README.md

---
language: jam
language_name: Jamaican Creole English
language_family: germanic_west_anglofrisian
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-germanic_west_anglofrisian
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: 4.524
  - name: best_isotropy
    type: isotropy
    value: 0.1451
  - name: vocabulary_size
    type: vocab
    value: 0
generated: 2026-01-10
---

# Jamaican Creole English - Wikilangs Models
## Comprehensive Research Report & Full Ablation Study

This repository contains NLP models trained and evaluated by Wikilangs, specifically on **Jamaican Creole English** 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** | 3.852x | 3.86 | 0.1007% | 191,616 |
| **16k** | 4.204x | 4.21 | 0.1099% | 175,540 |
| **32k** | 4.524x 🏆 | 4.53 | 0.1183% | 163,136 |

### Tokenization Examples

Below are sample sentences tokenized with each vocabulary size:

**Sample 1:** `David Guetta (riil niem: Pierre David Guetta; baan 7 Novemba a Paris) a wah Fren...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁david ▁gu et ta ▁( ri il ▁niem : ▁pier ... (+26 more)` | 36 |
| 16k | `▁david ▁guetta ▁( riil ▁niem : ▁pierre ▁david ▁guetta ; ... (+18 more)` | 28 |
| 32k | `▁david ▁guetta ▁( riil ▁niem : ▁pierre ▁david ▁guetta ; ... (+18 more)` | 28 |

**Sample 2:** `AnuovaHannover 100px Kantinent YuuropNieshan JoermaniParish 204.14 km² Anuova (J...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁an uov ah ann over ▁ 1 0 0 px ... (+36 more)` | 46 |
| 16k | `▁an uov ah ann over ▁ 1 0 0 px ... (+34 more)` | 44 |
| 32k | `▁anuovahann over ▁ 1 0 0 px ▁kantinent ▁yuuropnieshan ▁joerman ... (+29 more)` | 39 |

**Sample 3:** `Jumiekan lichicha intanashinali rinoun, wid di ailan a Jumieka biin di uom ar bo...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁jumiekan ▁lichicha ▁intanashinali ▁rin oun , ▁wid ▁di ▁ailan ▁a ... (+12 more)` | 22 |
| 16k | `▁jumiekan ▁lichicha ▁intanashinali ▁rinoun , ▁wid ▁di ▁ailan ▁a ▁jumieka ... (+11 more)` | 21 |
| 32k | `▁jumiekan ▁lichicha ▁intanashinali ▁rinoun , ▁wid ▁di ▁ailan ▁a ▁jumieka ... (+11 more)` | 21 |


### Key Findings

- **Best Compression:** 32k achieves 4.524x compression
- **Lowest UNK Rate:** 8k with 0.1007% 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 | 1,541 | 10.59 | 3,741 | 32.5% | 65.9% |
| **2-gram** | Subword | 238 🏆 | 7.89 | 1,403 | 70.0% | 99.7% |
| **3-gram** | Word | 1,509 | 10.56 | 3,102 | 32.2% | 65.7% |
| **3-gram** | Subword | 1,861 | 10.86 | 9,633 | 27.4% | 74.4% |
| **4-gram** | Word | 1,686 | 10.72 | 4,165 | 32.5% | 55.5% |
| **4-gram** | Subword | 9,243 | 13.17 | 41,304 | 13.9% | 41.0% |
| **5-gram** | Word | 591 | 9.21 | 2,198 | 46.6% | 71.4% |
| **5-gram** | Subword | 25,412 | 14.63 | 84,144 | 8.7% | 26.9% |

### Top 5 N-grams by Size

**2-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `a di` | 2,702 |
| 2 | `ina di` | 1,423 |
| 3 | `tu di` | 748 |
| 4 | `a wah` | 541 |
| 5 | `ah di` | 470 |

**3-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `askaadn tu di` | 213 |
| 2 | `wan a di` | 194 |
| 3 | `tu di sensos` | 193 |
| 4 | `di pravins a` | 187 |
| 5 | `kiastiil ahn león` | 185 |

**4-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `askaadn tu di sensos` | 193 |
| 2 | `kiastiil ahn león spien` | 184 |
| 3 | `ina di pravins a` | 183 |
| 4 | `di pravins a soria` | 183 |
| 5 | `spien askaadn tu di` | 182 |

**5-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `ine di miunisipaliti ab papyulieshan` | 182 |
| 2 | `sensos ine di miunisipaliti ab` | 182 |
| 3 | `di sensos ine di miunisipaliti` | 182 |
| 4 | `tu di sensos ine di` | 182 |
| 5 | `askaadn tu di sensos ine` | 182 |

**2-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `_ a` | 29,684 |
| 2 | `a _` | 25,927 |
| 3 | `i _` | 25,474 |
| 4 | `a n` | 21,538 |
| 5 | `_ d` | 20,084 |

**3-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `_ d i` | 15,507 |
| 2 | `d i _` | 13,620 |
| 3 | `_ a _` | 10,852 |
| 4 | `a n _` | 8,698 |
| 5 | `a h _` | 7,964 |

**4-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `_ d i _` | 12,752 |
| 2 | `a _ d i` | 5,069 |
| 3 | `_ a h _` | 4,411 |
| 4 | `_ i n a` | 4,365 |
| 5 | `i n a _` | 4,360 |

**5-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `a _ d i _` | 4,702 |
| 2 | `_ i n a _` | 4,109 |
| 3 | `_ a _ d i` | 2,835 |
| 4 | `s h a n _` | 2,596 |
| 5 | `e s h a n` | 2,001 |


### Key Findings

- **Best Perplexity:** 2-gram (subword) with 238
- **Entropy Trend:** Decreases with larger n-grams (more predictable)
- **Coverage:** Top-1000 patterns cover ~27% 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 | 0.8259 | 1.773 | 4.61 | 23,902 | 17.4% |
| **1** | Subword | 0.9178 | 1.889 | 6.25 | 632 | 8.2% |
| **2** | Word | 0.2166 | 1.162 | 1.43 | 109,227 | 78.3% |
| **2** | Subword | 0.9098 | 1.879 | 5.02 | 3,949 | 9.0% |
| **3** | Word | 0.0581 | 1.041 | 1.09 | 155,360 | 94.2% |
| **3** | Subword | 0.8329 | 1.781 | 3.69 | 19,800 | 16.7% |
| **4** | Word | 0.0168 🏆 | 1.012 | 1.02 | 167,194 | 98.3% |
| **4** | Subword | 0.6241 | 1.541 | 2.48 | 72,952 | 37.6% |

### Generated Text Samples (Word-based)

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

**Context Size 1:**

1. `di tuu taip fi di standad tex ahn staavieshan ahn florida ina wol 93 6 october`
2. `a chriiti nachrali kaaz bai deh riyolajikal prapati raits gruup a di chanspuot infrachokcha we no`
3. `ah kom a review of america otherwise extoernal duona an ina piepal basilika a eni memba`

**Context Size 2:**

1. `a di 63 siit ina paaliment yet di riil sakratiiz laka nof languij elefen distingguish kountebl ah`
2. `ina di naat ahn lan pahn di kraas fi di buk we im du wehn put tigeda`
3. `tu di yuuman vais ina ar wok jinarali inten fi bi a kaman kuol ah ud ah`

**Context Size 3:**

1. `askaadn tu di sensos ine di miunisipaliti ab papyulieshan a 53 inabitant a soria category jaagrafi`
2. `wan a di yonggis mongx di mieja wol rilijan wid uoba 2 4 bilian adierent nuo az kristian`
3. `tu di sensos ine di miunisipaliti ab papyulieshan a 28 inabitant a soria category jaagrafi`

**Context Size 4:**

1. `askaadn tu di sensos di siti ab a papyulieshan a 17 865 piipl`
2. `kiastiil ahn león spien askaadn tu di sensos ine di miunisipaliti ab papyulieshan a 114 inabitant a ...`
3. `di pravins a soria kiastiil ahn león spien askaadn tu di sensos di toun ab a papyulieshan a 2`


### Generated Text Samples (Subword-based)

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

**Context Size 1:**

1. `_(miesorish_pren`
2. `aash_pr_seng_kop`
3. `ip_ti_ma_tatiuse`

**Context Size 2:**

1. `_a_a_fahn_kos_a_d`
2. `a_di_frailica"._w`
3. `i_menis_jaid_an_a`

**Context Size 3:**

1. `_di_bot_ar_i,_ada_`
2. `di_np_nof_amoert_p`
3. `_a_no_nuo_impuot_s`

**Context Size 4:**

1. `_di_kans,_by_nubia_`
2. `a_di_dieta_di_sophy`
3. `_ah_ab_tuul._founli`


### Key Findings

- **Best Predictability:** Context-4 (word) with 98.3% predictability
- **Branching Factor:** Decreases with context size (more deterministic)
- **Memory Trade-off:** Larger contexts require more storage (72,952 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 | 10,520 |
| Total Tokens | 170,163 |
| Mean Frequency | 16.18 |
| Median Frequency | 3 |
| Frequency Std Dev | 187.26 |

### Most Common Words

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | di | 13,145 |
| 2 | a | 11,091 |
| 3 | ah | 4,442 |
| 4 | ina | 4,225 |
| 5 | fi | 2,654 |
| 6 | we | 1,934 |
| 7 | tu | 1,838 |
| 8 | wah | 1,390 |
| 9 | ar | 1,371 |
| 10 | az | 1,170 |

### Least Common Words (from vocabulary)

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | turn | 2 |
| 2 | episode | 2 |
| 3 | clips | 2 |
| 4 | schaffer | 2 |
| 5 | politico | 2 |
| 6 | youtube | 2 |
| 7 | archived | 2 |
| 8 | viral | 2 |
| 9 | klein | 2 |
| 10 | cancel | 2 |

### Zipf's Law Analysis

| Metric | Value |
|--------|-------|
| Zipf Coefficient | 1.0629 |
| R² (Goodness of Fit) | 0.987155 |
| Adherence Quality | **excellent** |

### Coverage Analysis

| Top N Words | Coverage |
|-------------|----------|
| Top 100 | 44.1% |
| Top 1,000 | 72.2% |
| Top 5,000 | 92.3% |
| Top 10,000 | 99.4% |

### Key Findings

- **Zipf Compliance:** R²=0.9872 indicates excellent adherence to Zipf's law
- **High Frequency Dominance:** Top 100 words cover 44.1% of corpus
- **Long Tail:** 520 words needed for remaining 0.6% 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 | 0.1451 🏆 | 0.5442 | N/A | N/A |
| **mono_64d** | 64 | 0.0312 | 0.5648 | N/A | N/A |
| **mono_128d** | 128 | 0.0054 | 0.5708 | N/A | N/A |
| **aligned_32d** | 32 | 0.1451 | 0.5158 | 0.0080 | 0.0920 |
| **aligned_64d** | 64 | 0.0312 | 0.5492 | 0.0140 | 0.1180 |
| **aligned_128d** | 128 | 0.0054 | 0.5682 | 0.0200 | 0.1320 |

### Key Findings

- **Best Isotropy:** mono_32d with 0.1451 (more uniform distribution)
- **Semantic Density:** Average pairwise similarity of 0.5522. Lower values indicate better semantic separation.
- **Alignment Quality:** Aligned models achieve up to 2.0% 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 | **4.300** | High morphological productivity | Reliable analysis |
| Idiomaticity Gap | **1.654** | High formulaic/idiomatic content | - |

### 6.2 Affix Inventory (Productive Units)

These are the most productive prefixes and suffixes identified by sampling the vocabulary for global substitutability patterns. A unit is considered an affix if stripping it leaves a valid stem that appears in other contexts.

#### Productive Prefixes
| Prefix | Examples |
|--------|----------|
| `-a` | akses, amplifai, araival |
| `-s` | staat, savlamaar, skyaaboro |
| `-i` | injri, inschument, ivenchal |
| `-p` | pavati, park, platfaam |
| `-m` | mahtah, mendeleev, migl |
| `-k` | kraitiiria, konghwaguk, kori |
| `-b` | buush, bahá, bizniz |
| `-r` | rizol, romance, room |

#### Productive Suffixes
| Suffix | Examples |
|--------|----------|
| `-n` | nuon, chrienin, yuumankain |
| `-an` | riilizieshan, porjan, dipikshan |
| `-s` | akses, viskyuos, takes |
| `-i` | amplifai, pavati, injri |
| `-a` | kraitiiria, tunisia, kyaa |
| `-l` | nigril, araival, rizol |
| `-t` | edit, staat, inschument |
| `-al` | araival, tioretikal, ivenchal |

### 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 |
|------|----------|------------------|----------|
| `schr` | 1.42x | 26 contexts | aschro, ischri, schres |
| `chra` | 1.37x | 28 contexts | chrai, chrak, exchra |
| `iesh` | 1.51x | 18 contexts | iesha, riesho, ieshan |
| `ikal` | 1.45x | 17 contexts | maikal, etikal, fizikal |
| `ment` | 1.36x | 19 contexts | mento, kament, moment |
| `toer` | 1.40x | 17 contexts | toerx, toerd, toerm |
| `iiri` | 1.42x | 16 contexts | tiiri, siiriz, siiriiz |
| `tiet` | 1.46x | 14 contexts | stiet, tieta, sitiet |
| `riti` | 1.40x | 15 contexts | priti, eritij, kritik |
| `shal` | 1.33x | 17 contexts | shalo, shalom, speshal |
| `esha` | 1.45x | 13 contexts | iesha, presha, ieshan |
| `isti` | 1.47x | 12 contexts | istil, istiet, sistim |

### 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 |
|--------|--------|-----------|----------|
| `-k` | `-n` | 107 words | kaatuun, kansenchrieshan |
| `-a` | `-n` | 81 words | alkalain, aprishieshan |
| `-k` | `-an` | 80 words | kansenchrieshan, konfederashan |
| `-i` | `-n` | 76 words | ingkluudn, imiin |
| `-p` | `-n` | 74 words | puoshan, pakistan |
| `-s` | `-n` | 73 words | susan, siblizieshan |
| `-i` | `-t` | 68 words | intoerprit, ikuivilent |
| `-r` | `-n` | 67 words | remain, riikan |
| `-a` | `-i` | 57 words | aatobayagrafi, ali |
| `-a` | `-an` | 55 words | aprishieshan, aaran |

### 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 |
|------|-----------------|------------|------|
| kantinyuiti | **`kantinyu-i-ti`** | 7.5 | `i` |
| kritikdem | **`kritik-d-em`** | 7.5 | `d` |
| apastalik | **`apast-al-ik`** | 7.5 | `al` |
| aatimisinin | **`aatimis-in-in`** | 7.5 | `in` |
| signifikans | **`signifik-an-s`** | 7.5 | `an` |
| plietanik | **`pliet-an-ik`** | 7.5 | `an` |
| suitsalan | **`suits-al-an`** | 7.5 | `al` |
| inishitiv | **`inishi-t-iv`** | 7.5 | `t` |
| distingtiv | **`disting-t-iv`** | 7.5 | `t` |
| afrikaanz | **`afrika-an-z`** | 7.5 | `an` |
| yuuropiian | **`yuuropi-i-an`** | 7.5 | `i` |
| salamanik | **`salam-an-ik`** | 7.5 | `an` |
| ilekchisiti | **`ilekchis-i-ti`** | 7.5 | `i` |
| afrikanis | **`afrik-an-is`** | 7.5 | `an` |
| chadishanal | **`chadish-an-al`** | 7.5 | `an` |

### 6.6 Linguistic Interpretation

> **Automated Insight:**
The language Jamaican Creole English shows high morphological productivity. The subword models are significantly more efficient than word models, suggesting a rich system of affixation or compounding.

> **Note on Idiomaticity:** The high Idiomaticity Gap suggests a large number of frequent multi-word expressions or formulaic sequences that are statistically distinct from their component parts.

---
## 7. Summary & Recommendations

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

### Production Recommendations

| Component | Recommended | Rationale |
|-----------|-------------|-----------|
| Tokenizer | **32k BPE** | Best compression (4.52x) |
| N-gram | **2-gram** | Lowest perplexity (238) |
| Markov | **Context-4** | Highest predictability (98.3%) |
| 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 05:49:27*