README.md

---
language: tr
language_name: Turkish
language_family: turkic_oghuz
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-turkic_oghuz
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: 4.777
  - name: best_isotropy
    type: isotropy
    value: 0.7797
  - name: vocabulary_size
    type: vocab
    value: 0
generated: 2026-01-18
---

# Turkish - Wikilangs Models
## Comprehensive Research Report & Full Ablation Study

This repository contains NLP models trained and evaluated by Wikilangs, specifically on **Turkish** 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

![Performance Dashboard](visualizations/performance_dashboard.png)

### Analysis and Evaluation

- [1. Tokenizer Evaluation](#1-tokenizer-evaluation)
- [2. N-gram Model Evaluation](#2-n-gram-model-evaluation)
- [3. Markov Chain Evaluation](#3-markov-chain-evaluation)
- [4. Vocabulary Analysis](#4-vocabulary-analysis)
- [5. Word Embeddings Evaluation](#5-word-embeddings-evaluation)
- [6. Morphological Analysis (Experimental)](#6--morphological-analysis-experimental)
- [7. Summary & Recommendations](#7-summary--recommendations)
- [Metrics Glossary](#appendix-metrics-glossary--interpretation-guide)
- [Visualizations Index](#visualizations-index)

---
## 1. Tokenizer Evaluation

![Tokenizer Compression](visualizations/tokenizer_compression.png)

![Tokenizer Fertility](visualizations/tokenizer_fertility.png)

![Tokenizer OOV](visualizations/tokenizer_oov.png)

![Total Tokens](visualizations/tokenizer_total_tokens.png)

### Results

| Vocab Size | Compression | Avg Token Len | UNK Rate | Total Tokens |
|------------|-------------|---------------|----------|--------------|
| **8k** | 3.702x | 3.70 | 0.0636% | 2,131,433 |
| **16k** | 4.112x | 4.11 | 0.0706% | 1,918,952 |
| **32k** | 4.477x | 4.48 | 0.0769% | 1,762,504 |
| **64k** | 4.777x 🏆 | 4.78 | 0.0820% | 1,651,752 |

### Tokenization Examples

Below are sample sentences tokenized with each vocabulary size:

**Sample 1:** `Bubiacoris Harpactorini oymağına bağlı bir böcek cinsidir. Kaynakça Dış bağlantı...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁bu bi ac or is ▁harp ac tor ini ▁oy ... (+13 more)` | 23 |
| 16k | `▁bu bi ac or is ▁harp ac tor ini ▁oy ... (+12 more)` | 22 |
| 32k | `▁bu bi ac or is ▁harp ac tor ini ▁oy ... (+12 more)` | 22 |
| 64k | `▁bu bi ac oris ▁harp actor ini ▁oym ağına ▁bağlı ... (+9 more)` | 19 |

**Sample 2:** `Monobothrium, Caryophyllaeidae familyasına bağlı bir hayvan cinsidir. Kaynakça D...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁mon ob oth ri um , ▁car y op hy ... (+16 more)` | 26 |
| 16k | `▁mon ob oth ri um , ▁car y ophy l ... (+14 more)` | 24 |
| 32k | `▁mon ob oth rium , ▁car y ophy l la ... (+13 more)` | 23 |
| 64k | `▁mon ob oth rium , ▁cary ophyl la e idae ... (+11 more)` | 21 |

**Sample 3:** `Spilophora, Spilophorini oymağına bağlı bir hayvan cinsidir. Kaynakça Dış bağlan...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁sp il oph ora , ▁sp il oph or ini ... (+14 more)` | 24 |
| 16k | `▁sp il oph ora , ▁sp il oph or ini ... (+13 more)` | 23 |
| 32k | `▁sp il oph ora , ▁sp il oph or ini ... (+13 more)` | 23 |
| 64k | `▁sp il ophora , ▁sp il oph or ini ▁oym ... (+11 more)` | 21 |


### Key Findings

- **Best Compression:** 64k achieves 4.777x compression
- **Lowest UNK Rate:** 8k with 0.0636% 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

![N-gram Perplexity](visualizations/ngram_perplexity.png)

![N-gram Unique](visualizations/ngram_unique.png)

![N-gram Coverage](visualizations/ngram_coverage.png)

### Results

| N-gram | Variant | Perplexity | Entropy | Unique N-grams | Top-100 Coverage | Top-1000 Coverage |
|--------|---------|------------|---------|----------------|------------------|-------------------|
| **2-gram** | Word | 517,762 | 18.98 | 3,151,475 | 4.5% | 12.1% |
| **2-gram** | Subword | 369 🏆 | 8.53 | 40,485 | 60.0% | 98.4% |
| **3-gram** | Word | 1,234,358 | 20.24 | 4,813,201 | 4.1% | 9.2% |
| **3-gram** | Subword | 3,553 | 11.79 | 274,497 | 20.2% | 62.5% |
| **4-gram** | Word | 2,217,879 | 21.08 | 7,386,404 | 3.8% | 8.3% |
| **4-gram** | Subword | 22,183 | 14.44 | 1,542,206 | 9.2% | 31.9% |
| **5-gram** | Word | 1,582,796 | 20.59 | 5,151,376 | 4.3% | 9.3% |
| **5-gram** | Subword | 97,109 | 16.57 | 5,359,470 | 5.4% | 19.6% |

### Top 5 N-grams by Size

**2-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `dış bağlantılar` | 350,729 |
| 2 | `kaynakça dış` | 266,247 |
| 3 | `bağlı bir` | 174,377 |
| 4 | `daha sonra` | 94,043 |
| 5 | `ya da` | 87,702 |

**3-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `kaynakça dış bağlantılar` | 265,490 |
| 2 | `cinsine bağlı bir` | 66,200 |
| 3 | `türüdür kaynakça dış` | 54,307 |
| 4 | `bağlı bir hayvan` | 46,777 |
| 5 | `amerika birleşik devletleri` | 46,424 |

**4-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `türüdür kaynakça dış bağlantılar` | 54,307 |
| 2 | `kaynakça dış bağlantılar tanımlanan` | 39,619 |
| 3 | `dış bağlantılar tanımlanan taksonlar` | 38,187 |
| 4 | `bağlı bir bitki türüdür` | 34,768 |
| 5 | `cinsine bağlı bir bitki` | 34,767 |

**5-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `kaynakça dış bağlantılar tanımlanan taksonlar` | 38,177 |
| 2 | `cinsine bağlı bir bitki türüdür` | 34,766 |
| 3 | `bitki türüdür kaynakça dış bağlantılar` | 33,470 |
| 4 | `bir bitki türüdür kaynakça dış` | 33,468 |
| 5 | `bağlı bir bitki türüdür kaynakça` | 33,157 |

**2-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `n _` | 17,593,330 |
| 2 | `a r` | 17,219,363 |
| 3 | `a n` | 15,720,144 |
| 4 | `e _` | 15,698,088 |
| 5 | `l a` | 14,911,706 |

**3-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `l a r` | 7,165,653 |
| 2 | `l e r` | 5,383,393 |
| 3 | `a n _` | 5,370,904 |
| 4 | `e r i` | 4,812,356 |
| 5 | `_ v e` | 4,389,139 |

**4-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `_ v e _` | 3,553,259 |
| 2 | `_ b i r` | 3,136,915 |
| 3 | `l a r ı` | 3,011,298 |
| 4 | `l e r i` | 2,857,164 |
| 5 | `ı n d a` | 2,744,063 |

**5-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `_ b i r _` | 2,224,652 |
| 2 | `ı n d a _` | 1,537,643 |
| 3 | `l a r ı _` | 1,490,277 |
| 4 | `l e r i _` | 1,341,301 |
| 5 | `l a r ı n` | 1,195,116 |


### Key Findings

- **Best Perplexity:** 2-gram (subword) with 369
- **Entropy Trend:** Decreases with larger n-grams (more predictable)
- **Coverage:** Top-1000 patterns cover ~20% of corpus
- **Recommendation:** 4-gram or 5-gram for best predictive performance

---
## 3. Markov Chain Evaluation

![Markov Entropy](visualizations/markov_entropy.png)

![Markov Contexts](visualizations/markov_contexts.png)

![Markov Branching](visualizations/markov_branching.png)

### Results

| Context | Variant | Avg Entropy | Perplexity | Branching Factor | Unique Contexts | Predictability |
|---------|---------|-------------|------------|------------------|-----------------|----------------|
| **1** | Word | 0.9720 | 1.962 | 14.51 | 3,106,535 | 2.8% |
| **1** | Subword | 1.4308 | 2.696 | 11.15 | 15,908 | 0.0% |
| **2** | Word | 0.3547 | 1.279 | 2.15 | 45,020,523 | 64.5% |
| **2** | Subword | 0.6247 | 1.542 | 4.03 | 177,402 | 37.5% |
| **3** | Word | 0.1200 | 1.087 | 1.24 | 96,715,794 | 88.0% |
| **3** | Subword | 0.6614 | 1.582 | 3.95 | 714,448 | 33.9% |
| **4** | Word | 0.0421 🏆 | 1.030 | 1.07 | 119,918,003 | 95.8% |
| **4** | Subword | 0.6640 | 1.584 | 3.51 | 2,821,176 | 33.6% |

### Generated Text Samples (Word-based)

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

**Context Size 1:**

1. `ve ağır bir şehirdir visayas bikol vikipedi maddeleri toplam 4 sumon olarak bilinmekteydi ve mevcut ...`
2. `bir puan durumu the protein kodlamayan genlerin etkisinin bu yolla temin etmek zorunda kalan ve yazı...`
3. `olarak başladı bir şekilde sona eren i hakîkat mecmuasında yayınlamıştır aktör adaylığı da kariyerin...`

**Context Size 2:**

1. `dış bağlantılar tanımlanan taksonlar j currie tarafından adlandırılmış taksonlar tanımlanan bitkiler...`
2. `kaynakça dış bağlantılar dünya su forumu bakanlar arası toplantılarına komitelerine konseylerine ve ...`
3. `bağlı bir beldeye dönüştü coğrafya köy adıyaman il merkezine 17 km uzaklıktadır nüfus yıllara göre m...`

**Context Size 3:**

1. `kaynakça dış bağlantılar tbmm internet sitesinde nilhan ayan doğumlular kadın milletvekilleri üniver...`
2. `cinsine bağlı bir bitki türüdür kaynakça dış bağlantılar tanımlanan taksonlar weed fowler tarafından...`
3. `türüdür kaynakça dış bağlantılar tanımlanan taksonlar jakob kaup tarafından adlandırılmış taksonlar ...`

**Context Size 4:**

1. `kaynakça dış bağlantılar tanımlanan taksonlar jakob kaup tarafından adlandırılmış taksonlar tanımlan...`
2. `bağlı bir bitki türüdür kaynakça dış bağlantılar florası florası florası florası florası florası flo...`
3. `cinsine bağlı bir bitki türüdür kaynakça dış bağlantılar florası tanımlanan bitkiler`


### Generated Text Samples (Subword-based)

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

**Context Size 1:**

1. `_dharirışman_ver`
2. `aoktasi_k_onanl_`
3. `er,_olurdık,_ve_`

**Context Size 2:**

1. `n_başmarımca:_"he`
2. `ar_sörlenişiklar_`
3. `an_pkkande_i̇ste_d`

**Context Size 3:**

1. `ları_şarkan_van_jo`
2. `an_aman_kada,_eyal`
3. `ler_için_iler_açta`

**Context Size 4:**

1. `_ve_kendi_hükûmet-e`
2. `_biridir._dışı_örne`
3. `larına_ve_inşasında`


### Key Findings

- **Best Predictability:** Context-4 (word) with 95.8% predictability
- **Branching Factor:** Decreases with context size (more deterministic)
- **Memory Trade-off:** Larger contexts require more storage (2,821,176 contexts)
- **Recommendation:** Context-3 or Context-4 for text generation

---
## 4. Vocabulary Analysis

![Zipf's Law](visualizations/zipf_law.png)

![Top Words](visualizations/top20_words.png)

![Coverage Curve](visualizations/vocab_coverage.png)

### Statistics

| Metric | Value |
|--------|-------|
| Vocabulary Size | 1,401,382 |
| Total Tokens | 145,506,395 |
| Mean Frequency | 103.83 |
| Median Frequency | 4 |
| Frequency Std Dev | 4551.48 |

### Most Common Words

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | ve | 3,563,467 |
| 2 | bir | 2,239,393 |
| 3 | olarak | 863,214 |
| 4 | da | 854,555 |
| 5 | bu | 838,172 |
| 6 | ile | 733,358 |
| 7 | de | 723,655 |
| 8 | 1 | 660,785 |
| 9 | için | 614,599 |
| 10 | the | 514,508 |

### Least Common Words (from vocabulary)

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | rawvixen | 2 |
| 2 | groupweb | 2 |
| 3 | castorum | 2 |
| 4 | othonlular | 2 |
| 5 | pnujsciewarty | 2 |
| 6 | mensuris | 2 |
| 7 | ponderibus | 2 |
| 8 | hexaplaric | 2 |
| 9 | titanozorların | 2 |
| 10 | noëp | 2 |

### Zipf's Law Analysis

| Metric | Value |
|--------|-------|
| Zipf Coefficient | 0.9525 |
| R² (Goodness of Fit) | 0.994421 |
| Adherence Quality | **excellent** |

### Coverage Analysis

| Top N Words | Coverage |
|-------------|----------|
| Top 100 | 21.0% |
| Top 1,000 | 44.4% |
| Top 5,000 | 63.7% |
| Top 10,000 | 71.8% |

### Key Findings

- **Zipf Compliance:** R²=0.9944 indicates excellent adherence to Zipf's law
- **High Frequency Dominance:** Top 100 words cover 21.0% of corpus
- **Long Tail:** 1,391,382 words needed for remaining 28.2% coverage

---
## 5. Word Embeddings Evaluation

![Embedding Isotropy](visualizations/embedding_isotropy.png)

![Similarity Matrix](visualizations/embedding_similarity.png)

![t-SNE Words](visualizations/tsne_words.png)

![t-SNE Sentences](visualizations/tsne_sentences.png)


### 5.1 Cross-Lingual Alignment

![Alignment Quality](visualizations/embedding_alignment_quality.png)

![Multilingual t-SNE](visualizations/embedding_tsne_multilingual.png)


### 5.2 Model Comparison

| Model | Dimension | Isotropy | Semantic Density | Alignment R@1 | Alignment R@10 |
|-------|-----------|----------|------------------|---------------|----------------|
| **mono_32d** | 32 | 0.7797 🏆 | 0.3582 | N/A | N/A |
| **mono_64d** | 64 | 0.7747 | 0.2849 | N/A | N/A |
| **mono_128d** | 128 | 0.7155 | 0.2346 | N/A | N/A |
| **aligned_32d** | 32 | 0.7797 | 0.3781 | 0.4900 | 0.8140 |
| **aligned_64d** | 64 | 0.7747 | 0.2854 | 0.6780 | 0.9160 |
| **aligned_128d** | 128 | 0.7155 | 0.2378 | 0.7720 | 0.9880 |

### Key Findings

- **Best Isotropy:** mono_32d with 0.7797 (more uniform distribution)
- **Semantic Density:** Average pairwise similarity of 0.2965. Lower values indicate better semantic separation.
- **Alignment Quality:** Aligned models achieve up to 77.2% 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 | **-0.527** | Low formulaic 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 |
|--------|----------|
| `-a` | alexandrovna, almazdan, alayoğlu |
| `-s` | saldırganlara, strimonas, stefańska |
| `-m` | marmarás, maiori, makura |
| `-k` | kentinte, kolonisiydiler, khmaer |
| `-ma` | marmarás, maiori, makura |
| `-t` | trioedd, teräsbetoni, tükettiklerini |
| `-b` | brakana, burgard, blaxland |
| `-ka` | karahasanuşağı, kabrinin, kapilvastu |

#### Productive Suffixes
| Suffix | Examples |
|--------|----------|
| `-n` | eklentilerinden, i̇smailîliğin, almazdan |
| `-a` | brakana, saldırganlara, göremiyorsa |
| `-r` | kolonisiydiler, dieringer, khmaer |
| `-e` | økonomiske, kentinte, dokuzsele |
| `-i` | zemberekli, eğilimlileri, yükselenleri |
| `-s` | marmarás, cortos, strimonas |
| `-en` | eklentilerinden, eyatlerinden, bezden |
| `-an` | almazdan, zayıfladıktan, sawaiyan |

### 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 |
|------|----------|------------------|----------|
| `utbo` | 2.90x | 41 contexts | futbol, futboll, futbola |
| `tbol` | 2.46x | 62 contexts | tboli, fotbol, futbol |
| `futb` | 2.98x | 27 contexts | futbol, futboll, futbola |
| `mışt` | 1.90x | 153 contexts | mıştı, mıştır, aşmıştı |
| `mler` | 1.60x | 310 contexts | imler, emler, dumler |
| `ılar` | 1.44x | 549 contexts | yılar, kılar, cılar |
| `bolc` | 2.73x | 24 contexts | bolca, bolcom, bolcan |
| `nakç` | 2.57x | 29 contexts | inakçı, oynakçı, konakçı |
| `sınd` | 1.75x | 125 contexts | sında, sınde, sındı |
| `ıştı` | 1.59x | 201 contexts | kıştı, mıştı, kıştım |
| `bağl` | 2.21x | 44 contexts | bağli, bağlı, bağla |
| `ılın` | 1.54x | 139 contexts | yılın, kılın, ılında |

### 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 |
|--------|--------|-----------|----------|
| `-k` | `-n` | 119 words | kelamcıların, kressentein |
| `-s` | `-n` | 96 words | swahn, scgn |
| `-s` | `-a` | 93 words | shibahara, stictigastra |
| `-a` | `-a` | 92 words | alaa, ayırdığında |
| `-s` | `-r` | 84 words | sanatsaldır, sodomiler |
| `-a` | `-n` | 84 words | aowin, avron |
| `-k` | `-a` | 78 words | kasnaklara, konaklamaya |
| `-k` | `-r` | 75 words | kürsüler, kuşatılmasıdır |
| `-s` | `-e` | 72 words | salomonmadeleine, sozialtechnologie |
| `-d` | `-n` | 71 words | destanıhaldun, digitalisation |

### 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 |
|------|-----------------|------------|------|
| makinalarına | **`makinaları-n-a`** | 7.5 | `n` |
| uzaklaşırken | **`uzaklaşır-k-en`** | 7.5 | `k` |
| kavuşacağına | **`kavuşacağı-n-a`** | 7.5 | `n` |
| kurslarına | **`kursları-n-a`** | 7.5 | `n` |
| willdenowia | **`willdenow-i-a`** | 7.5 | `i` |
| aktarımdan | **`aktarım-da-n`** | 7.5 | `da` |
| nicaeensis | **`nicaeen-s-is`** | 7.5 | `s` |
| falankstaki | **`falankst-a-ki`** | 7.5 | `a` |
| çekilirken | **`çekilir-k-en`** | 7.5 | `k` |
| toprakküre | **`toprakkü-r-e`** | 7.5 | `r` |
| luvicedeki | **`luvice-de-ki`** | 7.5 | `de` |
| irtifadaki | **`irtifad-a-ki`** | 7.5 | `a` |
| öschelbronn | **`öschelbro-n-n`** | 7.5 | `n` |
| ticketları | **`ticketl-a-rı`** | 7.5 | `a` |
| çalışarak | **`çalışa-ra-k`** | 7.5 | `ra` |

### 6.6 Linguistic Interpretation

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

---
## 7. Summary & Recommendations

![Performance Dashboard](visualizations/performance_dashboard.png)

### Production Recommendations

| Component | Recommended | Rationale |
|-----------|-------------|-----------|
| Tokenizer | **64k BPE** | Best compression (4.78x) |
| N-gram | **2-gram** | Lowest perplexity (369) |
| Markov | **Context-4** | Highest predictability (95.8%) |
| 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](https://huggingface.co/datasets/omarkamali/wikipedia-monthly) - a monthly snapshot of Wikipedia articles across 300+ languages.

### Project

A project by **[Wikilangs](https://wikilangs.org)** - Open-source NLP models for every Wikipedia language.

### Maintainer

[Omar Kamali](https://omarkamali.com) - [Omneity Labs](https://omneitylabs.com)

### Citation

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

```bibtex
@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](https://wikilangs.org)
- 🤗 Models: [huggingface.co/wikilangs](https://huggingface.co/wikilangs)
- 📊 Data: [wikipedia-monthly](https://huggingface.co/datasets/omarkamali/wikipedia-monthly)
- 👤 Author: [Omar Kamali](https://huggingface.co/omarkamali)
- 🤝 Sponsor: [Featherless AI](https://featherless.ai)
---
*Generated by Wikilangs Models Pipeline*

*Report Date: 2026-01-18 06:49:15*