README.md

---
language: tn
language_name: Tswana
language_family: bantu_southern
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-bantu_southern
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: 4.812
  - name: best_isotropy
    type: isotropy
    value: 0.8424
  - name: vocabulary_size
    type: vocab
    value: 0
generated: 2026-01-11
---

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

This repository contains NLP models trained and evaluated by Wikilangs, specifically on **Tswana** 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** | 4.418x | 4.42 | 0.0556% | 737,223 |
| **16k** | 4.593x | 4.59 | 0.0578% | 709,175 |
| **32k** | 4.727x | 4.73 | 0.0595% | 689,022 |
| **64k** | 4.812x 🏆 | 4.81 | 0.0606% | 676,881 |

### Tokenization Examples

Below are sample sentences tokenized with each vocabulary size:

**Sample 1:** `Need for Speed (NFS) ke motshameko wa motshikinyego o go thomiwang o o dirilweng...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁need ▁for ▁spe ed ▁( nf s ) ▁ke ▁motshameko ... (+22 more)` | 32 |
| 16k | `▁need ▁for ▁spe ed ▁( nf s ) ▁ke ▁motshameko ... (+19 more)` | 29 |
| 32k | `▁need ▁for ▁spe ed ▁( nf s ) ▁ke ▁motshameko ... (+19 more)` | 29 |
| 64k | `▁need ▁for ▁speed ▁( nf s ) ▁ke ▁motshameko ▁wa ... (+17 more)` | 27 |

**Sample 2:** `Bekkersdal ke toropo ya Gauteng e ko lefatsheng la Aforika Borwa. Metswedi`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁be k kers dal ▁ke ▁toropo ▁ya ▁gauteng ▁e ▁ko ... (+6 more)` | 16 |
| 16k | `▁be k kers dal ▁ke ▁toropo ▁ya ▁gauteng ▁e ▁ko ... (+6 more)` | 16 |
| 32k | `▁be k kers dal ▁ke ▁toropo ▁ya ▁gauteng ▁e ▁ko ... (+6 more)` | 16 |
| 64k | `▁bekkersdal ▁ke ▁toropo ▁ya ▁gauteng ▁e ▁ko ▁lefatsheng ▁la ▁aforika ... (+3 more)` | 13 |

**Sample 3:** `Osaka ke toropo kgolo kwa Japan. E na le baagi ba le`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁o saka ▁ke ▁toropo ▁kgolo ▁kwa ▁japan . ▁e ▁na ... (+4 more)` | 14 |
| 16k | `▁o saka ▁ke ▁toropo ▁kgolo ▁kwa ▁japan . ▁e ▁na ... (+4 more)` | 14 |
| 32k | `▁osaka ▁ke ▁toropo ▁kgolo ▁kwa ▁japan . ▁e ▁na ▁le ... (+3 more)` | 13 |
| 64k | `▁osaka ▁ke ▁toropo ▁kgolo ▁kwa ▁japan . ▁e ▁na ▁le ... (+3 more)` | 13 |


### Key Findings

- **Best Compression:** 64k achieves 4.812x compression
- **Lowest UNK Rate:** 8k with 0.0556% 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 | 7,155 | 12.80 | 61,361 | 28.5% | 48.8% |
| **2-gram** | Subword | 191 🏆 | 7.58 | 3,179 | 76.4% | 99.6% |
| **3-gram** | Word | 14,210 | 13.79 | 120,191 | 25.9% | 38.6% |
| **3-gram** | Subword | 1,323 | 10.37 | 26,297 | 38.5% | 81.3% |
| **4-gram** | Word | 23,873 | 14.54 | 216,515 | 24.9% | 33.3% |
| **4-gram** | Subword | 6,088 | 12.57 | 134,442 | 22.1% | 55.7% |
| **5-gram** | Word | 10,743 | 13.39 | 157,061 | 32.2% | 39.1% |
| **5-gram** | Subword | 18,500 | 14.18 | 344,305 | 15.2% | 39.9% |

### Top 5 N-grams by Size

**2-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `aforika borwa` | 32,436 |
| 2 | `toropo ya` | 30,077 |
| 3 | `ke toropo` | 29,904 |
| 4 | `ya gauteng` | 29,770 |
| 5 | `gauteng e` | 29,736 |

**3-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `ke toropo ya` | 29,751 |
| 2 | `ya gauteng e` | 29,733 |
| 3 | `gauteng e aforika` | 29,718 |
| 4 | `toropo ya gauteng` | 29,718 |
| 5 | `e aforika borwa` | 29,717 |

**4-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `ya gauteng e aforika` | 29,718 |
| 2 | `gauteng e aforika borwa` | 29,717 |
| 3 | `ke toropo ya gauteng` | 29,716 |
| 4 | `toropo ya gauteng e` | 29,716 |
| 5 | `mamelodi ke toropo ya` | 29,700 |

**5-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `ya gauteng e aforika borwa` | 29,717 |
| 2 | `ke toropo ya gauteng e` | 29,714 |
| 3 | `toropo ya gauteng e aforika` | 29,706 |
| 4 | `mamelodi ke toropo ya gauteng` | 29,700 |
| 5 | `borwa mamelodi ke toropo ya` | 29,699 |

**2-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `a _` | 935,894 |
| 2 | `e _` | 661,328 |
| 3 | `o _` | 427,244 |
| 4 | `l e` | 283,587 |
| 5 | `_ m` | 267,742 |

**3-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `_ l e` | 169,796 |
| 2 | `l e _` | 163,890 |
| 3 | `n g _` | 148,572 |
| 4 | `w a _` | 147,301 |
| 5 | `y a _` | 133,144 |

**4-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `_ y a _` | 122,807 |
| 2 | `_ l e _` | 121,639 |
| 3 | `e n g _` | 86,110 |
| 4 | `_ g o _` | 81,757 |
| 5 | `a _ b o` | 80,508 |

**5-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `o _ y a _` | 65,761 |
| 2 | `_ y a _ g` | 42,726 |
| 3 | `_ k w a _` | 39,822 |
| 4 | `a _ g o _` | 37,584 |
| 5 | `k a _ b o` | 37,508 |


### Key Findings

- **Best Perplexity:** 2-gram (subword) with 191
- **Entropy Trend:** Decreases with larger n-grams (more predictable)
- **Coverage:** Top-1000 patterns cover ~40% 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 | 0.9132 | 1.883 | 6.97 | 102,696 | 8.7% |
| **1** | Subword | 1.0155 | 2.022 | 7.94 | 975 | 0.0% |
| **2** | Word | 0.3523 | 1.277 | 2.10 | 714,400 | 64.8% |
| **2** | Subword | 0.9918 | 1.989 | 6.26 | 7,740 | 0.8% |
| **3** | Word | 0.1700 | 1.125 | 1.38 | 1,497,396 | 83.0% |
| **3** | Subword | 0.9000 | 1.866 | 4.58 | 48,443 | 10.0% |
| **4** | Word | 0.0886 🏆 | 1.063 | 1.16 | 2,060,334 | 91.1% |
| **4** | Subword | 0.6744 | 1.596 | 2.97 | 221,611 | 32.6% |

### Generated Text Samples (Word-based)

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

**Context Size 1:**

1. `ya ntlha wa citylife ka beilby porteus bishopo wa batjho ba ba amegang mo melawaneng ya`
2. `le balatedi bale mo dipolelong tsa itsholelo le tlhaeletsano pula botswana e diragalang bonnyane le ...`
3. `e aforika borwa mamelodi ke marang rang a le 357 quoting from the namibian via africabib`

**Context Size 2:**

1. `aforika borwa mamelodi ke toropo ya gauteng e aforika borwa mamelodi ke toropo ya gauteng e aforika`
2. `toropo ya gauteng e aforika borwa e tshwenyegile ka ditlamorago tse di nnang kwa kgaolong ya kweneng`
3. `ke toropo ya gauteng e aforika borwa mamelodi ke toropo ya gauteng e aforika borwa mamelodi ke`

**Context Size 3:**

1. `ke toropo ya gauteng e aforika borwa mamelodi ke toropo ya gauteng e aforika borwa mamelodi ke torop...`
2. `ya gauteng e aforika borwa mamelodi ke toropo ya gauteng e aforika borwa mamelodi ke toropo ya gaute...`
3. `toropo ya gauteng e aforika borwa mamelodi ke toropo ya gauteng e aforika borwa mamelodi ke toropo y...`

**Context Size 4:**

1. `ya gauteng e aforika borwa mamelodi ke toropo ya gauteng e aforika borwa mamelodi ke toropo ya gaute...`
2. `toropo ya gauteng e aforika borwa mamelodi ke toropo ya gauteng e aforika borwa mamelodi ke toropo y...`
3. `ke toropo ya gauteng e aforika borwa mamelodi ke toropo ya gauteng e aforika borwa mamelodi ke torop...`


### Generated Text Samples (Subword-based)

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

**Context Size 1:**

1. `_tlanga_ssatllhe`
2. `asophopolotlesha`
3. `eg,_ne_kgipave_d`

**Context Size 2:**

1. `a_mo_tlhabews_fet`
2. `e_neiratse_le_e_k`
3. `o_tekgo_ke_e_bof_`

**Context Size 3:**

1. `_le_e_a_nna_e_tor_`
2. `le_dipape_fa_tswa_`
3. `ng_e_aforika_di_mo`

**Context Size 4:**

1. `_ya_borwa._mamelodi`
2. `_le_mme_a_bonakgoba`
3. `eng_of_ethiopia_re,`


### Key Findings

- **Best Predictability:** Context-4 (word) with 91.1% predictability
- **Branching Factor:** Decreases with context size (more deterministic)
- **Memory Trade-off:** Larger contexts require more storage (221,611 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 | 51,001 |
| Total Tokens | 3,021,722 |
| Mean Frequency | 59.25 |
| Median Frequency | 4 |
| Frequency Std Dev | 1394.66 |

### Most Common Words

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | ya | 122,910 |
| 2 | le | 122,280 |
| 3 | e | 120,451 |
| 4 | a | 105,517 |
| 5 | go | 82,599 |
| 6 | ka | 70,434 |
| 7 | ba | 60,026 |
| 8 | ne | 54,685 |
| 9 | o | 51,263 |
| 10 | ke | 50,884 |

### Least Common Words (from vocabulary)

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | komit | 2 |
| 2 | duduzane | 2 |
| 3 | marčetić | 2 |
| 4 | prijedor | 2 |
| 5 | dnevne | 2 |
| 6 | novosti | 2 |
| 7 | greifenseelauf | 2 |
| 8 | makithing | 2 |
| 9 | benet | 2 |
| 10 | linnen | 2 |

### Zipf's Law Analysis

| Metric | Value |
|--------|-------|
| Zipf Coefficient | 1.1378 |
| R² (Goodness of Fit) | 0.995228 |
| Adherence Quality | **excellent** |

### Coverage Analysis

| Top N Words | Coverage |
|-------------|----------|
| Top 100 | 57.2% |
| Top 1,000 | 76.6% |
| Top 5,000 | 89.3% |
| Top 10,000 | 93.6% |

### Key Findings

- **Zipf Compliance:** R²=0.9952 indicates excellent adherence to Zipf's law
- **High Frequency Dominance:** Top 100 words cover 57.2% of corpus
- **Long Tail:** 41,001 words needed for remaining 6.4% 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 | 0.8424 | 0.3285 | N/A | N/A |
| **mono_64d** | 64 | 0.8282 | 0.2689 | N/A | N/A |
| **mono_128d** | 128 | 0.7325 | 0.2225 | N/A | N/A |
| **aligned_32d** | 32 | 0.8424 🏆 | 0.3391 | 0.0640 | 0.3560 |
| **aligned_64d** | 64 | 0.8282 | 0.2702 | 0.1760 | 0.5100 |
| **aligned_128d** | 128 | 0.7325 | 0.2209 | 0.2840 | 0.6440 |

### Key Findings

- **Best Isotropy:** aligned_32d with 0.8424 (more uniform distribution)
- **Semantic Density:** Average pairwise similarity of 0.2751. Lower values indicate better semantic separation.
- **Alignment Quality:** Aligned models achieve up to 28.4% 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 | **-0.020** | 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 |
|--------|----------|
| `-ma` | marcia, mahlatse, magudumana |
| `-m` | moinjineere, marcia, membrane |
| `-s` | sejaneng, still, stratification |
| `-b` | bontshiwang, busiwa, bongz |
| `-a` | adaptations, ausi, aug |
| `-di` | diitsholelo, distinguished, dikhwaere |
| `-mo` | moinjineere, motlabogi, monkeybone |
| `-t` | thapisitsweng, thakanyo, tedx |

#### Productive Suffixes
| Suffix | Examples |
|--------|----------|
| `-e` | christine, ratilwe, legotlhe |
| `-ng` | sejaneng, thapisitsweng, bontshiwang |
| `-a` | otjozondjupa, zuma, marcia |
| `-g` | rosberg, sejaneng, thapisitsweng |
| `-s` | vermeers, adaptations, focuses |
| `-o` | diitsholelo, phatlalatso, thakanyo |
| `-n` | zeaxanthin, stratification, defection |
| `-i` | shwahili, ausi, cpi |

### 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 |
|------|----------|------------------|----------|
| `tion` | 2.63x | 39 contexts | action, motion, notion |
| `tsen` | 2.13x | 60 contexts | tseno, tsene, tsena |
| `tlho` | 1.79x | 96 contexts | tlhoa, tlhopo, tlhora |
| `tshw` | 2.08x | 46 contexts | tshwa, ntshwa, tshweu |
| `otlh` | 1.78x | 67 contexts | otlhe, yotlhe, sotlhe |
| `tshe` | 1.76x | 68 contexts | ntshe, tsheko, tshele |
| `lhop` | 2.30x | 24 contexts | tlhopo, tlhopa, tlhopha |
| `otsw` | 1.86x | 43 contexts | otswa, rotswe, motswe |
| `hoph` | 2.25x | 21 contexts | tlhopha, tlhopho, tlhophe |
| `mets` | 1.81x | 43 contexts | metso, metsi, metse |
| `wana` | 1.98x | 30 contexts | swana, mowana, ntwana |
| `gwag` | 2.28x | 18 contexts | ngwag, gwaga, ngwago |

### 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 |
|--------|--------|-----------|----------|
| `-t` | `-g` | 121 words | tlhodileng, tlileng |
| `-t` | `-ng` | 120 words | tlhodileng, tlileng |
| `-t` | `-a` | 111 words | tshwaetswa, tsenngwa |
| `-t` | `-e` | 108 words | takirambudde, togolese |
| `-s` | `-e` | 95 words | setswerre, segololwane |
| `-b` | `-i` | 93 words | bogasi, bukhari |
| `-b` | `-e` | 90 words | blaze, banyamulenge |
| `-di` | `-o` | 84 words | ditshenolo, dikago |
| `-b` | `-g` | 83 words | benefitting, buang |
| `-b` | `-ng` | 81 words | benefitting, buang |

### 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 |
|------|-----------------|------------|------|
| botshepegi | **`botshepe-g-i`** | 7.5 | `g` |
| kgatlhego | **`kgatlhe-g-o`** | 7.5 | `g` |
| prehistoric | **`p-re-historic`** | 7.5 | `historic` |
| watergate | **`water-ga-te`** | 7.5 | `ga` |
| eletsegang | **`eletseg-a-ng`** | 7.5 | `a` |
| malahlela | **`malah-le-la`** | 7.5 | `le` |
| botswanago | **`botswana-g-o`** | 7.5 | `g` |
| ditlhagala | **`ditlhag-a-la`** | 7.5 | `a` |
| motshidisi | **`motshi-di-si`** | 7.5 | `di` |
| bosimegeng | **`bosimeg-e-ng`** | 7.5 | `e` |
| northeast | **`northea-s-t`** | 7.5 | `s` |
| diphethogo | **`diphetho-g-o`** | 7.5 | `g` |
| rwandaise | **`rwanda-i-se`** | 7.5 | `i` |
| utlwaleng | **`utlwa-le-ng`** | 7.5 | `le` |
| kgatlhile | **`kgatlh-i-le`** | 7.5 | `i` |

### 6.6 Linguistic Interpretation

> **Automated Insight:**
The language Tswana 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 (4.81x) |
| N-gram | **2-gram** | Lowest perplexity (191) |
| Markov | **Context-4** | Highest predictability (91.1%) |
| 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: 2026-01-11 01:22:30*