README.md

---
language: vec
language_name: Venetian
language_family: romance_galloitalic
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-romance_galloitalic
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: 3.863
  - name: best_isotropy
    type: isotropy
    value: 0.7720
  - name: vocabulary_size
    type: vocab
    value: 0
generated: 2026-01-11
---

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

This repository contains NLP models trained and evaluated by Wikilangs, specifically on **Venetian** 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** | 3.304x | 3.31 | 0.0784% | 181,229 |
| **16k** | 3.529x | 3.54 | 0.0837% | 169,663 |
| **32k** | 3.715x | 3.72 | 0.0881% | 161,162 |
| **64k** | 3.863x 🏆 | 3.87 | 0.0916% | 155,004 |

### Tokenization Examples

Below are sample sentences tokenized with each vocabulary size:

**Sample 1:** `El 256 (CCLVI en numeri romani) el xe on an del III secoło. Avegnimenti Nasesti ...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁el ▁ 2 5 6 ▁( ccl vi ▁en ▁numeri ... (+16 more)` | 26 |
| 16k | `▁el ▁ 2 5 6 ▁( ccl vi ▁en ▁numeri ... (+16 more)` | 26 |
| 32k | `▁el ▁ 2 5 6 ▁( ccl vi ▁en ▁numeri ... (+16 more)` | 26 |
| 64k | `▁el ▁ 2 5 6 ▁( ccl vi ▁en ▁numeri ... (+16 more)` | 26 |

**Sample 2:** `El 144 v.C. (CXLIV v.C par numari romani) el xe on an de el II secoło v.C.. Aveg...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁el ▁ 1 4 4 ▁v . c . ▁( ... (+32 more)` | 42 |
| 16k | `▁el ▁ 1 4 4 ▁v . c . ▁( ... (+31 more)` | 41 |
| 32k | `▁el ▁ 1 4 4 ▁v . c . ▁( ... (+31 more)` | 41 |
| 64k | `▁el ▁ 1 4 4 ▁v . c . ▁( ... (+31 more)` | 41 |

**Sample 3:** `el xe un comun del distreto de Lenzburg che el fa parte del canton Argovia in Sv...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁el ▁xe ▁un ▁comun ▁del ▁distreto ▁de ▁len z burg ... (+15 more)` | 25 |
| 16k | `▁el ▁xe ▁un ▁comun ▁del ▁distreto ▁de ▁len zburg ▁che ... (+14 more)` | 24 |
| 32k | `▁el ▁xe ▁un ▁comun ▁del ▁distreto ▁de ▁lenzburg ▁che ▁el ... (+13 more)` | 23 |
| 64k | `▁el ▁xe ▁un ▁comun ▁del ▁distreto ▁de ▁lenzburg ▁che ▁el ... (+13 more)` | 23 |


### Key Findings

- **Best Compression:** 64k achieves 3.863x compression
- **Lowest UNK Rate:** 8k with 0.0784% 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 | 4,312 | 12.07 | 91,618 | 40.5% | 59.1% |
| **2-gram** | Subword | 223 🏆 | 7.80 | 5,564 | 73.1% | 99.2% |
| **3-gram** | Word | 4,702 | 12.20 | 134,286 | 42.0% | 60.0% |
| **3-gram** | Subword | 1,552 | 10.60 | 41,266 | 35.7% | 78.3% |
| **4-gram** | Word | 4,657 | 12.19 | 186,223 | 41.4% | 63.1% |
| **4-gram** | Subword | 7,211 | 12.82 | 219,587 | 24.6% | 52.4% |
| **5-gram** | Word | 3,493 | 11.77 | 114,029 | 40.0% | 65.3% |
| **5-gram** | Subword | 22,392 | 14.45 | 608,866 | 19.7% | 41.0% |

### Top 5 N-grams by Size

**2-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `de ła` | 73,737 |
| 2 | `el xe` | 70,338 |
| 3 | `departemento de` | 68,217 |
| 4 | `del departemento` | 67,585 |
| 5 | `altri projeti` | 57,004 |

**3-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `del departemento de` | 67,534 |
| 2 | `el xe on` | 51,956 |
| 3 | `xe on comun` | 48,810 |
| 4 | `demogràfega altri projeti` | 42,469 |
| 5 | `evołusion demogràfega altri` | 42,466 |

**4-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `el xe on comun` | 48,761 |
| 2 | `evołusion demogràfega altri projeti` | 42,466 |
| 3 | `xe on comun de` | 41,994 |
| 4 | `che el fa parte` | 37,577 |
| 5 | `el fa parte del` | 37,224 |

**5-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `el xe on comun de` | 41,982 |
| 2 | `che el fa parte del` | 37,190 |
| 3 | `el fa parte del rejon` | 33,708 |
| 4 | `in fransa evołusion demogràfega altri` | 33,510 |
| 5 | `fransa evołusion demogràfega altri projeti` | 33,510 |

**2-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `e _` | 1,265,574 |
| 2 | `a _` | 993,554 |
| 3 | `_ d` | 907,290 |
| 4 | `d e` | 819,733 |
| 5 | `l _` | 515,433 |

**3-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `_ d e` | 746,746 |
| 2 | `e l _` | 427,367 |
| 3 | `d e _` | 422,586 |
| 4 | `o n _` | 229,675 |
| 5 | `_ e l` | 229,151 |

**4-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `_ d e _` | 408,249 |
| 2 | `_ e l _` | 225,182 |
| 3 | `_ ł a _` | 183,914 |
| 4 | `_ d e l` | 164,044 |
| 5 | `d e l _` | 159,382 |

**5-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `_ d e l _` | 159,062 |
| 2 | `p a r t e` | 129,822 |
| 3 | `o _ d e _` | 120,109 |
| 4 | `e _ ł a _` | 117,322 |
| 5 | `s i o n _` | 95,313 |


### Key Findings

- **Best Perplexity:** 2-gram (subword) with 223
- **Entropy Trend:** Decreases with larger n-grams (more predictable)
- **Coverage:** Top-1000 patterns cover ~41% 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 | 0.8047 | 1.747 | 5.60 | 282,129 | 19.5% |
| **1** | Subword | 0.8604 | 1.816 | 6.21 | 2,732 | 14.0% |
| **2** | Word | 0.2918 | 1.224 | 1.77 | 1,575,997 | 70.8% |
| **2** | Subword | 0.8328 | 1.781 | 5.24 | 16,965 | 16.7% |
| **3** | Word | 0.1167 | 1.084 | 1.22 | 2,791,577 | 88.3% |
| **3** | Subword | 0.7829 | 1.721 | 4.23 | 88,847 | 21.7% |
| **4** | Word | 0.0422 🏆 | 1.030 | 1.06 | 3,391,639 | 95.8% |
| **4** | Subword | 0.6983 | 1.623 | 3.15 | 375,778 | 30.2% |

### Generated Text Samples (Word-based)

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

**Context Size 1:**

1. `de ła provinsa de ła xe un sècoło v c co el ga susitò i ritràti`
2. `el xe na part abitasion privada che ła comunità autònoma de 89 abitanti del film montà`
3. `ła provinsa de 479 abitanti del primo caxo asołutivo ergativo asołutivo el fa parte del departemento`

**Context Size 2:**

1. `de ła provinsa de groninga na picenina organizasion ciamada dont make me feel brand new bag i`
2. `el xe on comun marcà del distreto de scheibbs del distreto de bruck an der leitha che`
3. `departemento de nord che el fa parte del rejon nova acuitania in fransa evołusion demogràfega altri ...`

**Context Size 3:**

1. `del departemento de haute saône che el fa parte del rejon alvergna rodano alpe in fransa evołusion d...`
2. `el xe on comun de ła spagna situà inte ła provinsa de alicante che ła fa parte de`
3. `xe on comun de 146 abitanti del departemento de lozère che el fa parte del del stato de`

**Context Size 4:**

1. `el xe on comun de 476 abitanti del departemento de vaucluse che el fa parte del rejon grand est`
2. `evołusion demogràfega altri projeti del departemento de drôme che el fa parte del stato de ła alta à...`
3. `xe on comun de 516 abitanti del departemento de côte d or che el fa parte del rejon ositània`


### Generated Text Samples (Subword-based)

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

**Context Size 1:**

1. `_lttintuzel-2_po`
2. `e_(li_onsetforo_`
3. `ali_densè_pare,_`

**Context Size 2:**

1. `e_oire_de_unìodo_`
2. `a_proverssensa_de`
3. `_deorquandopartom`

**Context Size 3:**

1. `_de_183_abitanti_d`
2. `el_bas-rhône-frang`
3. `de_ave_al_de_sento`

**Context Size 4:**

1. `_de_aisne_-_lujo_de`
2. `_el_fa_par_posti_de`
3. `_ła_u_partemento_de`


### 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 (375,778 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 | 119,267 |
| Total Tokens | 5,515,860 |
| Mean Frequency | 46.25 |
| Median Frequency | 4 |
| Frequency Std Dev | 1838.83 |

### Most Common Words

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | de | 422,791 |
| 2 | el | 251,936 |
| 3 | ła | 185,729 |
| 4 | del | 159,907 |
| 5 | xe | 95,799 |
| 6 | e | 88,103 |
| 7 | che | 86,802 |
| 8 | in | 85,859 |
| 9 | l | 73,523 |
| 10 | departemento | 68,444 |

### Least Common Words (from vocabulary)

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | güvenli | 2 |
| 2 | taşımacılık | 2 |
| 3 | sunuyoruz | 2 |
| 4 | edebilirsiniz | 2 |
| 5 | parça | 2 |
| 6 | sensorial | 2 |
| 7 | complicada | 2 |
| 8 | caregari | 2 |
| 9 | sabigotho | 2 |
| 10 | paułista | 2 |

### Zipf's Law Analysis

| Metric | Value |
|--------|-------|
| Zipf Coefficient | 1.0353 |
| R² (Goodness of Fit) | 0.998145 |
| Adherence Quality | **excellent** |

### Coverage Analysis

| Top N Words | Coverage |
|-------------|----------|
| Top 100 | 56.8% |
| Top 1,000 | 72.7% |
| Top 5,000 | 83.6% |
| Top 10,000 | 88.2% |

### Key Findings

- **Zipf Compliance:** R²=0.9981 indicates excellent adherence to Zipf's law
- **High Frequency Dominance:** Top 100 words cover 56.8% of corpus
- **Long Tail:** 109,267 words needed for remaining 11.8% 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 | 0.7685 | 0.3278 | N/A | N/A |
| **mono_64d** | 64 | 0.7720 🏆 | 0.2784 | N/A | N/A |
| **mono_128d** | 128 | 0.7461 | 0.2091 | N/A | N/A |
| **aligned_32d** | 32 | 0.7685 | 0.3249 | 0.0880 | 0.3700 |
| **aligned_64d** | 64 | 0.7720 | 0.2747 | 0.1500 | 0.4740 |
| **aligned_128d** | 128 | 0.7461 | 0.2092 | 0.2280 | 0.5740 |

### Key Findings

- **Best Isotropy:** mono_64d with 0.7720 (more uniform distribution)
- **Semantic Density:** Average pairwise similarity of 0.2707. Lower values indicate better semantic separation.
- **Alignment Quality:** Aligned models achieve up to 22.8% 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 | **0.594** | 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 |
|--------|----------|
| `-s` | sosiałizasion, scumisi, sarr |
| `-a` | antegamente, adeti, anthology |
| `-c` | cctv, cussìta, coṅkiṅ |
| `-p` | presidensa, palácio, pinin |
| `-m` | mathieu, mesonà, megało |
| `-ma` | mathieu, maxistero, maschi |
| `-b` | bajijo, baloo, bałene |
| `-ca` | cale, canałizasion, caronte |

#### Productive Suffixes
| Suffix | Examples |
|--------|----------|
| `-e` | garantise, erdre, antegamente |
| `-a` | taxa, fondarìa, presidensa |
| `-o` | energetico, palácio, successivo |
| `-i` | scumisi, laóri, lupi |
| `-n` | sosiałizasion, eugen, pinin |
| `-on` | sosiałizasion, canałizasion, musurareon |
| `-s` | infos, snows, gladys |
| `-te` | antegamente, facontinente, desferente |

### 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 |
|------|----------|------------------|----------|
| `ento` | 2.32x | 93 contexts | bento, vento, zento |
| `ment` | 1.96x | 170 contexts | menti, mento, mente |
| `altr` | 2.16x | 43 contexts | altri, altra, altre |
| `ltri` | 2.56x | 18 contexts | altri, altria, filtri |
| `emen` | 1.75x | 64 contexts | hemen, iemen, yemen |
| `ołus` | 2.58x | 15 contexts | mołuski, mołusco, sołusion |
| `omun` | 1.94x | 36 contexts | comun, komun, comune |
| `itan` | 1.53x | 95 contexts | titan, kitang, gitana |
| `ejon` | 2.32x | 17 contexts | rejon, lejon, prejon |
| `fega` | 2.03x | 25 contexts | fegato, sòfega, grafega |
| `comu` | 2.07x | 18 contexts | comun, comum, comune |
| `epar` | 1.69x | 35 contexts | separa, separà, separè |

### 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 |
|--------|--------|-----------|----------|
| `-c` | `-e` | 151 words | canpanarie, conosùe |
| `-c` | `-a` | 141 words | cołùnbia, cołonia |
| `-s` | `-o` | 125 words | sapporo, situato |
| `-s` | `-a` | 119 words | scrita, stamperia |
| `-c` | `-o` | 114 words | cantabrico, contatto |
| `-s` | `-e` | 112 words | sdrùciołe, severamente |
| `-p` | `-o` | 107 words | primo, perìgoło |
| `-p` | `-e` | 106 words | percepire, prostituzione |
| `-s` | `-i` | 104 words | sigismondi, squilli |
| `-c` | `-i` | 99 words | culti, conservatrici |

### 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 |
|------|-----------------|------------|------|
| costituindo | **`costitu-in-do`** | 7.5 | `in` |
| continuando | **`continu-an-do`** | 7.5 | `an` |
| tełevizore | **`tełeviz-o-re`** | 7.5 | `o` |
| marełéngua | **`ma-re-łéngua`** | 7.5 | `łéngua` |
| festixava | **`festix-a-va`** | 7.5 | `a` |
| anałòxego | **`anałòx-e-go`** | 7.5 | `e` |
| vendidori | **`vendid-o-ri`** | 7.5 | `o` |
| discontinuità | **`discontinu-i-tà`** | 7.5 | `i` |
| francobołi | **`francob-o-łi`** | 7.5 | `o` |
| charleroi | **`charler-o-i`** | 7.5 | `o` |
| sommières | **`sommiè-re-s`** | 7.5 | `re` |
| giacobini | **`giacob-i-ni`** | 7.5 | `i` |
| incorpando | **`incorp-an-do`** | 7.5 | `an` |
| sicatrise | **`sicat-ri-se`** | 7.5 | `ri` |
| partecipaxion | **`partecipax-i-on`** | 7.5 | `i` |

### 6.6 Linguistic Interpretation

> **Automated Insight:**
The language Venetian 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

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

### Production Recommendations

| Component | Recommended | Rationale |
|-----------|-------------|-----------|
| Tokenizer | **64k BPE** | Best compression (3.86x) |
| N-gram | **2-gram** | Lowest perplexity (223) |
| 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: 2026-01-11 03:08:09*