README.md

---
language: nv
language_name: Navajo
language_family: american_athabaskan
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_athabaskan
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: 3.722
  - name: best_isotropy
    type: isotropy
    value: 0.7658
  - name: vocabulary_size
    type: vocab
    value: 0
generated: 2026-01-10
---

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

This repository contains NLP models trained and evaluated by Wikilangs, specifically on **Navajo** 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** | 3.313x | 3.32 | 0.7428% | 222,258 |
| **16k** | 3.483x | 3.49 | 0.7810% | 211,391 |
| **32k** | 3.612x | 3.62 | 0.8101% | 203,814 |
| **64k** | 3.722x 🏆 | 3.73 | 0.8346% | 197,818 |

### Tokenization Examples

Below are sample sentences tokenized with each vocabulary size:

**Sample 1:** `Tółání Kʼish Chʼínítʼiʼ Tsé Chʼééchiiʼ yishtłizhii`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁tółání ▁k ʼ ish ▁ch ʼ ínít ʼ i ʼ ... (+7 more)` | 17 |
| 16k | `▁tółání ▁k ʼ ish ▁ch ʼ ínít ʼ i ʼ ... (+6 more)` | 16 |
| 32k | `▁tółání ▁k ʼ ish ▁ch ʼ ínít ʼ i ʼ ... (+6 more)` | 16 |
| 64k | `▁tółání ▁k ʼ ish ▁ch ʼ ínít ʼ i ʼ ... (+6 more)` | 16 |

**Sample 2:** `Naakaii Dootłʼizhii Bikéyahdę́ę́ʼ lókʼaatah naaʼahóóhai Tsiiʼyishbizhí Dineʼé Bi...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁naakaii ▁dootł ʼ izhii ▁bikéyahdę́ę́ ʼ ▁lók ʼ aatah ▁naa ... (+16 more)` | 26 |
| 16k | `▁naakaii ▁dootł ʼ izhii ▁bikéyahdę́ę́ ʼ ▁lók ʼ aatah ▁naa ... (+16 more)` | 26 |
| 32k | `▁naakaii ▁dootł ʼ izhii ▁bikéyahdę́ę́ ʼ ▁lók ʼ aatah ▁naa ... (+16 more)` | 26 |
| 64k | `▁naakaii ▁dootł ʼ izhii ▁bikéyahdę́ę́ ʼ ▁lók ʼ aatah ▁naa ... (+16 more)` | 26 |

**Sample 3:** `Azeeʼ haajinítsoh Azeeʼ haajinítsʼóóz Azeeʼ haajiní łibáhígíí`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁azee ʼ ▁haajiní tsoh ▁azee ʼ ▁haajiní ts ʼ óóz ... (+4 more)` | 14 |
| 16k | `▁azee ʼ ▁haajiní tsoh ▁azee ʼ ▁haajiní ts ʼ óóz ... (+4 more)` | 14 |
| 32k | `▁azee ʼ ▁haajinítsoh ▁azee ʼ ▁haajiní ts ʼ óóz ▁azee ... (+3 more)` | 13 |
| 64k | `▁azee ʼ ▁haajinítsoh ▁azee ʼ ▁haajiníts ʼ óóz ▁azee ʼ ... (+2 more)` | 12 |


### Key Findings

- **Best Compression:** 64k achieves 3.722x compression
- **Lowest UNK Rate:** 8k with 0.7428% 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 | 1,012 | 9.98 | 12,895 | 47.2% | 81.9% |
| **2-gram** | Subword | 222 🏆 | 7.79 | 1,668 | 72.2% | 99.8% |
| **3-gram** | Word | 2,466 | 11.27 | 30,460 | 36.6% | 67.1% |
| **3-gram** | Subword | 858 | 9.74 | 13,690 | 41.6% | 89.2% |
| **4-gram** | Word | 5,133 | 12.33 | 61,517 | 29.9% | 56.5% |
| **4-gram** | Subword | 1,964 | 10.94 | 55,169 | 29.2% | 77.2% |
| **5-gram** | Word | 7,471 | 12.87 | 67,722 | 25.5% | 51.1% |
| **5-gram** | Subword | 3,279 | 11.68 | 102,677 | 23.7% | 69.1% |

### Top 5 N-grams by Size

**2-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `ndaʼałkaahí dóó` | 18,966 |
| 2 | `dóó ééʼdeetįįhii` | 18,949 |
| 3 | `ééʼdeetįįhii éí` | 18,878 |
| 4 | `áádóó éí` | 18,437 |
| 5 | `dah yikahjí` | 18,133 |

**3-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `ndaʼałkaahí dóó ééʼdeetįįhii` | 18,948 |
| 2 | `dóó ééʼdeetįįhii éí` | 18,878 |
| 3 | `dah yikahjí atah` | 18,128 |
| 4 | `ánoolinígíí dóó bichʼiyąʼ` | 16,794 |
| 5 | `dóó bichʼiyąʼ díí` | 16,604 |

**4-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `ndaʼałkaahí dóó ééʼdeetįįhii éí` | 18,877 |
| 2 | `ánoolinígíí dóó bichʼiyąʼ díí` | 16,603 |
| 3 | `dah yikahjí atah yisdzoh` | 15,997 |
| 4 | `atah yisdzoh áádóó éí` | 13,441 |
| 5 | `yikahjí atah yisdzoh áádóó` | 13,428 |

**5-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `dah yikahjí atah yisdzoh áádóó` | 13,428 |
| 2 | `yikahjí atah yisdzoh áádóó éí` | 13,421 |
| 3 | `hólǫ́ ndaʼałkaahí dóó ééʼdeetįįhii éí` | 13,312 |
| 4 | `deiłníigo dayózhí ánoolinígíí dóó bichʼiyąʼ` | 12,295 |
| 5 | `dayózhí ánoolinígíí dóó bichʼiyąʼ díí` | 12,263 |

**2-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `í _` | 362,053 |
| 2 | `_ d` | 273,921 |
| 3 | `é í` | 184,110 |
| 4 | `_ é` | 173,881 |
| 5 | `_ b` | 173,418 |

**3-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `é í _` | 182,329 |
| 2 | `_ b i` | 160,761 |
| 3 | `_ é í` | 154,684 |
| 4 | `ó ó _` | 132,006 |
| 5 | `d ó ó` | 123,733 |

**4-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `_ é í _` | 154,592 |
| 2 | `d ó ó _` | 123,699 |
| 3 | `_ d ó ó` | 98,895 |
| 4 | `í g í í` | 52,301 |
| 5 | `g í í _` | 51,425 |

**5-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `_ d ó ó _` | 98,891 |
| 2 | `í g í í _` | 51,394 |
| 3 | `í _ d ó ó` | 48,361 |
| 4 | `i _ é í _` | 38,726 |
| 5 | `d ó ó _ é` | 38,444 |


### Key Findings

- **Best Perplexity:** 2-gram (subword) with 222
- **Entropy Trend:** Decreases with larger n-grams (more predictable)
- **Coverage:** Top-1000 patterns cover ~69% 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 | 0.5447 | 1.459 | 3.56 | 37,020 | 45.5% |
| **1** | Subword | 1.0994 | 2.143 | 8.42 | 395 | 0.0% |
| **2** | Word | 0.2649 | 1.202 | 1.82 | 130,895 | 73.5% |
| **2** | Subword | 1.0039 | 2.005 | 6.61 | 3,325 | 0.0% |
| **3** | Word | 0.1801 | 1.133 | 1.46 | 235,498 | 82.0% |
| **3** | Subword | 0.8364 | 1.786 | 3.94 | 21,977 | 16.4% |
| **4** | Word | 0.1277 🏆 | 1.093 | 1.29 | 339,354 | 87.2% |
| **4** | Subword | 0.5506 | 1.465 | 2.29 | 86,495 | 44.9% |

### Generated Text Samples (Word-based)

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

**Context Size 1:**

1. `éí łigai baʼáádígíí éí kéyah dah ndaaʼeełí łánídę́ę́ʼ tłʼiish dah yikahjí atah yisdzoh áádóó éí chʼi...`
2. `dóó chʼał dootłʼizhí bikédaayahdi tʼéiyá hólǫ́ ndaʼałkaahí dóó ééʼdeetįįhii éí diłhił shádiʼááh dóó ...`
3. `dah daalgai bitsiitsʼiin éí nahasdzáán tʼáá díkwíí mm áníłtso bitsʼíís éí yótʼáahdi tsídii tsídígíí ...`

**Context Size 2:**

1. `ndaʼałkaahí dóó ééʼdeetįįhii éí certhilauda benguelensis deiłníigo dayózhí ánoolinígíí dóó bichʼiyąʼ...`
2. `dóó ééʼdeetįįhii éí euscarthmus rufomarginatus deiłníigo dayózhí ánoolinígíí dóó bichʼiyąʼ díí naʼas...`
3. `ééʼdeetįįhii éí rhamphiophis oxyrhynchus deiłníigo dayózhí ánoolinígíí dóó bichʼiyąʼ díí tsídii biką...`

**Context Size 3:**

1. `ndaʼałkaahí dóó ééʼdeetįįhii éí dendropsophus koechlini deiłníigo dayózhí ánoolinígíí dóó bichʼiyąʼ ...`
2. `dóó ééʼdeetįįhii éí ptilopsis leucotis deiłníigo dayózhí ánoolinígíí dóó bichʼiyąʼ díí tłʼiish éí 30...`
3. `dah yikahjí atah yisdzoh áádóó éí naakaii łizhiní bikéyahdi hólǫ́ ndaʼałkaahí dóó ééʼdeetįįhii éí xe...`

**Context Size 4:**

1. `ndaʼałkaahí dóó ééʼdeetįįhii éí dendrolagus deiłníigo deiyózhí díí nahatʼeʼiitsoh éí 17 ałʼąą ádaatʼ...`
2. `ánoolinígíí dóó bichʼiyąʼ díí naʼashǫ́ʼii éí 4 5di asdzoh áníłtso bitsʼíís éí chʼilgo dootłʼizh bits...`
3. `dah yikahjí atah yisdzoh áádóó éí magí bitseeʼ noodǫ́zí bikéyahdi tʼéiyá hólǫ́ ndaʼałkaahí dóó ééʼde...`


### Generated Text Samples (Subword-based)

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

**Context Size 1:**

1. `_béíígaiy_"_yaʼ_`
2. `i_yąʼééí_ttą́._ée`
3. `í_éí_éí_tsh_áááʼ`

**Context Size 2:**

1. `í_dóó_atahdę́ę́ʼ_yę́`
2. `_dóó_bináhooly_oo`
3. `éí_bitoʼ_atah_yik`

**Context Size 3:**

1. `éí_naaʼałkaahí_éí_`
2. `_bitłʼaahjí_kélchí`
3. `_éí_naaznilzhin;_b`

**Context Size 4:**

1. `_éí_naashchʼąąʼ_éí_`
2. `dóó_éí_hólǫ́._ndaʼał`
3. `_dóó_ééʼdeetįįhii_é`


### Key Findings

- **Best Predictability:** Context-4 (word) with 87.2% predictability
- **Branching Factor:** Decreases with context size (more deterministic)
- **Memory Trade-off:** Larger contexts require more storage (86,495 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 | 15,109 |
| Total Tokens | 1,314,110 |
| Mean Frequency | 86.98 |
| Median Frequency | 4 |
| Frequency Std Dev | 1812.30 |

### Most Common Words

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | éí | 176,805 |
| 2 | dóó | 99,009 |
| 3 | dah | 28,837 |
| 4 | díí | 25,092 |
| 5 | bichʼiyąʼ | 23,153 |
| 6 | áádóó | 21,278 |
| 7 | ndaʼałkaahí | 19,035 |
| 8 | ééʼdeetįįhii | 18,949 |
| 9 | deiłníigo | 18,893 |
| 10 | atah | 18,728 |

### Least Common Words (from vocabulary)

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | milano | 2 |
| 2 | príncipe | 2 |
| 3 | butiama | 2 |
| 4 | àɖokun | 2 |
| 5 | yí | 2 |
| 6 | azɔ | 2 |
| 7 | àkpɔ̀ | 2 |
| 8 | gbɔ̀ | 2 |
| 9 | panafrikan | 2 |
| 10 | modèle | 2 |

### Zipf's Law Analysis

| Metric | Value |
|--------|-------|
| Zipf Coefficient | 1.3602 |
| R² (Goodness of Fit) | 0.987051 |
| Adherence Quality | **excellent** |

### Coverage Analysis

| Top N Words | Coverage |
|-------------|----------|
| Top 100 | 72.4% |
| Top 1,000 | 93.5% |
| Top 5,000 | 97.8% |
| Top 10,000 | 99.2% |

### Key Findings

- **Zipf Compliance:** R²=0.9871 indicates excellent adherence to Zipf's law
- **High Frequency Dominance:** Top 100 words cover 72.4% of corpus
- **Long Tail:** 5,109 words needed for remaining 0.8% 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 | 0.7658 🏆 | 0.3405 | N/A | N/A |
| **mono_64d** | 64 | 0.6030 | 0.2817 | N/A | N/A |
| **mono_128d** | 128 | 0.1964 | 0.2867 | N/A | N/A |
| **aligned_32d** | 32 | 0.7658 | 0.3269 | 0.0120 | 0.1440 |
| **aligned_64d** | 64 | 0.6030 | 0.2833 | 0.0280 | 0.2120 |
| **aligned_128d** | 128 | 0.1964 | 0.2859 | 0.0960 | 0.2700 |

### Key Findings

- **Best Isotropy:** mono_32d with 0.7658 (more uniform distribution)
- **Semantic Density:** Average pairwise similarity of 0.3008. Lower values indicate better semantic separation.
- **Alignment Quality:** Aligned models achieve up to 9.6% 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 | **-0.261** | 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` | allotment, amá, apodora |
| `-bi` | bikʼa, bichʼoshtsoh, bitsʼáozʼaʼ |
| `-d` | diastema, dryocalamus, deezlíníidi |
| `-b` | bílátaʼiitsóóh, bikʼa, bí |
| `-t` | tséhaagééd, tóńlį́, tʼiistsooítah |
| `-s` | sylvilagus, sturnira, sturnus |
| `-n` | natalobatrachus, neomixis, nahonitłʼahii |
| `-c` | certhiaxis, chʼiltaalzhahii, chʼahí |

#### Productive Suffixes
| Suffix | Examples |
|--------|----------|
| `-s` | himalayensis, sylvilagus, femoralis |
| `-us` | sylvilagus, dryocalamus, sturnus |
| `-í` | wálázhiní, bí, magítʼą́ʼí |
| `-i` | tséʼałnáoztʼiʼíidi, deezlíníidi, chʼiltaalzhahii |
| `-a` | sturnira, fuscicauda, bikʼa |
| `-is` | himalayensis, femoralis, ichthyophis |
| `-ii` | chʼiltaalzhahii, dáághahii, nahonitłʼahii |
| `-íí` | yeeyáʼdaałtíʼígíí, díkiwíí, dadijoolígíí |

### 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 |
|------|----------|------------------|----------|
| `ikah` | 2.28x | 8 contexts | yikahí, yikahji, yikahjí |
| `itsʼ` | 1.33x | 31 contexts | bitsʼáh, bitsʼoh, ditsʼoz |
| `tsʼí` | 1.63x | 14 contexts | tsʼídá, tsʼííh, tsʼímah |
| `éyah` | 1.67x | 13 contexts | kéyah, kéyahdi, hakéyah |
| `iłní` | 1.98x | 8 contexts | deiłní, nihiłní, ádeiłní |
| `sʼíí` | 1.87x | 9 contexts | tsʼííh, bitsʼíí, atsʼíís |
| `yika` | 2.28x | 5 contexts | yikał, yikahí, yikahji |
| `kahj` | 2.28x | 5 contexts | yikahji, yikahjí, daakahjí |
| `kéya` | 1.67x | 9 contexts | kéyah, kéyahdi, hakéyah |
| `níig` | 1.81x | 7 contexts | níigo, aníigo, aaníigo |
| `iníg` | 2.05x | 5 contexts | kinígíí, ádinígíí, nízinígíí |
| `bich` | 1.44x | 11 contexts | bichʼįʼ, bichąąʼ, bichʼil |

### 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` | `-s` | 249 words | chrysops, clematis |
| `-p` | `-s` | 243 words | platymantis, parvirostris |
| `-d` | `-í` | 213 words | dinilbáhí, dziłghą́ʼí |
| `-a` | `-s` | 184 words | arvalis, antrozous |
| `-n` | `-í` | 184 words | naalzheehígíí, naʼazísí |
| `-s` | `-s` | 156 words | sclerurus, scytodes |
| `-p` | `-us` | 138 words | perspicillatus, pteruthius |
| `-c` | `-us` | 131 words | castaneus, chroicocephalus |
| `-c` | `-a` | 126 words | crocata, cyanoleuca |
| `-t` | `-í` | 123 words | tłʼohtsʼózí, tłʼohwaaʼí |

### 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 |
|------|-----------------|------------|------|
| daʼałhosh | **`daʼałho-s-h`** | 7.5 | `s` |
| moluccensis | **`moluccen-s-is`** | 7.5 | `s` |
| daatsʼísí | **`daatsʼí-s-í`** | 7.5 | `s` |
| sminthopsis | **`sminthop-s-is`** | 7.5 | `s` |
| barbadensis | **`barbaden-s-is`** | 7.5 | `s` |
| chʼoshtsoh | **`chʼosht-s-oh`** | 7.5 | `s` |
| leucopsis | **`leucop-s-is`** | 7.5 | `s` |
| pretiosus | **`pretio-s-us`** | 7.5 | `s` |
| dlǫ́ʼiitsoh | **`dlǫ́ʼiit-s-oh`** | 7.5 | `s` |
| dinilzhinhgo | **`dinilzhin-h-go`** | 7.5 | `h` |
| mąʼiikʼǫsh | **`mąʼiikʼǫ-s-h`** | 7.5 | `s` |
| portoricensis | **`portoricen-s-is`** | 7.5 | `s` |
| natalensis | **`natalen-s-is`** | 7.5 | `s` |
| yildeełítsoh | **`yildeełít-s-oh`** | 7.5 | `s` |
| iichʼąhiitsʼósí | **`iichʼąhiitsʼó-s-í`** | 7.5 | `s` |

### 6.6 Linguistic Interpretation

> **Automated Insight:**
The language Navajo 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 (3.72x) |
| N-gram | **2-gram** | Lowest perplexity (222) |
| Markov | **Context-4** | Highest predictability (87.2%) |
| 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: 2026-01-10 16:24:15*