README.md

---
language: om
language_name: Oromo
language_family: cushitic
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-cushitic
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: 4.912
  - name: best_isotropy
    type: isotropy
    value: 0.8881
  - name: vocabulary_size
    type: vocab
    value: 0
generated: 2026-01-10
---

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

This repository contains NLP models trained and evaluated by Wikilangs, specifically on **Oromo** 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** | 3.925x | 3.93 | 0.7743% | 539,824 |
| **16k** | 4.303x | 4.30 | 0.8489% | 492,401 |
| **32k** | 4.614x | 4.62 | 0.9103% | 459,184 |
| **64k** | 4.912x 🏆 | 4.91 | 0.9692% | 431,282 |

### Tokenization Examples

Below are sample sentences tokenized with each vocabulary size:

**Sample 1:** `Giinii-Bisaa’uu biyya Afrikaa keessa jirtu. President: Umaró Umbaló Sissoko Prim...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁giinii - b isaa ’ uu ▁biyya ▁afrikaa ▁keessa ▁jirtu ... (+26 more)` | 36 |
| 16k | `▁giinii - b isaa ’ uu ▁biyya ▁afrikaa ▁keessa ▁jirtu ... (+21 more)` | 31 |
| 32k | `▁giinii - b isaa ’ uu ▁biyya ▁afrikaa ▁keessa ▁jirtu ... (+19 more)` | 29 |
| 64k | `▁giinii - bisaa ’ uu ▁biyya ▁afrikaa ▁keessa ▁jirtu . ... (+11 more)` | 21 |

**Sample 2:** `Godinni Qeellam Wallagaa kan argamu Oromiyaa keessatti. Wabii Oromiyaa Qellami w...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁godinni ▁q eell am ▁wall agaa ▁kan ▁argamu ▁oromiyaa ▁keessatti ... (+15 more)` | 25 |
| 16k | `▁godinni ▁qeell am ▁wall agaa ▁kan ▁argamu ▁oromiyaa ▁keessatti . ... (+13 more)` | 23 |
| 32k | `▁godinni ▁qeellam ▁wallagaa ▁kan ▁argamu ▁oromiyaa ▁keessatti . ▁wabii ▁oromiyaa ... (+8 more)` | 18 |
| 64k | `▁godinni ▁qeellam ▁wallagaa ▁kan ▁argamu ▁oromiyaa ▁keessatti . ▁wabii ▁oromiyaa ... (+6 more)` | 16 |

**Sample 3:** `Indoneeshiyaan biyya Eshiyaa bahaatti argamtu. Indoneeshiyaan odolotaarraa kan u...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁ind on eesh iyaan ▁biyya ▁eshiyaa ▁bahaatti ▁argamtu . ▁ind ... (+25 more)` | 35 |
| 16k | `▁indon eesh iyaan ▁biyya ▁eshiyaa ▁bahaatti ▁argamtu . ▁indon eesh ... (+22 more)` | 32 |
| 32k | `▁indoneeshiyaan ▁biyya ▁eshiyaa ▁bahaatti ▁argamtu . ▁indoneeshiyaan ▁odol ot aarraa ... (+18 more)` | 28 |
| 64k | `▁indoneeshiyaan ▁biyya ▁eshiyaa ▁bahaatti ▁argamtu . ▁indoneeshiyaan ▁odolotaarraa ▁kan ▁uummamte ... (+16 more)` | 26 |


### Key Findings

- **Best Compression:** 64k achieves 4.912x compression
- **Lowest UNK Rate:** 8k with 0.7743% 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 | 5,143 | 12.33 | 11,123 | 19.4% | 44.4% |
| **2-gram** | Subword | 213 🏆 | 7.73 | 2,915 | 73.1% | 99.4% |
| **3-gram** | Word | 6,773 | 12.73 | 11,742 | 15.4% | 34.2% |
| **3-gram** | Subword | 1,522 | 10.57 | 17,267 | 31.9% | 80.6% |
| **4-gram** | Word | 27,247 | 14.73 | 32,157 | 4.6% | 12.9% |
| **4-gram** | Subword | 7,489 | 12.87 | 73,751 | 15.1% | 48.6% |
| **5-gram** | Word | 24,982 | 14.61 | 27,382 | 3.0% | 11.0% |
| **5-gram** | Subword | 24,392 | 14.57 | 164,495 | 9.3% | 29.9% |

### Top 5 N-grams by Size

**2-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `ta u` | 3,055 |
| 2 | `yoo ta` | 2,546 |
| 3 | `ta e` | 1,182 |
| 4 | `danda a` | 874 |
| 5 | `kan akka` | 737 |

**3-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `yoo ta u` | 2,255 |
| 2 | `haa ta u` | 342 |
| 3 | `ta u malee` | 285 |
| 4 | `yeroo baay ee` | 252 |
| 5 | `of keessaa qaba` | 224 |

**4-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `haa ta u malee` | 282 |
| 2 | `yoo ta u kunis` | 213 |
| 3 | `qabu yoo ta u` | 167 |
| 4 | `ta uu danda a` | 135 |
| 5 | `tokko yoo ta u` | 128 |

**5-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `kan qabu yoo ta u` | 103 |
| 2 | `keessaa tokko yoo ta u` | 91 |
| 3 | `kan qaban yoo ta u` | 61 |
| 4 | `of keessaa qabu yoo ta` | 41 |
| 5 | `keessaa qabu yoo ta u` | 41 |

**2-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `a a` | 191,675 |
| 2 | `a _` | 157,464 |
| 3 | `a n` | 95,295 |
| 4 | `i _` | 91,344 |
| 5 | `n _` | 80,089 |

**3-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `a a _` | 71,357 |
| 2 | `a n _` | 40,890 |
| 3 | `a a n` | 34,146 |
| 4 | `i i _` | 34,015 |
| 5 | `t t i` | 25,977 |

**4-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `t t i _` | 21,400 |
| 2 | `a a n _` | 17,797 |
| 3 | `_ k a n` | 16,358 |
| 4 | `e e s s` | 15,993 |
| 5 | `a t t i` | 15,383 |

**5-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `a t t i _` | 13,078 |
| 2 | `e e s s a` | 12,757 |
| 3 | `_ k a n _` | 11,199 |
| 4 | `k e e s s` | 10,958 |
| 5 | `_ k e e s` | 10,805 |


### Key Findings

- **Best Perplexity:** 2-gram (subword) with 213
- **Entropy Trend:** Decreases with larger n-grams (more predictable)
- **Coverage:** Top-1000 patterns cover ~30% 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 | 0.7695 | 1.705 | 4.74 | 79,366 | 23.1% |
| **1** | Subword | 1.4423 | 2.718 | 10.26 | 768 | 0.0% |
| **2** | Word | 0.1820 | 1.134 | 1.36 | 375,403 | 81.8% |
| **2** | Subword | 0.8632 | 1.819 | 4.70 | 7,870 | 13.7% |
| **3** | Word | 0.0437 | 1.031 | 1.07 | 508,948 | 95.6% |
| **3** | Subword | 0.7482 | 1.680 | 3.52 | 36,941 | 25.2% |
| **4** | Word | 0.0138 🏆 | 1.010 | 1.02 | 541,591 | 98.6% |
| **4** | Subword | 0.5894 | 1.505 | 2.53 | 129,933 | 41.1% |

### Generated Text Samples (Word-based)

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

**Context Size 1:**

1. `fi niwutiroonii kan maatilee gurguddaa macroscopic anatomy fi wantoota qoyyobiyyoo hojjatu ta ee war...`
2. `kan maddu hedduu kan akka barachuu irra taxon dha innis dheerina ujummoo dhigaa ademma kunis yaada`
3. `ta uun leenjisaa yoo ta u guuttiiqoota poozatiiviidhaan chaarjii saffisaan hir isa sirrii irratti ka...`

**Context Size 2:**

1. `ta u malee siwaahiliin ummatta hanga million 40 guddatte haa ta u kan oromiyaa dabalatee heera mataa`
2. `yoo ta u xurbaa alfaa lamaa fi isaa ol walitti makamuun narvii lafee dugda of keessaa qabu`
3. `ta e garuu kofni isaa sirrii ta een fide namoonni jireenya godaansaa dhiisanii ganda dhaabbataa uumm...`

**Context Size 3:**

1. `yoo ta u hiikni isaanii kofa sirrii jechuudha rektaangiliin rogoonni isaa afran dheerinni isaa walfa...`
2. `haa ta u malee rakkoon haaraan dhufe cunqursaa lakkoofsaa tyranny of numbers naannee walxaxaa ijaaru...`
3. `ta u malee walii galtee dhabuun isaa fi birgaadeer jeneraal abdulkaariim qaasim gidduutti uumameen m...`

**Context Size 4:**

1. `haa ta u malee addeessi lafa irraa fagoo jira fageenyi inni lafarraa qabus kilomeetira 384 000 ol ta...`
2. `yoo ta u kunis dhuudhaa haaraa dhaabbatummaa saffisa ifaa fi dhuudhaa duraan beekamaa ture dhuudhaa ...`
3. `qabu yoo ta u yeroo baay ee meeshaalee bulchiinsaatiin kan qindaa anidha jijjiiramni adeemsa kana ke...`


### Generated Text Samples (Subword-based)

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

**Context Size 1:**

1. `a_awalmbarseeers`
2. `_ga_isufin_saan_`
3. `idalee_haaaammi_`

**Context Size 2:**

1. `aan._dhaniilixa_n`
2. `a_gamu_dhummaaloo`
3. `anii_dhaa_laawudh`

**Context Size 3:**

1. `aa_biran._see_itti`
2. `an_dina_maloojiiti`
3. `aane_galuuf_fakkee`

**Context Size 4:**

1. `tti_buuf_itti_oomis`
2. `aan_fakkasummataa_q`
3. `_kan_ariiroo_mancaa`


### Key Findings

- **Best Predictability:** Context-4 (word) with 98.6% predictability
- **Branching Factor:** Decreases with context size (more deterministic)
- **Memory Trade-off:** Larger contexts require more storage (129,933 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 | 33,399 |
| Total Tokens | 552,518 |
| Mean Frequency | 16.54 |
| Median Frequency | 3 |
| Frequency Std Dev | 147.85 |

### Most Common Words

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | fi | 12,823 |
| 2 | kan | 11,436 |
| 3 | ta | 8,265 |
| 4 | akka | 6,344 |
| 5 | keessatti | 5,232 |
| 6 | u | 4,498 |
| 7 | yoo | 4,135 |
| 8 | hin | 3,898 |
| 9 | yeroo | 3,580 |
| 10 | tokko | 3,377 |

### Least Common Words (from vocabulary)

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | prezidaantii | 2 |
| 2 | kasasiyoonaa | 2 |
| 3 | feltrinelli | 2 |
| 4 | rizaabii | 2 |
| 5 | raamyaan | 2 |
| 6 | rootarii | 2 |
| 7 | eegsa | 2 |
| 8 | alta | 2 |
| 9 | kuusii | 2 |
| 10 | maalamaa | 2 |

### Zipf's Law Analysis

| Metric | Value |
|--------|-------|
| Zipf Coefficient | 1.0202 |
| R² (Goodness of Fit) | 0.995340 |
| Adherence Quality | **excellent** |

### Coverage Analysis

| Top N Words | Coverage |
|-------------|----------|
| Top 100 | 32.2% |
| Top 1,000 | 59.7% |
| Top 5,000 | 79.9% |
| Top 10,000 | 87.7% |

### Key Findings

- **Zipf Compliance:** R²=0.9953 indicates excellent adherence to Zipf's law
- **High Frequency Dominance:** Top 100 words cover 32.2% of corpus
- **Long Tail:** 23,399 words needed for remaining 12.3% 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 | 0.8881 🏆 | 0.3121 | N/A | N/A |
| **mono_64d** | 64 | 0.7740 | 0.2857 | N/A | N/A |
| **mono_128d** | 128 | 0.1998 | 0.2448 | N/A | N/A |
| **aligned_32d** | 32 | 0.8881 | 0.3137 | 0.0140 | 0.1340 |
| **aligned_64d** | 64 | 0.7740 | 0.2922 | 0.0300 | 0.1800 |
| **aligned_128d** | 128 | 0.1998 | 0.2360 | 0.0660 | 0.2500 |

### Key Findings

- **Best Isotropy:** mono_32d with 0.8881 (more uniform distribution)
- **Semantic Density:** Average pairwise similarity of 0.2807. Lower values indicate better semantic separation.
- **Alignment Quality:** Aligned models achieve up to 6.6% 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 | **-0.633** | 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` | alex, amaaradhaan, ambaa |
| `-s` | silocones, soovieyeet, sexual |
| `-b` | beekamaafi, baatriin, balaaleffata |
| `-d` | duruma, document, dheedee |
| `-m` | milaa, moment, materials |
| `-ma` | materials, massachusetts, malcolm |
| `-ba` | baatriin, balaaleffata, baasti |
| `-g` | garraan, guutuu, guute |

#### Productive Suffixes
| Suffix | Examples |
|--------|----------|
| `-a` | milaa, qilleensarra, duruma |
| `-i` | beekamaafi, keetoonoonni, miiti |
| `-n` | tekinooloojiin, garraan, cimaadhaan |
| `-ii` | naqanii, giriikii, bismazii |
| `-aa` | milaa, wiifaa, inaariyaa |
| `-an` | garraan, cimaadhaan, firoottan |
| `-e` | caamee, kaaye, dheedee |
| `-ti` | miiti, baasti, totti |

### 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 |
|------|----------|------------------|----------|
| `araa` | 1.85x | 101 contexts | haraa, karaa, qaraa |
| `rraa` | 1.90x | 85 contexts | erraa, urraa, orraa |
| `mmaa` | 2.12x | 47 contexts | ummaa, ammaa, lummaa |
| `eess` | 1.68x | 124 contexts | eessa, keessa, eessoo |
| `gudd` | 2.12x | 34 contexts | guddo, gudda, guddae |
| `aala` | 1.58x | 114 contexts | yaala, jaala, gaala |
| `arga` | 1.75x | 66 contexts | marga, argan, argaa |
| `rrat` | 2.24x | 25 contexts | irrati, urratti, arratti |
| `okko` | 2.26x | 24 contexts | tokko, tokkof, tokkos |
| `ratt` | 1.90x | 45 contexts | iratti, barattu, biratti |
| `chuu` | 1.77x | 50 contexts | achuu, kichuu, glchuu |
| `jedh` | 2.23x | 20 contexts | jedhu, jedhe, jedha |

### 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 |
|--------|--------|-----------|----------|
| `-b` | `-a` | 163 words | balbala, biyyootaa |
| `-h` | `-a` | 142 words | hubachaa, harkisuudha |
| `-a` | `-a` | 141 words | arba, araboota |
| `-d` | `-a` | 137 words | dhooqa, dandeessa |
| `-a` | `-i` | 132 words | ashaabi, argamni |
| `-d` | `-i` | 129 words | dargageessi, dirreetti |
| `-m` | `-a` | 128 words | milliyoona, molekuloota |
| `-s` | `-a` | 121 words | shaffaaxa, shayxaana |
| `-d` | `-n` | 115 words | daawwannaan, dirreewwan |
| `-g` | `-a` | 110 words | genera, gamaa |

### 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 |
|------|-----------------|------------|------|
| burtukaanaa | **`burtuka-an-aa`** | 7.5 | `an` |
| funfannaa | **`funfan-n-aa`** | 7.5 | `n` |
| dhalattoonni | **`dhalattoon-n-i`** | 7.5 | `n` |
| aannannoo | **`aannan-n-oo`** | 7.5 | `n` |
| tambiixni | **`tambiix-n-i`** | 7.5 | `n` |
| fooyyessanii | **`fooyyess-an-ii`** | 7.5 | `an` |
| yaadoonni | **`yaadoon-n-i`** | 7.5 | `n` |
| academies | **`academ-i-es`** | 7.5 | `i` |
| karibiyaanii | **`karibiya-an-ii`** | 7.5 | `an` |
| enciclopèdia | **`enciclopèd-i-a`** | 7.5 | `i` |
| laakkawanii | **`laakkaw-an-ii`** | 7.5 | `an` |
| jijjiramni | **`jijjiram-n-i`** | 7.5 | `n` |
| raajeffannoo | **`raajeffan-n-oo`** | 7.5 | `n` |
| unionoromia | **`unionorom-i-a`** | 7.5 | `i` |
| raadiyoona | **`raadiyoo-n-a`** | 7.5 | `n` |

### 6.6 Linguistic Interpretation

> **Automated Insight:**
The language Oromo 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 (4.91x) |
| N-gram | **2-gram** | Lowest perplexity (213) |
| Markov | **Context-4** | Highest predictability (98.6%) |
| 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 16:39:48*