ꩻ	ခန္ဓာႏတန်ယို-နဝ်ꩻ	4.5	ခန္ဓာႏတန်ယို
ရောင်ထာꩻနဝ်ꩻ	ရောင်ထာꩻ-နဝ်ꩻ	4.5	ရောင်ထာꩻ
ခယ်ႏမူႏနဝ်ꩻ	ခယ်ႏမူႏ-နဝ်ꩻ	4.5	ခယ်ႏမူႏ
အနာႏဂတ်နဝ်ꩻ	အနာႏဂတ်-နဝ်ꩻ	4.5	အနာႏဂတ်
ရဟန်ꩻသာႏမဏေႏနဝ်ꩻ	ရဟန်ꩻသာႏမဏေႏ-နဝ်ꩻ	4.5	ရဟန်ꩻသာႏမဏေႏ
ထွို့ꩻစွဲႏနဝ်ꩻ	ထွို့ꩻစွဲႏ-နဝ်ꩻ	4.5	ထွို့ꩻစွဲႏ
စူမွူးနဝ်ꩻ	စူမွူး-နဝ်ꩻ	4.5	စူမွူး
ပွိုးနဝ်ꩻ	ပွိုး-နဝ်ꩻ	4.5	ပွိုး
သင်္ဃာႏတောႏနဝ်ꩻ	သင်္ဃာႏတေ-ာႏ-နဝ်ꩻ	3.0	သင်္ဃာႏတေ

6.6 Linguistic Interpretation

Automated Insight: The language Pa'o Karen 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 (4.85x)
N-gram	2-gram	Lowest perplexity (1,398)
Markov	Context-4	Highest predictability (99.1%)
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: 2026-01-03 19:13:44