README.md

---
language: kcg
language_name: Tyap
language_family: atlantic_other
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-atlantic_other
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: 4.834
  - name: best_isotropy
    type: isotropy
    value: 0.3873
  - name: vocabulary_size
    type: vocab
    value: 0
generated: 2026-01-10
---

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

This repository contains NLP models trained and evaluated by Wikilangs, specifically on **Tyap** 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** | 4.149x | 4.15 | 0.1551% | 192,111 |
| **16k** | 4.452x | 4.46 | 0.1664% | 179,058 |
| **32k** | 4.706x | 4.71 | 0.1760% | 169,365 |
| **64k** | 4.834x 🏆 | 4.84 | 0.1807% | 164,911 |

### Tokenization Examples

Below are sample sentences tokenized with each vocabulary size:

**Sample 1:** `Atanii yet mam hwa kunin kyak avwou mun tsatsak ladi mang talata . Wikimedians Z...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁at ani i ▁yet ▁mam ▁hwa ▁ku nin ▁kyak ▁avwo ... (+11 more)` | 21 |
| 16k | `▁atanii ▁yet ▁mam ▁hwa ▁ku nin ▁kyak ▁avwou ▁mun ▁tsatsak ... (+8 more)` | 18 |
| 32k | `▁atanii ▁yet ▁mam ▁hwa ▁kunin ▁kyak ▁avwou ▁mun ▁tsatsak ▁ladi ... (+6 more)` | 16 |
| 64k | `▁atanii ▁yet ▁mam ▁hwa ▁kunin ▁kyak ▁avwou ▁mun ▁tsatsak ▁ladi ... (+6 more)` | 16 |

**Sample 2:** `Zong (á̱ ka ndyuut zwong a̱ni) yet jen nang a̱yin nswan a̱fa a̱khwot di̱ mi̱n ya...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁zong ▁( á̱ ▁ka ▁ndyuut ▁z wong ▁a̱ni ) ▁yet ... (+19 more)` | 29 |
| 16k | `▁zong ▁( á̱ ▁ka ▁ndyuut ▁z wong ▁a̱ni ) ▁yet ... (+19 more)` | 29 |
| 32k | `▁zong ▁( á̱ ▁ka ▁ndyuut ▁z wong ▁a̱ni ) ▁yet ... (+19 more)` | 29 |
| 64k | `▁zong ▁( á̱ ▁ka ▁ndyuut ▁zwong ▁a̱ni ) ▁yet ▁jen ... (+18 more)` | 28 |

**Sample 3:** `Ci̱ncai yet a̱cyuang ga̱swan ba̱ ya ka̱tako a̱ni. Ya̱fang`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁c i̱n c ai ▁yet ▁a̱cyuang ▁ga̱s wan ▁ba̱ ▁ya ... (+5 more)` | 15 |
| 16k | `▁c i̱n c ai ▁yet ▁a̱cyuang ▁ga̱swan ▁ba̱ ▁ya ▁ka̱tak ... (+4 more)` | 14 |
| 32k | `▁ci̱ncai ▁yet ▁a̱cyuang ▁ga̱swan ▁ba̱ ▁ya ▁ka̱tako ▁a̱ni . ▁ya̱fang` | 10 |
| 64k | `▁ci̱ncai ▁yet ▁a̱cyuang ▁ga̱swan ▁ba̱ ▁ya ▁ka̱tako ▁a̱ni . ▁ya̱fang` | 10 |


### Key Findings

- **Best Compression:** 64k achieves 4.834x compression
- **Lowest UNK Rate:** 8k with 0.1551% 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 | 2,665 | 11.38 | 5,715 | 23.5% | 60.8% |
| **2-gram** | Subword | 265 🏆 | 8.05 | 1,919 | 66.6% | 99.3% |
| **3-gram** | Word | 3,873 | 11.92 | 6,453 | 18.1% | 48.8% |
| **3-gram** | Subword | 1,877 | 10.87 | 12,850 | 30.0% | 74.6% |
| **4-gram** | Word | 6,271 | 12.61 | 8,735 | 12.5% | 36.0% |
| **4-gram** | Subword | 8,185 | 13.00 | 52,350 | 17.1% | 47.5% |
| **5-gram** | Word | 3,808 | 11.89 | 4,846 | 13.7% | 41.8% |
| **5-gram** | Subword | 20,242 | 14.31 | 96,936 | 11.3% | 34.0% |

### Top 5 N-grams by Size

**2-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `nang á̱` | 1,002 |
| 2 | `di̱ fam` | 924 |
| 3 | `á̱ ku` | 675 |
| 4 | `a̱ si̱` | 657 |
| 5 | `ku yet` | 653 |

**3-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `di̱ fam a̱tak` | 234 |
| 2 | `nang á̱ ku` | 230 |
| 3 | `di̱ fam a̱za` | 209 |
| 4 | `nang á̱ ngyei` | 200 |
| 5 | `ya̱fang a̱ka̱fwuop nta` | 196 |

**4-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `zwat swak ma̱ng sweang` | 86 |
| 2 | `kyiak neet ma̱ a̱lyia̱` | 82 |
| 3 | `wiki bootcamp season 1` | 80 |
| 4 | `di̱ fam a̱za hu` | 72 |
| 5 | `di̱ fam a̱tyin hu` | 70 |

**5-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `neet ma̱ a̱lyia̱ ba̱ng si̱` | 62 |
| 2 | `á̱ lyen ma̱ng a̱lyoot a̱gwomna̱ti` | 62 |
| 3 | `kyiak neet ma̱ a̱lyia̱ ba̱ng` | 59 |
| 4 | `in tyap romanian and english` | 58 |
| 5 | `together in tyap romanian and` | 58 |

**2-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `_ a̱` | 38,395 |
| 2 | `n g` | 35,424 |
| 3 | `a n` | 31,339 |
| 4 | `t _` | 27,103 |
| 5 | `a _` | 26,601 |

**3-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `n g _` | 26,180 |
| 2 | `a n g` | 17,304 |
| 3 | `e t _` | 10,561 |
| 4 | `_ m a̱` | 8,983 |
| 5 | `a t _` | 7,766 |

**4-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `a n g _` | 13,963 |
| 2 | `y i a̱ _` | 6,492 |
| 3 | `a̱ n g _` | 6,360 |
| 4 | `_ m a̱ n` | 6,098 |
| 5 | `m a̱ n g` | 5,713 |

**5-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `m a̱ n g _` | 5,692 |
| 2 | `_ m a̱ n g` | 5,676 |
| 3 | `_ y e t _` | 4,648 |
| 4 | `n a n g _` | 3,924 |
| 5 | `b y i n _` | 3,628 |


### Key Findings

- **Best Perplexity:** 2-gram (subword) with 265
- **Entropy Trend:** Decreases with larger n-grams (more predictable)
- **Coverage:** Top-1000 patterns cover ~34% 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 | 0.7557 | 1.688 | 4.71 | 28,147 | 24.4% |
| **1** | Subword | 0.9793 | 1.972 | 6.49 | 911 | 2.1% |
| **2** | Word | 0.2473 | 1.187 | 1.54 | 132,079 | 75.3% |
| **2** | Subword | 0.8642 | 1.820 | 4.83 | 5,908 | 13.6% |
| **3** | Word | 0.0833 | 1.059 | 1.13 | 202,426 | 91.7% |
| **3** | Subword | 0.7719 | 1.708 | 3.49 | 28,551 | 22.8% |
| **4** | Word | 0.0300 🏆 | 1.021 | 1.04 | 228,145 | 97.0% |
| **4** | Subword | 0.5638 | 1.478 | 2.31 | 99,668 | 43.6% |

### Generated Text Samples (Word-based)

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

**Context Size 1:**

1. `ma̱ng a̱lyoot a̱liza̱nda mi̱ a̱bibyia̱ njen nang si̱tet ba̱yelsa shyia̱ cet a̱gwaza a̱nyiung di̱ fam...`
2. `ku nihon shong kaswuo a̱ni nggu a̱tyoli sa̱mwila a̱cyia̱ shong mediterranean ba̱ nyia̱ a̱yaafim ku s...`
3. `yet a̱tyulyuut ma̱ng a̱za jenshyung si̱tet ka̱duna a̱tak shong www stoa org dead keys in the`

**Context Size 2:**

1. `nang á̱ ku mbwuo lyulyoot a̱ni ni̱nia yet guadalajara monterrey puebla toluca tijuana ciudad juárez ...`
2. `di̱ fam a̱byin jenshyung a̱siya a̱sa̱khwot nhu na a̱ni tamah si̱ ci a̱pyie ngu nang kham nsaai`
3. `á̱ ku mi̱n a̱ khwuat a̱nietca̱tshot a̱niet khwo mba tai a̱ ku ngyei gini potuga a̱ni ma̱nang`

**Context Size 3:**

1. `di̱ fam a̱tak hu a̱za afrika si̱ myian a̱ja a̱wot di̱ fam a̱tak si̱tet ka̱duna naijeriya a̱ nyia̱`
2. `nang á̱ ku byin nggu a̱tali̱gan a̱ga̱mi tshshekari was born in taligan magamia zangon kataf to paren...`
3. `di̱ fam a̱za hu naat kyai a̱sa̱khwot caina a̱tak hu yet kyai a̱sa̱khwot ku shyia̱ di̱ ngaan fam`

**Context Size 4:**

1. `zwat swak ma̱ng sweang yet a̱tyukwai nfwuo á̱niet naijeriya wa a̱nyan wa yet byiek a̱kwak a̱son á̱gw...`
2. `kyiak neet ma̱ a̱lyia̱ ba̱ng si̱ tat a̱ ku ba̱ng cucuk a̱gwomna̱ti jhyang di̱n jen ji̱ ku swak a̱ni`
3. `di̱ fam a̱za hu a̱fganistan di̱ fam a̱tyin hu ka̱ ka̱u di̱ si̱sak nang lili a̱byin ka yet a̱ni`


### Generated Text Samples (Subword-based)

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

**Context Size 1:**

1. `_á̱_fabefandera_e`
2. `anwu.,_hwunre,_m`
3. `ngbamang_mi̱ta_á̱k`

**Context Size 2:**

1. `_a̱fangba̱_ny-fwuo_`
2. `ng_hi_biya_bya_si̱`
3. `ang_á̱ni._ya̱u_vin_`

**Context Size 3:**

1. `ng_a̱yaaethe_part_o`
2. `angka̱i_a̱khai_ba_,_`
3. `et_a̱lyen_shong_a̱ku`

**Context Size 4:**

1. `ang_gini_ka̱sitibin_`
2. `yia̱_a̱yaapi̱rotidia._`
3. `a̱ng_si̱_swak_mi̱_suso`


### Key Findings

- **Best Predictability:** Context-4 (word) with 97.0% predictability
- **Branching Factor:** Decreases with context size (more deterministic)
- **Memory Trade-off:** Larger contexts require more storage (99,668 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 | 11,223 |
| Total Tokens | 236,752 |
| Mean Frequency | 21.10 |
| Median Frequency | 3 |
| Frequency Std Dev | 149.81 |

### Most Common Words

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | ma̱ng | 5,701 |
| 2 | ku | 5,107 |
| 3 | yet | 4,705 |
| 4 | si̱ | 3,684 |
| 5 | a̱ni | 3,615 |
| 6 | hu | 3,553 |
| 7 | á̱ | 3,391 |
| 8 | nang | 3,386 |
| 9 | a̱ | 3,096 |
| 10 | ka | 2,820 |

### Least Common Words (from vocabulary)

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | tockus | 2 |
| 2 | erythrorhynchus | 2 |
| 3 | atu | 2 |
| 4 | luwut | 2 |
| 5 | akad | 2 |
| 6 | أبو | 2 |
| 7 | نواس | 2 |
| 8 | nuwās | 2 |
| 9 | a̱tyoka̱u | 2 |
| 10 | basi̱li̱kata | 2 |

### Zipf's Law Analysis

| Metric | Value |
|--------|-------|
| Zipf Coefficient | 1.1596 |
| R² (Goodness of Fit) | 0.992895 |
| Adherence Quality | **excellent** |

### Coverage Analysis

| Top N Words | Coverage |
|-------------|----------|
| Top 100 | 48.0% |
| Top 1,000 | 78.4% |
| Top 5,000 | 93.6% |
| Top 10,000 | 99.0% |

### Key Findings

- **Zipf Compliance:** R²=0.9929 indicates excellent adherence to Zipf's law
- **High Frequency Dominance:** Top 100 words cover 48.0% of corpus
- **Long Tail:** 1,223 words needed for remaining 1.0% 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 | 0.3873 🏆 | 0.4467 | N/A | N/A |
| **mono_64d** | 64 | 0.0916 | 0.4260 | N/A | N/A |
| **mono_128d** | 128 | 0.0123 | 0.4367 | N/A | N/A |
| **aligned_32d** | 32 | 0.3873 | 0.4319 | 0.0240 | 0.1440 |
| **aligned_64d** | 64 | 0.0916 | 0.4421 | 0.0200 | 0.1440 |
| **aligned_128d** | 128 | 0.0123 | 0.4376 | 0.0160 | 0.1340 |

### Key Findings

- **Best Isotropy:** mono_32d with 0.3873 (more uniform distribution)
- **Semantic Density:** Average pairwise similarity of 0.4368. Lower values indicate better semantic separation.
- **Alignment Quality:** Aligned models achieve up to 2.4% 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 | **0.229** | 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` | a̱tyuweang, a̱ka̱satyok, american |
| `-n` | nia, ning, na̠ |
| `-s` | sardi, songs, swot |
| `-m` | ma̱m, ma̱li̱daviya, mabyin |
| `-k` | kwaimam, kpantyin, kwom |
| `-b` | bendel, bu, buzău |
| `-t` | tyantung, ta̱lyi̱ri̱p, tunis |
| `-ma` | ma̱m, ma̱li̱daviya, mabyin |

#### Productive Suffixes
| Suffix | Examples |
|--------|----------|
| `-a` | nia, ania, ma̱li̱daviya |
| `-n` | american, a̱yangka̱nan, rénmín |
| `-ng` | ga̱swúong, a̱tyuweang, tyantung |
| `-t` | lilyuut, felt, list |
| `-g` | ga̱swúong, a̱tyuweang, tyantung |
| `-i` | yhui, a̱yaazoni, a̱vwui |
| `-s` | prayers, songs, français |
| `-e` | fare, harare, senate |

### 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 |
|------|----------|------------------|----------|
| `yang` | 1.38x | 56 contexts | gyang, lyang, jyang |
| `wang` | 1.62x | 25 contexts | gwang, nwang, swang |
| `eang` | 1.59x | 26 contexts | keang, weang, teang |
| `tion` | 1.88x | 13 contexts | action, nation, notion |
| `wuan` | 1.50x | 23 contexts | swuan, fwuan, vwuan |
| `yiak` | 1.67x | 16 contexts | tyiak, kyiak, byiak |
| `yiat` | 1.56x | 18 contexts | tyiat, lyiat, kyiat |
| `wuon` | 1.51x | 19 contexts | fwuon, vwuon, bwuon |
| `hyan` | 1.69x | 11 contexts | nhyan, ghyang, hihyan |
| `nshy` | 1.33x | 14 contexts | nshye, nshya, nshyie |
| `kean` | 1.50x | 9 contexts | keang, keana, a̱kean |
| `nyiu` | 1.48x | 9 contexts | a̱nyiu, nyiung, anyiuk |

### 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 |
|--------|--------|-----------|----------|
| `-a` | `-g` | 194 words | anbang, a̱tyubwuanng |
| `-a` | `-ng` | 193 words | anbang, a̱tyubwuanng |
| `-a` | `-t` | 166 words | a̱gwut, a̱tat |
| `-a` | `-i` | 144 words | a̱ta̱nii, agwii |
| `-a` | `-a` | 137 words | alata, a̱jiya |
| `-a` | `-n` | 131 words | afwun, a̱zabyin |
| `-a` | `-k` | 104 words | acucuk, akanok |
| `-a` | `-an` | 53 words | ashan, american |
| `-c` | `-s` | 41 words | collins, caucasus |
| `-k` | `-a` | 41 words | kola, ki̱risi̱ta |

### 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 |
|------|-----------------|------------|------|
| marketing | **`market-i-ng`** | 7.5 | `i` |
| kuzangmam | **`kuzang-m-am`** | 7.5 | `m` |
| a̱ka̱safang | **`a̱ka̱saf-a-ng`** | 7.5 | `a` |
| kyangtutu | **`kyangtu-t-u`** | 7.5 | `t` |
| ka̱zaktan | **`ka̱zak-t-an`** | 7.5 | `t` |
| á̱nietnzop | **`á̱nietnz-o-p`** | 7.5 | `o` |
| christians | **`christi-an-s`** | 7.5 | `an` |
| atakjenshyung | **`at-ak-jenshyung`** | 7.5 | `jenshyung` |
| nvwuomaat | **`nvwuom-a-at`** | 7.5 | `a` |
| institution | **`institut-i-on`** | 7.5 | `i` |
| a̱tyulyiai | **`a̱tyuly-i-ai`** | 7.5 | `i` |
| nggwoneam | **`nggwon-e-am`** | 7.5 | `e` |
| a̱nyanyan | **`a̱nyan-ya-n`** | 6.0 | `a̱nyan` |
| africaines | **`africa-in-es`** | 6.0 | `africa` |
| a̱kwokwak | **`a̱kwok-wa-k`** | 6.0 | `a̱kwok` |

### 6.6 Linguistic Interpretation

> **Automated Insight:**
The language Tyap 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 (4.83x) |
| N-gram | **2-gram** | Lowest perplexity (265) |
| Markov | **Context-4** | Highest predictability (97.0%) |
| 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 07:27:32*