README.md

---
language: tl
language_name: Filipino
language_family: austronesian_philippine_central
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-austronesian_philippine_central
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: 4.787
  - name: best_isotropy
    type: isotropy
    value: 0.8025
  - name: vocabulary_size
    type: vocab
    value: 0
generated: 2026-01-11
---

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

This repository contains NLP models trained and evaluated by Wikilangs, specifically on **Filipino** 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** | 3.870x | 3.87 | 0.0846% | 1,144,874 |
| **16k** | 4.258x | 4.26 | 0.0930% | 1,040,653 |
| **32k** | 4.570x | 4.57 | 0.0998% | 969,681 |
| **64k** | 4.787x 🏆 | 4.79 | 0.1046% | 925,672 |

### Tokenization Examples

Below are sample sentences tokenized with each vocabulary size:

**Sample 1:** `Ang Anastasius I o Anastasio I ay maaaring tumukoy kina: Anastasius I (emperador...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁ang ▁ana sta si us ▁i ▁o ▁ana sta sio ... (+25 more)` | 35 |
| 16k | `▁ang ▁anasta sius ▁i ▁o ▁anasta sio ▁i ▁ay ▁maaaring ... (+17 more)` | 27 |
| 32k | `▁ang ▁anasta sius ▁i ▁o ▁anasta sio ▁i ▁ay ▁maaaring ... (+15 more)` | 25 |
| 64k | `▁ang ▁anastasius ▁i ▁o ▁anastasio ▁i ▁ay ▁maaaring ▁tumukoy ▁kina ... (+11 more)` | 21 |

**Sample 2:** `Ang alupihan ay tumutukoy sa mga sumusunod: alupihan, hayop na maraming mga paa ...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁ang ▁a lu pi han ▁ay ▁tumutukoy ▁sa ▁mga ▁sumusunod ... (+23 more)` | 33 |
| 16k | `▁ang ▁alu pi han ▁ay ▁tumutukoy ▁sa ▁mga ▁sumusunod : ... (+19 more)` | 29 |
| 32k | `▁ang ▁alu pi han ▁ay ▁tumutukoy ▁sa ▁mga ▁sumusunod : ... (+19 more)` | 29 |
| 64k | `▁ang ▁alupihan ▁ay ▁tumutukoy ▁sa ▁mga ▁sumusunod : ▁alupihan , ... (+15 more)` | 25 |

**Sample 3:** `Tumutukoy ang Getafe sa: Getafe, Bohol, Pilipinas Getafe, Espanya`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁tumutukoy ▁ang ▁ge ta fe ▁sa : ▁ge ta fe ... (+9 more)` | 19 |
| 16k | `▁tumutukoy ▁ang ▁ge ta fe ▁sa : ▁ge ta fe ... (+9 more)` | 19 |
| 32k | `▁tumutukoy ▁ang ▁ge ta fe ▁sa : ▁ge ta fe ... (+9 more)` | 19 |
| 64k | `▁tumutukoy ▁ang ▁geta fe ▁sa : ▁geta fe , ▁bohol ... (+6 more)` | 16 |


### Key Findings

- **Best Compression:** 64k achieves 4.787x compression
- **Lowest UNK Rate:** 8k with 0.0846% 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 | 47,186 | 15.53 | 318,514 | 13.3% | 28.2% |
| **2-gram** | Subword | 197 🏆 | 7.62 | 12,564 | 75.1% | 99.3% |
| **3-gram** | Word | 194,690 | 17.57 | 626,197 | 5.1% | 14.4% |
| **3-gram** | Subword | 1,562 | 10.61 | 73,993 | 36.4% | 76.3% |
| **4-gram** | Word | 444,151 | 18.76 | 1,007,564 | 4.2% | 10.1% |
| **4-gram** | Subword | 8,805 | 13.10 | 386,404 | 20.7% | 48.0% |
| **5-gram** | Word | 288,906 | 18.14 | 622,946 | 5.8% | 12.4% |
| **5-gram** | Subword | 34,036 | 15.05 | 1,176,700 | 12.2% | 33.2% |

### Top 5 N-grams by Size

**2-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `ng mga` | 122,547 |
| 2 | `sa mga` | 92,284 |
| 3 | `ang mga` | 86,243 |
| 4 | `ay isang` | 47,028 |
| 5 | `mula sa` | 45,918 |

**3-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `sa pamamagitan ng` | 15,624 |
| 2 | `sa lalawigan ng` | 8,276 |
| 3 | `sa pagitan ng` | 8,017 |
| 4 | `mga sanggunian mga` | 7,752 |
| 5 | `iba t ibang` | 7,698 |

**4-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `mga panlabas na link` | 5,294 |
| 2 | `sanggunian mga panlabas na` | 4,753 |
| 3 | `mga sanggunian mga panlabas` | 4,623 |
| 4 | `munisipalidad sa lalawigan ng` | 3,555 |
| 5 | `comune komuna o munisipalidad` | 3,547 |

**5-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `mga sanggunian mga panlabas na` | 4,621 |
| 2 | `sanggunian mga panlabas na link` | 4,299 |
| 3 | `comune komuna o munisipalidad sa` | 3,419 |
| 4 | `sa mga sumusunod na munisipalidad` | 3,189 |
| 5 | `ay isang comune komuna o` | 3,156 |

**2-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `n g` | 3,917,952 |
| 2 | `a n` | 3,737,257 |
| 3 | `g _` | 3,418,646 |
| 4 | `a _` | 3,186,790 |
| 5 | `_ n` | 2,406,716 |

**3-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `n g _` | 3,291,039 |
| 2 | `a n g` | 2,010,994 |
| 3 | `_ s a` | 1,072,670 |
| 4 | `_ n a` | 1,030,586 |
| 5 | `_ n g` | 987,165 |

**4-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `a n g _` | 1,606,671 |
| 2 | `_ n g _` | 960,600 |
| 3 | `_ s a _` | 872,495 |
| 4 | `_ n a _` | 613,902 |
| 5 | `_ a n g` | 594,113 |

**5-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `_ a n g _` | 585,381 |
| 2 | `_ m g a _` | 498,790 |
| 3 | `n g _ p a` | 315,071 |
| 4 | `g _ m g a` | 277,715 |
| 5 | `n g _ m g` | 277,460 |


### Key Findings

- **Best Perplexity:** 2-gram (subword) with 197
- **Entropy Trend:** Decreases with larger n-grams (more predictable)
- **Coverage:** Top-1000 patterns cover ~33% 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 | 0.8300 | 1.778 | 7.53 | 527,629 | 17.0% |
| **1** | Subword | 0.9447 | 1.925 | 6.25 | 10,325 | 5.5% |
| **2** | Word | 0.3582 | 1.282 | 2.25 | 3,967,765 | 64.2% |
| **2** | Subword | 0.5676 | 1.482 | 3.43 | 64,498 | 43.2% |
| **3** | Word | 0.1673 | 1.123 | 1.38 | 8,894,925 | 83.3% |
| **3** | Subword | 0.5929 | 1.508 | 3.39 | 221,050 | 40.7% |
| **4** | Word | 0.0699 🏆 | 1.050 | 1.12 | 12,295,618 | 93.0% |
| **4** | Subword | 0.6330 | 1.551 | 3.10 | 748,630 | 36.7% |

### Generated Text Samples (Word-based)

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

**Context Size 1:**

1. `ng magaang mga nakamit kasunod ng mga tren kiha 20 second movement noong heograpiya ang timog`
2. `sa kasaysayan ng diyos at idinagdag ang pagkakasakit namatay ang estado sa hilaga lungsod sa benta`
3. `ang bayan sa silangang eslabong kaharian maaring magbayad ng pagkakaroon o mala pabilog harapang nak...`

**Context Size 2:**

1. `ng mga tao sinasabi na parang gunting pagguguntingan kalish nancy the nice guys holly march sa isang`
2. `sa mga katangian ng larangang ito bagaman ang christ ang pananampalataya sa diyos sapagkat nawalan n...`
3. `ang mga teoretikal na edukasyon na si tenzin gyatso ang ikawalong baitang 13 taon chronology of afri...`

**Context Size 3:**

1. `sa pamamagitan ng plots and distribusyon ng isang natutunghayan ang eigen ay sarili sa aleman mainam...`
2. `sa lalawigan ng cuneo sa rehiyon ng lazio na matatagpuan mga timog ng mantua matatagpuan sa isang bu...`
3. `sa pagitan ng dalawang organismo sa kaso ng isang kurtinang pang shower ang kurtina ay iyon ding nag...`

**Context Size 4:**

1. `mga panlabas na link opisyal na website thayers gazetteer international school of painting drawing a...`
2. `sanggunian mga panlabas na link opisyal na website bayan at lungsod sa pilipinas subalit bilang kara...`
3. `mga sanggunian mga panlabas na link plundering desire articles interviews release reviews live revie...`


### Generated Text Samples (Subword-based)

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

**Context Size 1:**

1. `anikuw_ng_shinit`
2. `_likataltung_nga`
3. `nfedinasonyahepa`

**Context Size 2:**

1. `ng_noong_ga_sang_`
2. `ana_mga_markilang`
3. `g_magpumish._puna`

**Context Size 3:**

1. `ng_malawan_nasakup`
2. `ang_tagpuanibersiy`
3. `_sa_ay_mayroon_tum`

**Context Size 4:**

1. `ang_pagtuunawaganap`
2. `_ng_telepono,_dahil`
3. `_sa_mga_panahong_om`


### Key Findings

- **Best Predictability:** Context-4 (word) with 93.0% predictability
- **Branching Factor:** Decreases with context size (more deterministic)
- **Memory Trade-off:** Larger contexts require more storage (748,630 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 | 223,605 |
| Total Tokens | 15,229,985 |
| Mean Frequency | 68.11 |
| Median Frequency | 4 |
| Frequency Std Dev | 3743.03 |

### Most Common Words

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | ng | 962,341 |
| 2 | sa | 881,526 |
| 3 | ang | 628,027 |
| 4 | na | 621,434 |
| 5 | mga | 506,055 |
| 6 | ay | 352,169 |
| 7 | at | 351,974 |
| 8 | isang | 180,575 |
| 9 | noong | 112,415 |
| 10 | ito | 97,397 |

### Least Common Words (from vocabulary)

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | madiclum | 2 |
| 2 | festivalpinakamahusay | 2 |
| 3 | siboryo | 2 |
| 4 | slazenger | 2 |
| 5 | yuwji | 2 |
| 6 | mandoriao | 2 |
| 7 | buzinkai | 2 |
| 8 | hiveswap | 2 |
| 9 | writerin | 2 |
| 10 | sskp | 2 |

### Zipf's Law Analysis

| Metric | Value |
|--------|-------|
| Zipf Coefficient | 1.0072 |
| R² (Goodness of Fit) | 0.995022 |
| Adherence Quality | **excellent** |

### Coverage Analysis

| Top N Words | Coverage |
|-------------|----------|
| Top 100 | 44.9% |
| Top 1,000 | 64.0% |
| Top 5,000 | 79.3% |
| Top 10,000 | 85.3% |

### Key Findings

- **Zipf Compliance:** R²=0.9950 indicates excellent adherence to Zipf's law
- **High Frequency Dominance:** Top 100 words cover 44.9% of corpus
- **Long Tail:** 213,605 words needed for remaining 14.7% 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 | 0.8025 | 0.3575 | N/A | N/A |
| **mono_64d** | 64 | 0.7423 | 0.3056 | N/A | N/A |
| **mono_128d** | 128 | 0.6846 | 0.2378 | N/A | N/A |
| **aligned_32d** | 32 | 0.8025 🏆 | 0.3655 | 0.3000 | 0.7020 |
| **aligned_64d** | 64 | 0.7423 | 0.2994 | 0.4300 | 0.8300 |
| **aligned_128d** | 128 | 0.6846 | 0.2419 | 0.5400 | 0.8680 |

### Key Findings

- **Best Isotropy:** aligned_32d with 0.8025 (more uniform distribution)
- **Semantic Density:** Average pairwise similarity of 0.3013. Lower values indicate better semantic separation.
- **Alignment Quality:** Aligned models achieve up to 54.0% 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 | **5.000** | High morphological productivity | Reliable analysis |
| Idiomaticity Gap | **-0.628** | 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` | mangangasiwa, maruja, masangkot |
| `-a` | aggie, arkimedes, antoni |
| `-s` | suleiman, sutan, steri |
| `-d` | democratikong, dugong, dlä |
| `-pa` | paranorman, parañaquelungsod, panti |
| `-m` | mánudagur, mundhum, moluccan |
| `-na` | nakalilitong, nangangagat, nagpupunyagi |
| `-ka` | kabuwanan, kalokohang, kalbaryo |

#### Productive Suffixes
| Suffix | Examples |
|--------|----------|
| `-ng` | improvising, democratikong, sikiyatriyang |
| `-n` | buogn, suleiman, sutan |
| `-a` | echeverría, periyodontista, tasya |
| `-g` | improvising, democratikong, sikiyatriyang |
| `-s` | rudolfensis, gulbis, arkimedes |
| `-o` | campochiaro, villonco, incognito |
| `-e` | aggie, batake, zakopane |
| `-an` | suleiman, sutan, paranorman |

### 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 |
|------|----------|------------------|----------|
| `inak` | 2.61x | 78 contexts | inako, pinak, inakma |
| `angg` | 2.17x | 161 contexts | sangg, angge, anggi |
| `inag` | 2.25x | 112 contexts | sinag, tinag, inagi |
| `agka` | 2.24x | 106 contexts | nagka, magka, sagka |
| `ngga` | 2.16x | 122 contexts | ungga, angga, tingga |
| `atag` | 2.19x | 110 contexts | patag, latag, datag |
| `agpa` | 2.21x | 92 contexts | pagpa, magpa, agpay |
| `angk` | 1.90x | 168 contexts | angka, sangka, sangko |
| `tion` | 2.15x | 82 contexts | tiong, ation, tione |
| `alaw` | 2.01x | 105 contexts | galaw, kalaw, alaws |
| `asyo` | 2.07x | 90 contexts | basyo, rasyo, tasyo |
| `inas` | 1.84x | 127 contexts | sinas, rinas, linas |

### 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 |
|--------|--------|-----------|----------|
| `-pa` | `-g` | 84 words | pankalakalang, pangangatawang |
| `-s` | `-n` | 76 words | saksakyan, sulangan |
| `-s` | `-a` | 73 words | sharmiela, semigallia |
| `-pa` | `-ng` | 71 words | pankalakalang, pangangatawang |
| `-pa` | `-n` | 71 words | pamain, paparusahan |
| `-na` | `-g` | 68 words | nagnangalang, napakabantog |
| `-pa` | `-a` | 66 words | pagkokomplementa, pamina |
| `-a` | `-a` | 66 words | alionushka, atienza |
| `-ka` | `-n` | 66 words | kasalukyan, karangyaan |
| `-ma` | `-g` | 66 words | masong, mabubuwag |

### 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 |
|------|-----------------|------------|------|
| napakalapot | **`napakalap-o-t`** | 7.5 | `o` |
| makapagpapatisod | **`makapagpapatis-o-d`** | 7.5 | `o` |
| montmirail | **`montmira-i-l`** | 7.5 | `i` |
| magtutuos | **`magtutu-o-s`** | 7.5 | `o` |
| kinaroroonang | **`kinaroroon-a-ng`** | 7.5 | `a` |
| sampaybakod | **`sampaybak-o-d`** | 7.5 | `o` |
| obergefell | **`obergefe-l-l`** | 7.5 | `l` |
| tinablang | **`tinab-la-ng`** | 7.5 | `la` |
| nababayarang | **`nababayar-a-ng`** | 7.5 | `a` |
| masmataas | **`ma-s-mataas`** | 7.5 | `mataas` |
| maghuhugas | **`maghuhu-g-as`** | 7.5 | `g` |
| napakakipot | **`napakakip-o-t`** | 7.5 | `o` |
| inglewood | **`inglewo-o-d`** | 7.5 | `o` |
| concerned | **`concer-n-ed`** | 7.5 | `n` |
| internationally | **`international-l-y`** | 7.5 | `l` |

### 6.6 Linguistic Interpretation

> **Automated Insight:**
The language Filipino 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 (4.79x) |
| N-gram | **2-gram** | Lowest perplexity (197) |
| Markov | **Context-4** | Highest predictability (93.0%) |
| 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: 2026-01-11 02:21:03*