README.md

---
language: gn
language_name: Guarani
language_family: american_guarani
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-american_guarani
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: 4.358
  - name: best_isotropy
    type: isotropy
    value: 0.8633
  - name: vocabulary_size
    type: vocab
    value: 0
generated: 2026-01-04
---

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

This repository contains NLP models trained and evaluated by Wikilangs, specifically on **Guarani** 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** | 3.636x | 3.64 | 0.0335% | 587,801 |
| **16k** | 3.949x | 3.95 | 0.0364% | 541,088 |
| **32k** | 4.196x | 4.20 | 0.0387% | 509,272 |
| **64k** | 4.358x 🏆 | 4.36 | 0.0402% | 490,302 |

### Tokenization Examples

Below are sample sentences tokenized with each vocabulary size:

**Sample 1:** `21 jasyapy ha'e papoapyha ára arygua. Arete Tembiasa Teñõi Mano`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁ 2 1 ▁jasyapy ▁ha ' e ▁papoapy ha ▁ára ... (+6 more)` | 16 |
| 16k | `▁ 2 1 ▁jasyapy ▁ha ' e ▁papoapyha ▁ára ▁arygua ... (+5 more)` | 15 |
| 32k | `▁ 2 1 ▁jasyapy ▁ha ' e ▁papoapyha ▁ára ▁arygua ... (+5 more)` | 15 |
| 64k | `▁ 2 1 ▁jasyapy ▁ha ' e ▁papoapyha ▁ára ▁arygua ... (+5 more)` | 15 |

**Sample 2:** `- ary. Oararecha'akue Hernán Guggiari - 20 jasykõi Ramón Artemio Bracho - 8 jasy...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁- ▁ary . ▁oararecha ' akue ▁her n án ▁gu ... (+21 more)` | 31 |
| 16k | `▁- ▁ary . ▁oararecha ' akue ▁hernán ▁guggiari ▁- ▁ ... (+15 more)` | 25 |
| 32k | `▁- ▁ary . ▁oararecha ' akue ▁hernán ▁guggiari ▁- ▁ ... (+15 more)` | 25 |
| 64k | `▁- ▁ary . ▁oararecha ' akue ▁hernán ▁guggiari ▁- ▁ ... (+15 more)` | 25 |

**Sample 3:** `Reconquista arasẽme tava Argentina retãme. Oĩhína tetãvore Santa Fe-me. Ko távap...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁re con qu ista ▁ara sẽme ▁tava ▁argentina ▁retãme . ... (+21 more)` | 31 |
| 16k | `▁re con quista ▁arasẽme ▁tava ▁argentina ▁retãme . ▁oĩhína ▁tetãvore ... (+19 more)` | 29 |
| 32k | `▁recon quista ▁arasẽme ▁tava ▁argentina ▁retãme . ▁oĩhína ▁tetãvore ▁santa ... (+18 more)` | 28 |
| 64k | `▁reconquista ▁arasẽme ▁tava ▁argentina ▁retãme . ▁oĩhína ▁tetãvore ▁santa ▁fe ... (+17 more)` | 27 |


### Key Findings

- **Best Compression:** 64k achieves 4.358x compression
- **Lowest UNK Rate:** 8k with 0.0335% 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 | 7,309 | 12.84 | 21,357 | 19.0% | 43.3% |
| **2-gram** | Subword | 341 🏆 | 8.41 | 3,339 | 59.7% | 98.6% |
| **3-gram** | Word | 10,967 | 13.42 | 25,888 | 15.3% | 36.1% |
| **3-gram** | Subword | 2,785 | 11.44 | 26,207 | 23.8% | 67.7% |
| **4-gram** | Word | 23,875 | 14.54 | 45,756 | 10.4% | 26.4% |
| **4-gram** | Subword | 14,052 | 13.78 | 126,719 | 12.1% | 38.9% |
| **5-gram** | Word | 17,503 | 14.10 | 31,812 | 11.9% | 28.3% |
| **5-gram** | Subword | 42,696 | 15.38 | 295,741 | 7.7% | 26.0% |

### Top 5 N-grams by Size

**2-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `ha e` | 7,977 |
| 2 | `pegua ary` | 3,060 |
| 3 | `mba e` | 3,024 |
| 4 | `ary reñói` | 2,730 |
| 5 | `mandu apy` | 2,204 |

**3-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `ha e peteĩ` | 2,053 |
| 2 | `tetãvore joapykuéra pegua` | 1,816 |
| 3 | `pegua ary reñói` | 1,571 |
| 4 | `pegua ary omano` | 1,034 |
| 5 | `pegua ñemano ary` | 977 |

**4-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `tetã peteĩ reko amérikagua` | 864 |
| 2 | `peteĩ reko amérikagua pegua` | 827 |
| 3 | `tetãvore joapykuéra pegua ary` | 552 |
| 4 | `eapohára tetãvore joapykuéra pegua` | 387 |
| 5 | `mba eapohára tetãvore joapykuéra` | 387 |

**5-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `tetã peteĩ reko amérikagua pegua` | 824 |
| 2 | `mba eapohára tetãvore joapykuéra pegua` | 387 |
| 3 | `ojehechákuri árape 5 jasypateĩ ary` | 272 |
| 4 | `tetãvore joapykuéra pegua ary reñói` | 244 |
| 5 | `ára ohasa va erã opa` | 242 |

**2-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `a _` | 226,579 |
| 2 | `e _` | 129,090 |
| 3 | `h a` | 102,213 |
| 4 | `_ o` | 98,932 |
| 5 | `r a` | 97,275 |

**3-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `_ h a` | 57,720 |
| 2 | `h a _` | 49,670 |
| 3 | `g u a` | 45,557 |
| 4 | `v a _` | 39,267 |
| 5 | `r a _` | 33,034 |

**4-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `_ h a _` | 33,891 |
| 2 | `e g u a` | 18,031 |
| 3 | `g u a _` | 15,376 |
| 4 | `a _ h a` | 14,782 |
| 5 | `a r y _` | 13,812 |

**5-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `p e g u a` | 12,475 |
| 2 | `k u é r a` | 11,652 |
| 3 | `_ p e g u` | 11,448 |
| 4 | `_ p e t e` | 10,472 |
| 5 | `u é r a _` | 10,450 |


### Key Findings

- **Best Perplexity:** 2-gram (subword) with 341
- **Entropy Trend:** Decreases with larger n-grams (more predictable)
- **Coverage:** Top-1000 patterns cover ~26% 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 | 0.7677 | 1.703 | 5.04 | 105,382 | 23.2% |
| **1** | Subword | 0.8888 | 1.852 | 6.40 | 1,525 | 11.1% |
| **2** | Word | 0.2175 | 1.163 | 1.51 | 529,122 | 78.3% |
| **2** | Subword | 0.8410 | 1.791 | 5.29 | 9,756 | 15.9% |
| **3** | Word | 0.0735 | 1.052 | 1.13 | 794,049 | 92.7% |
| **3** | Subword | 0.8224 | 1.768 | 4.14 | 51,620 | 17.8% |
| **4** | Word | 0.0287 🏆 | 1.020 | 1.05 | 891,878 | 97.1% |
| **4** | Subword | 0.6549 | 1.575 | 2.78 | 213,613 | 34.5% |

### Generated Text Samples (Word-based)

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

**Context Size 1:**

1. `ha ombotuicha ha e kuéra ary omemby iména he i hũ kangy osapukái térã ambuéva chína`
2. `e hetãve hag̃ua paraguaýpe paraguay ii ha e ojapi ha uruguái ha mba apópe ko mbo`
3. `ary eddie izzard lewis cass ojokuaikuaáva uruguaigua maría trigueros haihára de paraguay tierra este...`

**Context Size 2:**

1. `ha e vaka ñemongakuaa ha mba apohára kuñanguéra tetãuáva upépe opu ãta umi artista uruguái chile ha`
2. `pegua ary reñói kami baterista hapõ pegua de la sombra la ciudad del este ypyetépe ha e`
3. `mba e ehechami rrúsia oñemomba e hag̃ua peteĩ ñemongeta periodístandi he i jey chupe ary jave ha`

**Context Size 3:**

1. `ha e peteĩ temiandu oreko mava jejapo ỹva mava omboaje ha oporangareko ambue tekove ombohovái peteĩ ...`
2. `tetãvore joapykuéra pegua ary reñói robert traylor baloncestista amérika retãvorekuéra joaju kuarahy...`
3. `pegua ary reñói émile michel cioran karai arandu nihilista rumáña pegua ary reñói mayía rodríguez mi...`

**Context Size 4:**

1. `tetã peteĩ reko amérikagua pegua takayuki morimoto vakapipopo ha ãhára japonés alexander ludwig acto...`
2. `peteĩ reko amérikagua pegua youri tielemans vakapipopo ha ãhára belga ary reñói joaquín capilla clav...`
3. `tetãvore joapykuéra pegua ary reñói josé pimentel llerenas líder sindical méhiko pegua ary reñói bed...`


### Generated Text Samples (Subword-based)

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

**Context Size 1:**

1. `_tia_n_hajorpix_`
2. `a’eoñe_áisévoki_`
3. `ed_spõgéruendach`

**Context Size 2:**

1. `a_oipoytépeguastr`
2. `e_po_frikatépegui`
3. `ha_urikaty_cubla_`

**Context Size 3:**

1. `_ha_ne_ã_upéa_esta`
2. `ha_yuri_imba'eha_o`
3. `guasu,_juan_crisab`

**Context Size 4:**

1. `_ha_ndaikatu_hectác`
2. `egua-pe_ha_ha_ja'ui`
3. `gua_(ñe’ẽmegua,_hen`


### Key Findings

- **Best Predictability:** Context-4 (word) with 97.1% predictability
- **Branching Factor:** Decreases with context size (more deterministic)
- **Memory Trade-off:** Larger contexts require more storage (213,613 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 | 43,448 |
| Total Tokens | 966,378 |
| Mean Frequency | 22.24 |
| Median Frequency | 3 |
| Frequency Std Dev | 299.53 |

### Most Common Words

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | ha | 46,095 |
| 2 | e | 14,500 |
| 3 | ary | 14,366 |
| 4 | de | 12,762 |
| 5 | pegua | 11,407 |
| 6 | pe | 9,844 |
| 7 | mba | 9,415 |
| 8 | ko | 8,744 |
| 9 | peteĩ | 8,686 |
| 10 | umi | 8,281 |

### Least Common Words (from vocabulary)

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | músika | 2 |
| 2 | jokohakue | 2 |
| 3 | oytúvre | 2 |
| 4 | monoꞌõ | 2 |
| 5 | konkúrso | 2 |
| 6 | kayꞌuhápe | 2 |
| 7 | rekoporã | 2 |
| 8 | vérso | 2 |
| 9 | juhujey | 2 |
| 10 | oñemoñeꞌẽpoty | 2 |

### Zipf's Law Analysis

| Metric | Value |
|--------|-------|
| Zipf Coefficient | 1.0723 |
| R² (Goodness of Fit) | 0.996343 |
| Adherence Quality | **excellent** |

### Coverage Analysis

| Top N Words | Coverage |
|-------------|----------|
| Top 100 | 35.4% |
| Top 1,000 | 63.8% |
| Top 5,000 | 81.6% |
| Top 10,000 | 88.3% |

### Key Findings

- **Zipf Compliance:** R²=0.9963 indicates excellent adherence to Zipf's law
- **High Frequency Dominance:** Top 100 words cover 35.4% of corpus
- **Long Tail:** 33,448 words needed for remaining 11.7% 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 | 0.8633 🏆 | 0.3274 | N/A | N/A |
| **mono_64d** | 64 | 0.8216 | 0.2580 | N/A | N/A |
| **mono_128d** | 128 | 0.5389 | 0.2262 | N/A | N/A |
| **aligned_32d** | 32 | 0.8633 | 0.3251 | 0.0680 | 0.2820 |
| **aligned_64d** | 64 | 0.8216 | 0.2581 | 0.0660 | 0.3620 |
| **aligned_128d** | 128 | 0.5389 | 0.2204 | 0.1580 | 0.4560 |

### Key Findings

- **Best Isotropy:** mono_32d with 0.8633 (more uniform distribution)
- **Semantic Density:** Average pairwise similarity of 0.2692. Lower values indicate better semantic separation.
- **Alignment Quality:** Aligned models achieve up to 15.8% 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 | **0.091** | 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 |
|--------|----------|
| `-oj` | ojehecharamoite, ojeguerahava, ojapose |
| `-oñ` | oñemohendárõguare, oñemongakuaáva, oñemboguapýkuri |
| `-oñe` | oñemohendárõguare, oñemongakuaáva, oñemboguapýkuri |

#### Productive Suffixes
| Suffix | Examples |
|--------|----------|
| `-a` | evahína, larnaka, retãmegua |
| `-e` | uvekitãñe, rakãngue, siouxsie |
| `-va` | ojeguerahava, omoguahẽva, oitýva |
| `-pe` | jokuairapépe, nekomatape, kysepukúpe |
| `-ra` | oliveira, tembiasahára, quimera |
| `-ha` | ñemoha, iñaranduha, ijyvateha |

### 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 |
|------|----------|------------------|----------|
| `rand` | 2.01x | 65 contexts | randy, brand, grand |
| `hech` | 2.02x | 55 contexts | hecha, hecho, ohecha |
| `ñemb` | 1.97x | 54 contexts | ñembý, ñemba, ñemby |
| `oñem` | 1.94x | 47 contexts | oñemo, oñemu, oñema |
| `kuér` | 1.72x | 75 contexts | kuéra, kuére, okuéra |
| `guer` | 1.73x | 73 contexts | guero, guera, gueru |
| `guas` | 1.64x | 76 contexts | águas, aguas, guasu |
| `uéra` | 1.85x | 42 contexts | kuéra, okuéra, ũkuéra |
| `ragu` | 1.65x | 57 contexts | rague, aragua, prague |
| `pegu` | 1.81x | 39 contexts | pegua, pegue, peguaa |
| `guar` | 1.63x | 59 contexts | guarã, guare, guara |
| `asyp` | 2.67x | 11 contexts | asypo, rasypa, jasypo |

### 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 |
|--------|--------|-----------|----------|
| `-oj` | `-a` | 78 words | ojereha, ojapóha |
| `-oñ` | `-a` | 67 words | oñembokuatiáva, oñemoporãva |
| `-oj` | `-va` | 45 words | ojapokuaáva, ojehechava |
| `-oñ` | `-va` | 36 words | oñembokuatiáva, oñemoporãva |
| `-oj` | `-e` | 27 words | ojejerure, ojelee |
| `-oñ` | `-e` | 26 words | oñombohovakérõguare, oñepyrũvaꞌekue |
| `-oj` | `-ha` | 15 words | ojereha, ojapóha |
| `-oñ` | `-ha` | 7 words | oñemondeháicha, oñemoambuéicha |
| `-oj` | `-pe` | 6 words | ojeipuruhápe, ojapohaguépe |
| `-oñ` | `-pe` | 6 words | oñemohendahápe, oñesãmbyhyhápe |

### 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 |
|------|-----------------|------------|------|
| peteĩhape | **`peteĩ-ha-pe`** | 6.0 | `peteĩ` |
| ojeguerekoha | **`oj-eguereko-ha`** | 6.0 | `eguereko` |
| ikatutaha | **`ikatuta-ha`** | 4.5 | `ikatuta` |
| amérikape | **`amérika-pe`** | 4.5 | `amérika` |
| posadaspe | **`posadas-pe`** | 4.5 | `posadas` |
| oñeñorairõ | **`oñe-ñorairõ`** | 4.5 | `ñorairõ` |
| áuteriape | **`áuteria-pe`** | 4.5 | `áuteria` |
| malvinape | **`malvina-pe`** | 4.5 | `malvina` |
| hekomarãva | **`hekomarã-va`** | 4.5 | `hekomarã` |
| ojopokóvo | **`oj-opokóvo`** | 4.5 | `opokóvo` |
| encarnaciónpe | **`encarnación-pe`** | 4.5 | `encarnación` |
| oñeñembosarái | **`oñe-ñembosarái`** | 4.5 | `ñembosarái` |
| arahentínape | **`arahentína-pe`** | 4.5 | `arahentína` |
| ijyvateha | **`ijyvate-ha`** | 4.5 | `ijyvate` |
| ojegueraha | **`oj-egue-ra-ha`** | 4.5 | `egue` |

### 6.6 Linguistic Interpretation

> **Automated Insight:**
The language Guarani 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 (4.36x) |
| N-gram | **2-gram** | Lowest perplexity (341) |
| Markov | **Context-4** | Highest predictability (97.1%) |
| 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](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: 2026-01-04 15:26:15*