Gothic - Wikilangs Models

Comprehensive Research Report & Full Ablation Study

This repository contains NLP models trained and evaluated by Wikilangs, specifically on Gothic 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	2.525x	2.53	0.0669%	260,190
16k	2.674x	2.68	0.0708%	245,725
32k	2.884x 🏆	2.89	0.0764%	227,819

Tokenization Examples

Below are sample sentences tokenized with each vocabulary size:

Sample 1: 𐌺𐌰𐌽𐌰𐌳𐌰 𐌹𐍃𐍄 𐌻𐌰𐌽𐌳 𐌰𐌽𐌰 𐌰𐌹𐍂𐌸𐌰𐌳𐌰𐌹𐌻𐌰𐌹 𐌽𐌰𐌿𐍂𐌸𐌰𐌼𐌰𐌹𐍂𐌹𐌺𐌰 𐌾𐌰𐌷 𐌲𐌰𐌼𐌰𐍂𐌺𐍉𐌸 𐌲𐌰𐌲𐌰𐌷𐌰𐍆𐍄𐌹𐌳𐌰 𐍂𐌴𐌹𐌺𐌾𐌰𐌹. ...

Vocab	Tokens	Count
8k	▁𐌺𐌰𐌽𐌰𐌳𐌰 ▁𐌹𐍃𐍄 ▁𐌻𐌰𐌽𐌳 ▁𐌰𐌽𐌰 ▁𐌰𐌹𐍂𐌸𐌰𐌳𐌰𐌹𐌻 𐌰𐌹 ▁𐌽𐌰𐌿𐍂𐌸 𐌰𐌼𐌰𐌹𐍂𐌹𐌺 𐌰 ▁𐌾𐌰𐌷 ... (+20 more)	30
16k	▁𐌺𐌰𐌽𐌰𐌳𐌰 ▁𐌹𐍃𐍄 ▁𐌻𐌰𐌽𐌳 ▁𐌰𐌽𐌰 ▁𐌰𐌹𐍂𐌸𐌰𐌳𐌰𐌹𐌻𐌰𐌹 ▁𐌽𐌰𐌿𐍂𐌸 𐌰𐌼𐌰𐌹𐍂𐌹𐌺𐌰 ▁𐌾𐌰𐌷 ▁𐌲𐌰𐌼𐌰𐍂𐌺𐍉𐌸 ▁𐌲𐌰𐌲𐌰𐌷𐌰𐍆𐍄𐌹𐌳𐌰 ... (+16 more)	26
32k	▁𐌺𐌰𐌽𐌰𐌳𐌰 ▁𐌹𐍃𐍄 ▁𐌻𐌰𐌽𐌳 ▁𐌰𐌽𐌰 ▁𐌰𐌹𐍂𐌸𐌰𐌳𐌰𐌹𐌻𐌰𐌹 ▁𐌽𐌰𐌿𐍂𐌸𐌰𐌼𐌰𐌹𐍂𐌹𐌺𐌰 ▁𐌾𐌰𐌷 ▁𐌲𐌰𐌼𐌰𐍂𐌺𐍉𐌸 ▁𐌲𐌰𐌲𐌰𐌷𐌰𐍆𐍄𐌹𐌳𐌰 ▁𐍂𐌴𐌹𐌺𐌾𐌰𐌹 ... (+12 more)	22

Sample 2: 𐌰𐍀𐌻𐍃 — 𐌰𐌺𐍂𐌰𐌽 𐌰𐍀𐌻𐌰𐌱𐌰𐌲𐌼𐌴 𐌾𐌰𐌷 𐍅𐌰𐌹𐌻𐌰𐌺𐌿𐌽𐌸𐌰 𐍆𐍉𐌳𐌴𐌹𐌽𐍃 𐌹𐍃𐍄·

Vocab	Tokens	Count
8k	▁𐌰𐍀𐌻𐍃 ▁— ▁𐌰𐌺𐍂𐌰𐌽 ▁𐌰𐍀 𐌻 𐌰𐌱𐌰𐌲𐌼𐌴 ▁𐌾𐌰𐌷 ▁𐍅𐌰𐌹𐌻 𐌰𐌺𐌿𐌽𐌸𐌰 ▁𐍆𐍉𐌳𐌴𐌹𐌽𐍃 ... (+2 more)	12
16k	▁𐌰𐍀𐌻𐍃 ▁— ▁𐌰𐌺𐍂𐌰𐌽 ▁𐌰𐍀 𐌻 𐌰𐌱𐌰𐌲𐌼𐌴 ▁𐌾𐌰𐌷 ▁𐍅𐌰𐌹𐌻 𐌰𐌺𐌿𐌽𐌸𐌰 ▁𐍆𐍉𐌳𐌴𐌹𐌽𐍃 ... (+2 more)	12
32k	▁𐌰𐍀𐌻𐍃 ▁— ▁𐌰𐌺𐍂𐌰𐌽 ▁𐌰𐍀𐌻𐌰𐌱𐌰𐌲𐌼𐌴 ▁𐌾𐌰𐌷 ▁𐍅𐌰𐌹𐌻𐌰𐌺𐌿𐌽𐌸𐌰 ▁𐍆𐍉𐌳𐌴𐌹𐌽𐍃 ▁𐌹𐍃𐍄 ·	9

Sample 3: 𐌺𐌰𐌿𐌻𐌿𐌼𐌱𐌾𐌰 (Colombia) 𐌹𐍃𐍄 𐌻𐌰𐌽𐌳 𐌹𐌽 𐍃𐌿𐌽𐌸𐍂𐌰𐌰𐌼𐌰𐌹𐍂𐌹𐌺𐌰𐌹. 𐌰𐌼𐌴𐍂𐌹𐌺𐌰 This page is brought t...

Vocab	Tokens	Count
8k	▁𐌺𐌰𐌿𐌻𐌿𐌼𐌱 𐌾𐌰 ▁( col om b ia ) ▁𐌹𐍃𐍄 ▁𐌻𐌰𐌽𐌳 ... (+19 more)	29
16k	▁𐌺𐌰𐌿𐌻𐌿𐌼𐌱𐌾𐌰 ▁( colombia ) ▁𐌹𐍃𐍄 ▁𐌻𐌰𐌽𐌳 ▁𐌹𐌽 ▁𐍃𐌿𐌽𐌸𐍂𐌰𐌰𐌼𐌰𐌹𐍂𐌹𐌺𐌰𐌹 . ▁𐌰𐌼𐌴𐍂𐌹𐌺𐌰 ... (+12 more)	22
32k	▁𐌺𐌰𐌿𐌻𐌿𐌼𐌱𐌾𐌰 ▁( colombia ) ▁𐌹𐍃𐍄 ▁𐌻𐌰𐌽𐌳 ▁𐌹𐌽 ▁𐍃𐌿𐌽𐌸𐍂𐌰𐌰𐌼𐌰𐌹𐍂𐌹𐌺𐌰𐌹 . ▁𐌰𐌼𐌴𐍂𐌹𐌺𐌰 ... (+10 more)	20

Key Findings

• Best Compression: 32k achieves 2.884x compression
• Lowest UNK Rate: 8k with 0.0669% 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	773	9.60	1,213	36.4%	92.9%
2-gram	Subword	546 🏆	9.09	2,316	47.1%	96.7%
3-gram	Word	630	9.30	1,041	40.1%	98.0%
3-gram	Subword	4,140	12.02	14,315	17.0%	56.1%
4-gram	Word	3,152	11.62	3,669	12.9%	38.4%
4-gram	Subword	17,609	14.10	51,785	8.9%	30.1%
5-gram	Word	2,230	11.12	2,508	13.1%	46.3%
5-gram	Subword	36,495	15.16	84,401	6.7%	21.6%

Top 5 N-grams by Size

2-grams (Word):

Rank	N-gram	Count
1	i to	325
2	wv i	315
3	akin to	129
4	iii to	106
5	𐌹𐌽 𐌰𐌼𐌰𐌹𐍂𐌹𐌺𐌰𐌹	102

3-grams (Word):

Rank	N-gram	Count
1	wv i to	276
2	akin to eng	78
3	sv vii to	64
4	sv iii to	61
5	𐌹𐍃𐍄 𐌻𐌰𐌽𐌳 𐌹𐌽	54

4-grams (Word):

Rank	N-gram	Count
1	𐌸𐍉𐌶𐌴𐌹 𐌰𐌻𐌻𐍉𐍃 𐍅𐌹𐌺𐌹𐍀𐌰𐌹𐌳𐌾𐍉𐍃 𐍃𐌺𐌿𐌻𐌿𐌽	48
2	𐌰𐌻𐌻𐍉𐍃 𐍅𐌹𐌺𐌹𐍀𐌰𐌹𐌳𐌾𐍉𐍃 𐍃𐌺𐌿𐌻𐌿𐌽 𐌷𐌰𐌱𐌰𐌽	48
3	𐍃𐌴𐌹𐌳𐍉 𐌸𐍉𐌶𐌴𐌹 𐌰𐌻𐌻𐍉𐍃 𐍅𐌹𐌺𐌹𐍀𐌰𐌹𐌳𐌾𐍉𐍃	48
4	𐌹𐌽 𐌰𐌼𐌰𐌹𐍂𐌹𐌺𐌰𐌹 𐌲𐌰𐍅𐌹𐍃𐍃𐌴𐌹𐍃 www	48
5	𐌹𐌽 𐌰𐌼𐌰𐌹𐍂𐌹𐌺𐌰𐌹 𐌷𐌰𐌿𐌱𐌹𐌳𐌰𐌱𐌰𐌿𐍂𐌲𐍃 𐌹𐍃𐍄	40

5-grams (Word):

Rank	N-gram	Count
1	𐍃𐌴𐌹𐌳𐍉 𐌸𐍉𐌶𐌴𐌹 𐌰𐌻𐌻𐍉𐍃 𐍅𐌹𐌺𐌹𐍀𐌰𐌹𐌳𐌾𐍉𐍃 𐍃𐌺𐌿𐌻𐌿𐌽	48
2	𐌸𐍉𐌶𐌴𐌹 𐌰𐌻𐌻𐍉𐍃 𐍅𐌹𐌺𐌹𐍀𐌰𐌹𐌳𐌾𐍉𐍃 𐍃𐌺𐌿𐌻𐌿𐌽 𐌷𐌰𐌱𐌰𐌽	48
3	𐌹𐍃𐍄 𐌲𐌰𐍅𐌹 𐌹𐌽 𐌰𐌼𐌰𐌹𐍂𐌹𐌺𐌰𐌹 𐌷𐌰𐌿𐌱𐌹𐌳𐌰𐌱𐌰𐌿𐍂𐌲𐍃	36
4	𐌲𐌰𐍅𐌹 𐌹𐌽 𐌰𐌼𐌰𐌹𐍂𐌹𐌺𐌰𐌹 𐌷𐌰𐌿𐌱𐌹𐌳𐌰𐌱𐌰𐌿𐍂𐌲𐍃 𐌹𐍃𐍄	36
5	𐌷𐌰𐌿𐌱𐌹𐌳𐌰𐌱𐌰𐌿𐍂𐌲𐍃 𐌾𐌰𐌷 𐍃𐍉 𐌼𐌰𐌹𐍃𐍄𐍉 𐌱𐌰𐌿𐍂𐌲𐍃	21

2-grams (Subword):

Rank	N-gram	Count
1	, _	17,634
2	. _	14,540
3	𐌰 𐌹	7,870
4	𐍃 _	7,637
5	𐌹 𐍃	6,470

3-grams (Subword):

Rank	N-gram	Count
1	_ - _	2,452
2	n , _	2,251
3	s , _	2,187
4	𐌹 𐌽 _	2,125
5	, _ s	2,064

4-grams (Subword):

Rank	N-gram	Count
1	_ 𐌹 𐌽 _	1,670
2	_ t o _	1,483
3	_ 𐌾 𐌰 𐌷	1,475
4	𐌾 𐌰 𐌷 _	1,472
5	a n , _	1,390

5-grams (Subword):

Rank	N-gram	Count
1	_ 𐌾 𐌰 𐌷 _	1,469
2	_ 𐌹 𐍃 𐍄 _	1,060
3	_ t h e _	885
4	, _ t o _	881
5	_ o e . _	839

Key Findings

• Best Perplexity: 2-gram (subword) with 546
• Entropy Trend: Decreases with larger n-grams (more predictable)
• Coverage: Top-1000 patterns cover ~22% 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	0.5463	1.460	2.78	26,779	45.4%
1	Subword	1.3185	2.494	9.24	600	0.0%
2	Word	0.1349	1.098	1.22	73,655	86.5%
2	Subword	0.9989	1.999	5.20	5,543	0.1%
3	Word	0.0401	1.028	1.06	89,056	96.0%
3	Subword	0.7885	1.727	3.23	28,771	21.2%
4	Word	0.0157 🏆	1.011	1.02	93,235	98.4%
4	Subword	0.5184	1.432	2.05	92,872	48.2%

Generated Text Samples (Word-based)

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

Context Size 1:

1. 𐌹𐌽 𐍅𐌹𐍃𐍄𐍂𐌰𐌹 𐌰𐍃𐌹𐌰𐌹 𐌽𐌴𐌷𐍅𐌿𐌽𐌳𐍉𐍃 𐌿𐍆𐌰𐍂 500 𐍆𐌰𐌿𐍂𐌰 𐍇𐍂𐌹𐍃𐍄𐌰𐌿 𐍃𐌰 𐌼𐌰𐌹𐍃𐍄𐌰 𐌰𐌻𐌻𐌰𐌹𐌶𐌴 𐌰𐌹𐍅𐌴 𐍃𐌴𐌹𐌳𐍉 𐌸𐍉𐌶𐌴𐌹 𐌵𐌹𐌼𐌰𐌽𐌳 𐍆𐍂𐌰𐌼
2. to tame 170 182 354 fulla ga náitjan wv i am trying to call cry aloud
3. 𐌾𐌰𐌷 𐌰𐌽𐌸𐌰𐍂𐌰𐌹𐌼 𐌱𐌰𐍂𐌱𐌰𐍂𐌹𐍅𐌴 𐌸𐌰𐌹𐌴𐌹 𐌺𐌿𐌽𐌽𐌰𐌽 𐍈𐌰 𐌹𐌽 𐌾𐌴𐍂𐌰 𐌿𐍃𐍅𐌰𐌹𐍂𐍀𐌰𐌽 𐌼𐌰𐌷𐍄𐌴𐌹𐌲 𐍅𐌰𐍃 𐌸𐌰𐍄𐌴𐌹 𐌰𐍂𐌰𐌱𐌹𐍃𐌺𐌰 𐍂𐌰𐌶𐌳𐌰 𐍂𐌰𐌶𐌳𐌰 𐌿𐌺𐍂𐌰...

Context Size 2:

1. i to lighten 424 ohg lohazzen láun sn pay reward 22 141 175 211 oe ht a
2. wv i see ga eitjan eits aj white 140 165 oe hwt ohg hw 329a an av
3. akin to eng ask treat shamefully oe ntan ohg neien ga nasjan wv i to permit allow

Context Size 3:

1. wv i to give light 63 85 105 320 oe lehtan liuhten liusan sv ii see af skiuban
2. akin to eng arrow arrow arjan distantly akin to lat anima spirit pant comp uzanan exhale and anda
3. sv vii to call to one profess confess acknowledge give thanks to and háusjan wv i to sin

Context Size 4:

1. 𐍃𐌴𐌹𐌳𐍉 𐌸𐍉𐌶𐌴𐌹 𐌰𐌻𐌻𐍉𐍃 𐍅𐌹𐌺𐌹𐍀𐌰𐌹𐌳𐌾𐍉𐍃 𐍃𐌺𐌿𐌻𐌿𐌽 𐌷𐌰𐌱𐌰𐌽 𐍃𐌴𐌹𐌳𐍉 𐌸𐍉𐌶𐌴𐌹 𐌰𐌻𐌻𐍉𐍃 𐍅𐌹𐌺𐌹𐍀𐌰𐌹𐌳𐌾𐍉𐍃 𐍃𐌺𐌿𐌻𐌿𐌽 𐌷𐌰𐌱𐌰𐌽 𐌱𐌰𐌽𐌳𐌰𐍂𐌴𐌹𐌺𐌾𐌹𐍃
2. 𐌹𐌽 𐌰𐌼𐌰𐌹𐍂𐌹𐌺𐌰𐌹 𐌲𐌰𐍅𐌹𐍃𐍃𐌴𐌹𐍃 www stpaul gov
3. 𐌸𐍉𐌶𐌴𐌹 𐌰𐌻𐌻𐍉𐍃 𐍅𐌹𐌺𐌹𐍀𐌰𐌹𐌳𐌾𐍉𐍃 𐍃𐌺𐌿𐌻𐌿𐌽 𐌷𐌰𐌱𐌰𐌽 𐍃𐌴𐌹𐌳𐍉 𐌸𐍉𐌶𐌴𐌹 𐌰𐌻𐌻𐍉𐍃 𐍅𐌹𐌺𐌹𐍀𐌰𐌹𐌳𐌾𐍉𐍃 𐍃𐌺𐌿𐌻𐌿𐌽 𐌷𐌰𐌱𐌰𐌽 𐌱𐌰𐌽𐌳𐌰𐍂𐌴𐌹𐌺𐌾𐌹𐍃

Generated Text Samples (Subword-based)

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

Context Size 1:

1. _sl_1_scoperutce
2. 𐌰𐌹𐌺𐌿𐌸_mago_𐌸𐌰_k,
3. 𐌹𐍈𐌰𐌷𐌹_(*wve._bal

Context Size 2:

1. ,_𐍃𐌴𐌹𐌽𐍃_𐌾𐌰𐌳𐌰,_ble
2. ._oe._arkjan_ram,
3. 𐌰𐌹._infornarusess

Context Size 3:

1. _-_chimess,_munia)
2. n,_with_kaúlustriv
3. s,_mallmers_but_at

Context Size 4:

1. _𐌹𐌽_𐌰𐌼𐌰𐌹𐍂𐌹𐌺𐌹𐍃_𐌿𐌽𐌳_𐌳
2. _to_restone_...hadu
3. _𐌾𐌰𐌷_𐌻𐌹𐌿𐌲𐍉𐍃𐌻𐌰𐌱𐌹𐍃𐌺𐌹𐍃

Key Findings

• Best Predictability: Context-4 (word) with 98.4% predictability
• Branching Factor: Decreases with context size (more deterministic)
• Memory Trade-off: Larger contexts require more storage (92,872 contexts)
• Recommendation: Context-3 or Context-4 for text generation

---

4. Vocabulary Analysis

Statistics

Metric	Value
Vocabulary Size	10,445
Total Tokens	85,682
Mean Frequency	8.20
Median Frequency	3
Frequency Std Dev	41.75

Most Common Words

Rank	Word	Frequency
1	𐌹𐌽	1,691
2	to	1,570
3	𐌾𐌰𐌷	1,478
4	𐌹𐍃𐍄	1,269
5	the	906
6	i	903
7	oe	851
8	ohg	841
9	a	719
10	𐍅𐌰𐍃	616

Least Common Words (from vocabulary)

Rank	Word	Frequency
1	𐌳𐌿𐍄𐍄𐌴	2
2	𐍆𐌹𐌲𐌲𐍂𐌰𐌽𐍃	2
3	𐍃𐌹𐌿𐌺𐌰𐌹𐌶𐌴	2
4	𐌺𐌿𐌺𐌾𐌰𐌽𐌳	2
5	𐌷𐌰𐌹𐍄𐌹𐍃	2
6	𐍃𐌿𐌽𐌸𐍂𐌹𐍃	2
7	𐌷𐌹𐌱𐌰𐌹𐍂𐌾𐍉𐍃	2
8	citerior	2
9	ulterior	2
10	𐌸𐌿𐍂𐌺𐌴𐌹𐍃	2

Zipf's Law Analysis

Metric	Value
Zipf Coefficient	0.8663
R² (Goodness of Fit)	0.982156
Adherence Quality	excellent

Coverage Analysis

Top N Words	Coverage
Top 100	33.8%
Top 1,000	63.2%
Top 5,000	86.7%
Top 10,000	99.0%

Key Findings

• Zipf Compliance: R²=0.9822 indicates excellent adherence to Zipf's law
• High Frequency Dominance: Top 100 words cover 33.8% of corpus
• Long Tail: 445 words needed for remaining 1.0% 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	0.1831 🏆	0.4505	N/A	N/A
mono_64d	64	0.0766	0.4301	N/A	N/A
mono_128d	128	0.0136	0.4355	N/A	N/A
aligned_32d	32	0.1831	0.4429	0.0080	0.0680
aligned_64d	64	0.0766	0.4301	0.0080	0.0740
aligned_128d	128	0.0136	0.4348	0.0160	0.0900

Key Findings

• Best Isotropy: mono_32d with 0.1831 (more uniform distribution)
• Semantic Density: Average pairwise similarity of 0.4373. Lower values indicate better semantic separation.
• Alignment Quality: Aligned models achieve up to 1.6% 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	1.146	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

Productive Suffixes

Suffix	Examples
-an	ocean, wan, hauhjan
-𐌽𐍃	𐌵𐌴𐌽𐍃, 𐌺𐌰𐌷𐍅𐌴𐌹𐌽𐍃, 𐌱𐍂𐌿𐌺𐌴𐌹𐌽𐍃

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	2.06x	24 contexts	there, other, others
𐌰𐌿𐍂𐌳	1.98x	18 contexts	𐍅𐌰𐌿𐍂𐌳, 𐍅𐌰𐌿𐍂𐌳𐌴, 𐍅𐌰𐌿𐍂𐌳𐌰
tion	2.11x	14 contexts	option, motion, nation
𐌴𐌹𐌽𐌰	1.83x	16 contexts	𐌺𐌴𐌹𐌽𐌰, 𐌼𐌴𐌹𐌽𐌰, 𐍅𐌴𐌹𐌽𐌰
𐍅𐌰𐌿𐍂	1.80x	14 contexts	𐍅𐌰𐌿𐍂𐌳, 𐍅𐌰𐌿𐍂𐌳𐌴, 𐍅𐌰𐌿𐍂𐌳𐌰
𐌿𐌳𐌰𐌽	2.08x	9 contexts	𐌲𐌿𐌳𐌰𐌽𐍃, 𐌸𐌹𐌿𐌳𐌰𐌽, 𐌸𐌹𐌿𐌳𐌰𐌽𐍃
𐌹𐌿𐌳𐌰	1.71x	14 contexts	𐌻𐌹𐌿𐌳𐌰, 𐌸𐌹𐌿𐌳𐌰, 𐌸𐌹𐌿𐌳𐌰𐌹
𐌾𐌰𐌽𐌳	1.62x	16 contexts	𐍃𐍉𐌺𐌾𐌰𐌽𐌳, 𐍅𐌰𐌲𐌾𐌰𐌽𐌳, 𐌼𐌰𐍄𐌾𐌰𐌽𐌳
𐍂𐌰𐌶𐌳	1.98x	9 contexts	𐍂𐌰𐌶𐌳𐍉, 𐍂𐌰𐌶𐌳𐌰, 𐍂𐌰𐌶𐌳𐍉𐌼
𐌹𐌽𐌰𐌹	1.88x	10 contexts	𐌰𐌹𐌽𐌰𐌹, 𐍃𐌹𐌽𐌰𐌹, 𐍃𐌴𐌹𐌽𐌰𐌹
𐌷𐌰𐌱𐌰	1.91x	9 contexts	𐌷𐌰𐌱𐌰𐌽, 𐌷𐌰𐌱𐌰𐌼, 𐌷𐌰𐌱𐌰𐌹𐌸
𐍂𐌴𐌹𐌺	1.82x	10 contexts	𐍂𐌴𐌹𐌺𐍃, 𐍂𐌴𐌹𐌺𐌹, 𐍂𐌴𐌹𐌺𐌹𐍃

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.

No significant affix co-occurrences detected.

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
𐍃𐌺𐌰𐌿𐌽𐌴𐌹𐌽𐍃	𐍃𐌺𐌰𐌿𐌽𐌴𐌹-𐌽𐍃	4.5	𐍃𐌺𐌰𐌿𐌽𐌴𐌹
𐍆𐍂𐌿𐌼𐌹𐍃𐍄𐍉𐌽𐍃	𐍆𐍂𐌿𐌼𐌹𐍃𐍄𐍉-𐌽𐍃	4.5	𐍆𐍂𐌿𐌼𐌹𐍃𐍄𐍉
𐌼𐌿𐌽𐌳𐍂𐌴𐌹𐌽𐍃	𐌼𐌿𐌽𐌳𐍂𐌴𐌹-𐌽𐍃	4.5	𐌼𐌿𐌽𐌳𐍂𐌴𐌹
𐌰𐌿𐍃𐍄𐍂𐌰𐌲𐌿𐍄𐌰𐌽𐍃	𐌰𐌿𐍃𐍄𐍂𐌰𐌲𐌿𐍄𐌰-𐌽𐍃	4.5	𐌰𐌿𐍃𐍄𐍂𐌰𐌲𐌿𐍄𐌰
𐌰𐌽𐌳𐌽𐌿𐌼𐌰𐌽𐍃	𐌰𐌽𐌳𐌽𐌿𐌼𐌰-𐌽𐍃	1.5	𐌰𐌽𐌳𐌽𐌿𐌼𐌰
𐌲𐌰𐌲𐌰𐌷𐌰𐍆𐍄𐌾𐌰𐌽𐌳𐌰𐌽𐍃	𐌲𐌰𐌲𐌰𐌷𐌰𐍆𐍄𐌾𐌰𐌽𐌳𐌰-𐌽𐍃	1.5	𐌲𐌰𐌲𐌰𐌷𐌰𐍆𐍄𐌾𐌰𐌽𐌳𐌰
porthpean	porthpe-an	1.5	porthpe
barbarian	barbari-an	1.5	barbari
scandinavian	scandinavi-an	1.5	scandinavi
𐍆𐍂𐌹𐌾𐌰𐍄𐌹𐌼𐍂𐌴𐌹𐌽𐍃	𐍆𐍂𐌹𐌾𐌰𐍄𐌹𐌼𐍂𐌴𐌹-𐌽𐍃	1.5	𐍆𐍂𐌹𐌾𐌰𐍄𐌹𐌼𐍂𐌴𐌹
𐌷𐍂𐌿𐌲𐌾𐌰𐌱𐌰𐌹𐌽𐌰𐌽𐍃	𐌷𐍂𐌿𐌲𐌾𐌰𐌱𐌰𐌹𐌽𐌰-𐌽𐍃	1.5	𐌷𐍂𐌿𐌲𐌾𐌰𐌱𐌰𐌹𐌽𐌰
𐌼𐌰𐌾𐌰𐌹𐌽𐌾𐍉𐌽𐍃	𐌼𐌰𐌾𐌰𐌹𐌽𐌾𐍉-𐌽𐍃	1.5	𐌼𐌰𐌾𐌰𐌹𐌽𐌾𐍉
macmillan	macmill-an	1.5	macmill
𐌼𐌹𐌻𐌿𐌺𐍃𐍆𐍉𐌳𐌾𐌰𐌽𐍃	𐌼𐌹𐌻𐌿𐌺𐍃𐍆𐍉𐌳𐌾𐌰-𐌽𐍃	1.5	𐌼𐌹𐌻𐌿𐌺𐍃𐍆𐍉𐌳𐌾𐌰
𐌽𐌹𐍂𐌱𐌰𐌽𐌹𐌽𐍃	𐌽𐌹𐍂𐌱𐌰𐌽𐌹-𐌽𐍃	1.5	𐌽𐌹𐍂𐌱𐌰𐌽𐌹

6.6 Linguistic Interpretation

Automated Insight: The language Gothic 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	32k BPE	Best compression (2.88x)
N-gram	2-gram	Lowest perplexity (546)
Markov	Context-4	Highest predictability (98.4%)
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: 2026-01-04 15:24:37