README.md

---
language: yi
language_name: Yiddish
language_family: germanic_west_continental
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_continental
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: 4.552
  - name: best_isotropy
    type: isotropy
    value: 0.8430
  - name: vocabulary_size
    type: vocab
    value: 0
generated: 2026-01-11
---

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

This repository contains NLP models trained and evaluated by Wikilangs, specifically on **Yiddish** 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** | 3.841x | 3.84 | 0.1120% | 631,919 |
| **16k** | 4.158x | 4.16 | 0.1213% | 583,788 |
| **32k** | 4.393x | 4.40 | 0.1282% | 552,468 |
| **64k** | 4.552x 🏆 | 4.55 | 0.1328% | 533,206 |

### Tokenization Examples

Below are sample sentences tokenized with each vocabulary size:

**Sample 1:** `געשעענישן געבוירן נפטר געווארן קאלענדאר צום באמערקן רעפערענצן`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁געשעענישן ▁געבוירן ▁נפטר ▁געווארן ▁קאלענדאר ▁צום ▁באמערקן ▁רעפערענצן` | 8 |
| 16k | `▁געשעענישן ▁געבוירן ▁נפטר ▁געווארן ▁קאלענדאר ▁צום ▁באמערקן ▁רעפערענצן` | 8 |
| 32k | `▁געשעענישן ▁געבוירן ▁נפטר ▁געווארן ▁קאלענדאר ▁צום ▁באמערקן ▁רעפערענצן` | 8 |
| 64k | `▁געשעענישן ▁געבוירן ▁נפטר ▁געווארן ▁קאלענדאר ▁צום ▁באמערקן ▁רעפערענצן` | 8 |

**Sample 2:** `געשעענישן געבוירן 24סטן יאנואר - פרידריך דער גרויסער, מלך פון פרייסן (געש' 28סטן...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁געשעענישן ▁געבוירן ▁ 2 4 סטן ▁יאנואר ▁- ▁פרידריך ▁דער ... (+27 more)` | 37 |
| 16k | `▁געשעענישן ▁געבוירן ▁ 2 4 סטן ▁יאנואר ▁- ▁פרידריך ▁דער ... (+27 more)` | 37 |
| 32k | `▁געשעענישן ▁געבוירן ▁ 2 4 סטן ▁יאנואר ▁- ▁פרידריך ▁דער ... (+25 more)` | 35 |
| 64k | `▁געשעענישן ▁געבוירן ▁ 2 4 סטן ▁יאנואר ▁- ▁פרידריך ▁דער ... (+25 more)` | 35 |

**Sample 3:** `א מענטש האט אויף יעדער האנט פינף פינגער. זעט אויך פינגער (פוס) אנאטאמיע`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁א ▁מענטש ▁האט ▁אויף ▁יעדער ▁האנט ▁פינף ▁פ ינגער . ... (+10 more)` | 20 |
| 16k | `▁א ▁מענטש ▁האט ▁אויף ▁יעדער ▁האנט ▁פינף ▁פינגער . ▁זעט ... (+6 more)` | 16 |
| 32k | `▁א ▁מענטש ▁האט ▁אויף ▁יעדער ▁האנט ▁פינף ▁פינגער . ▁זעט ... (+6 more)` | 16 |
| 64k | `▁א ▁מענטש ▁האט ▁אויף ▁יעדער ▁האנט ▁פינף ▁פינגער . ▁זעט ... (+6 more)` | 16 |


### Key Findings

- **Best Compression:** 64k achieves 4.552x compression
- **Lowest UNK Rate:** 8k with 0.1120% 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 | 21,980 | 14.42 | 83,327 | 13.3% | 32.3% |
| **2-gram** | Subword | 275 🏆 | 8.10 | 6,028 | 68.2% | 98.3% |
| **3-gram** | Word | 61,497 | 15.91 | 131,301 | 6.0% | 17.5% |
| **3-gram** | Subword | 2,102 | 11.04 | 45,237 | 31.8% | 72.4% |
| **4-gram** | Word | 130,494 | 16.99 | 212,902 | 3.8% | 10.8% |
| **4-gram** | Subword | 10,721 | 13.39 | 208,071 | 17.9% | 44.3% |
| **5-gram** | Word | 103,402 | 16.66 | 145,493 | 3.1% | 10.1% |
| **5-gram** | Subword | 36,498 | 15.16 | 485,661 | 10.7% | 29.1% |

### Top 5 N-grams by Size

**2-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `פון די` | 13,720 |
| 2 | `איז געווען` | 11,141 |
| 3 | `אין די` | 9,304 |
| 4 | `איז א` | 8,395 |
| 5 | `אין דער` | 8,145 |

**3-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `איז געווען א` | 2,689 |
| 2 | `די ווייב פון` | 2,393 |
| 3 | `א זון פון` | 2,168 |
| 4 | `ער איז געווען` | 1,847 |
| 5 | `איז געווען דער` | 1,502 |

**4-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `א זון פון הרב` | 1,309 |
| 2 | `די ווייב פון רבי` | 1,223 |
| 3 | `די ווייב פון הרב` | 967 |
| 4 | `א טאכטער פון הרב` | 935 |
| 5 | `איז געבוירן געווארן אין` | 602 |

**5-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `אנציקלופדיה לחכמי גליציה מאיר וונדר` | 383 |
| 2 | `ביז צום סוף יאר בלייבן` | 365 |
| 3 | `צום סוף יאר בלייבן נאך` | 364 |
| 4 | `און צו זיין מוטער מרת` | 357 |
| 5 | `אינעם גרעגאריאנישן קאלענדאר ביז צום` | 336 |

**2-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `ן _` | 767,030 |
| 2 | `_ א` | 671,834 |
| 3 | `ע ר` | 443,218 |
| 4 | `ר _` | 336,097 |
| 5 | `ט _` | 319,572 |

**3-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `_ א י` | 258,659 |
| 2 | `ע ר _` | 253,758 |
| 3 | `ו ן _` | 215,735 |
| 4 | `_ א ו` | 163,745 |
| 5 | `ן _ א` | 160,680 |

**4-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `_ ד י _` | 111,007 |
| 2 | `פ ו ן _` | 108,221 |
| 3 | `_ פ ו ן` | 105,513 |
| 4 | `א י ן _` | 97,976 |
| 5 | `_ א י ן` | 97,190 |

**5-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `_ פ ו ן _` | 105,410 |
| 2 | `_ א י ן _` | 97,087 |
| 3 | `_ א ו ן _` | 88,981 |
| 4 | `_ א י ז _` | 80,703 |
| 5 | `_ ד ע ר _` | 61,940 |


### Key Findings

- **Best Perplexity:** 2-gram (subword) with 275
- **Entropy Trend:** Decreases with larger n-grams (more predictable)
- **Coverage:** Top-1000 patterns cover ~29% 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 | 0.8935 | 1.858 | 7.14 | 157,053 | 10.6% |
| **1** | Subword | 1.0780 | 2.111 | 9.14 | 1,976 | 0.0% |
| **2** | Word | 0.3611 | 1.284 | 2.03 | 1,117,409 | 63.9% |
| **2** | Subword | 0.8485 | 1.801 | 5.31 | 18,020 | 15.1% |
| **3** | Word | 0.1429 | 1.104 | 1.26 | 2,254,596 | 85.7% |
| **3** | Subword | 0.7997 | 1.741 | 3.95 | 95,571 | 20.0% |
| **4** | Word | 0.0521 🏆 | 1.037 | 1.08 | 2,835,381 | 94.8% |
| **4** | Subword | 0.6110 | 1.527 | 2.70 | 377,001 | 38.9% |

### Generated Text Samples (Word-based)

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

**Context Size 1:**

1. `די שיאיטן און דער אונגווארער רב איבער 400 ווען ער איז רייך פון קיִעוו איר דער`
2. `פון סערעט וויזשניץ הרב שלמה לעלאווער רבי חיים פון די אויבנדערמאנטע פראדוקטן פאר א שעה קיין`
3. `אין די געגנט האט זיך ענדליך פארבעסערט אירע צוויי יאר נאכן טויט איז ווען אלע יאהר`

**Context Size 2:**

1. `פון די פארן זענען דער אב בית דין איז מזכה סענדער פון ר אברהם פרידמאן איז רבי`
2. `איז געווען דער איינציגער זון צו זיין פאטער ביי דער צווייטער משגיח רבי מאיר איז נפטר געווארן`
3. `אין די שפיץ שעה ן פון טאג און גאנצע פעקלעך וויכטיגע פיילס אויסגעשפרייט אויף 5 604 פיס`

**Context Size 3:**

1. `איז געווען א שטארקע וואוקס אין קליינע ביזנעסער צו וועלכע די סקאטישע האבן אבער געקרומט מיט די נאז`
2. `די ווייב פון רבי דוד זיינע ספרים שני המאורות רבי שבתי פון ראשקוב פון די חבריא קדישא פון`
3. `א זון פון דעם לאנגן הרב יהושע א זון פון ר מרדכי צמח דער בעל העובדא אויך האט`

**Context Size 4:**

1. `א זון פון הרב אליעזר הורוויץ רב אין ריגליץ און סטאשאוו א זון פון הרב מנחם מענדיל האגער ה`
2. `די ווייב פון רבי אברהם יעקב פרידמאן ה אלול ה תרפ ח י ט טבת תשע ג סאדיגורער רבי`
3. `די ווייב פון הרב דוד שפירא פון סאנוק רבי דוד שפירא ה שבט ה תרל ז כ ב סיון`


### Generated Text Samples (Subword-based)

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

**Context Size 1:**

1. `_דער_ע"נערען_טי_`
2. `יזײַ_גנירער_מצול_`
3. `אזער_רס_ס_ארקרן_`

**Context Size 2:**

1. `ן_שפרושים_טיציג,_`
2. `_איזשעה_אלן_אידי_`
3. `ער_די_סישטיביט_זע`

**Context Size 3:**

1. `_אין_אויף_דער_זיין`
2. `ער_פילמען:_piel_(ק`
3. `ון_רפון_נאנציעסן_ד`

**Context Size 4:**

1. `_די_היינט_זיידעם_גר`
2. `פון_סטעיט_נאך_געווא`
3. `_פון_תורה_אויב_זי"ע`


### Key Findings

- **Best Predictability:** Context-4 (word) with 94.8% predictability
- **Branching Factor:** Decreases with context size (more deterministic)
- **Memory Trade-off:** Larger contexts require more storage (377,001 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 | 69,606 |
| Total Tokens | 3,320,646 |
| Mean Frequency | 47.71 |
| Median Frequency | 4 |
| Frequency Std Dev | 1020.60 |

### Most Common Words

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | די | 112,921 |
| 2 | פון | 105,938 |
| 3 | אין | 97,977 |
| 4 | און | 89,450 |
| 5 | איז | 81,968 |
| 6 | א | 72,112 |
| 7 | דער | 63,946 |
| 8 | האט | 50,599 |
| 9 | ער | 32,997 |
| 10 | צו | 30,909 |

### Least Common Words (from vocabulary)

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | בראַזיליאַנער | 2 |
| 2 | ראקכאָולד | 2 |
| 3 | אַראָפּגעוואָרפן | 2 |
| 4 | xai | 2 |
| 5 | גראָק | 2 |
| 6 | מאַננעס | 2 |
| 7 | מאָטאָרספּאָרט | 2 |
| 8 | ספּיר | 2 |
| 9 | באקער | 2 |
| 10 | דזשוהורי | 2 |

### Zipf's Law Analysis

| Metric | Value |
|--------|-------|
| Zipf Coefficient | 1.1137 |
| R² (Goodness of Fit) | 0.995903 |
| Adherence Quality | **excellent** |

### Coverage Analysis

| Top N Words | Coverage |
|-------------|----------|
| Top 100 | 44.7% |
| Top 1,000 | 69.3% |
| Top 5,000 | 85.0% |
| Top 10,000 | 90.4% |

### Key Findings

- **Zipf Compliance:** R²=0.9959 indicates excellent adherence to Zipf's law
- **High Frequency Dominance:** Top 100 words cover 44.7% of corpus
- **Long Tail:** 59,606 words needed for remaining 9.6% 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 | 0.8392 | 0.3748 | N/A | N/A |
| **mono_64d** | 64 | 0.8430 | 0.2765 | N/A | N/A |
| **mono_128d** | 128 | 0.7897 | 0.1920 | N/A | N/A |
| **aligned_32d** | 32 | 0.8392 | 0.3531 | 0.0140 | 0.1620 |
| **aligned_64d** | 64 | 0.8430 🏆 | 0.2622 | 0.0220 | 0.2260 |
| **aligned_128d** | 128 | 0.7897 | 0.1928 | 0.0940 | 0.3060 |

### Key Findings

- **Best Isotropy:** aligned_64d with 0.8430 (more uniform distribution)
- **Semantic Density:** Average pairwise similarity of 0.2752. Lower values indicate better semantic separation.
- **Alignment Quality:** Aligned models achieve up to 9.4% 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 | **-0.653** | 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 |
|--------|----------|
| `-א` | אבֿות, אויפגעקלערטע, אויסצושפרייטן |
| `-מ` | מעדל, מעכוצעף, מיעוטים |
| `-ה` | האסאקאוו, המחדש, המאוחדת |
| `-ב` | באדינוג, בפה, בוכה |
| `-פ` | פאלעסיע, פאָרויס, פרינציפל |
| `-קא` | קאראקטיר, קאבאל, קאמיוניטי |
| `-בא` | באדינוג, באַטײַט, באַזונדערע |
| `-גע` | געווארענע, געשעט, געהרגט |

#### Productive Suffixes
| Suffix | Examples |
|--------|----------|
| `-ן` | יומן, טעלעפאָן, רעקאדירן |
| `-ער` | 93סטער, שעפעטיווקער, כאראקטער |
| `-ר` | 93סטער, קאראקטיר, שעפעטיווקער |
| `-ע` | פאלעסיע, סוואליאווע, אויפגעקלערטע |
| `-ט` | דערהויפּט, אנשטאלט, אריבערגעטוישט |
| `-ען` | צוריקגעווינען, פיליפיניען, אריינצונעמען |
| `-ס` | סענסארס, פאָרויס, מאָדיפֿיקאַציעס |
| `-ג` | באדינוג, סטאַרטינג, פרובירנדיג |

### 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 |
|------|----------|------------------|----------|
| `אנגע` | 1.82x | 57 contexts | לאנגע, טאנגע, אנגעל |
| `שראל` | 2.40x | 18 contexts | ישראל, ישראלי, לישראל |
| `עווע` | 1.59x | 84 contexts | געווע, מעווע, סטעווע |
| `יבער` | 1.49x | 102 contexts | ליבער, טיבער, ציבער |
| `געוו` | 1.57x | 62 contexts | געווע, געוון, געוויס |
| `דישע` | 1.67x | 47 contexts | ידישע, ײדישע, מאדישע |
| `ידיש` | 1.80x | 33 contexts | אידיש, ידישע, יידיש |
| `ייער` | 1.53x | 62 contexts | אייער, פייער, מייער |
| `נדער` | 1.33x | 94 contexts | אנדער, ענדער, אנדערט |
| `יינע` | 1.41x | 70 contexts | ריינע, דיינע, ניינע |
| `נגען` | 1.74x | 26 contexts | הענגען, גאנגען, שענגען |
| `קומע` | 1.62x | 27 contexts | קומען, קומעט, באקומע |

### 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 |
|--------|--------|-----------|----------|
| `-א` | `-ן` | 361 words | אנצייכענען, אויסגעשריבן |
| `-פ` | `-ן` | 176 words | פאקוסירן, פקדון |
| `-א` | `-ט` | 176 words | אוניוועריסטעט, איזאלירט |
| `-א` | `-ע` | 135 words | אראמישע, אראביקע |
| `-א` | `-ר` | 105 words | אדר, אגריקולטורער |
| `-פ` | `-ע` | 99 words | פארדייטע, פרייליכסטע |
| `-פ` | `-ט` | 93 words | פאָרמאַט, פובליצירט |
| `-פ` | `-ר` | 91 words | פאראמעדיקער, פאנגער |
| `-א` | `-ער` | 91 words | אגריקולטורער, אינטערוויוער |
| `-א` | `-ען` | 90 words | אנצייכענען, אריבערציען |

### 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 |
|------|-----------------|------------|------|
| באוואוינטסטע | **`באוואוינט-ס-טע`** | 7.5 | `ס` |
| אויסגעלאסן | **`אויסגעלא-ס-ן`** | 7.5 | `ס` |
| דרויסנדיקסטע | **`דרויסנדיק-ס-טע`** | 7.5 | `ס` |
| וועלטטייל | **`וועלטטי-י-ל`** | 7.5 | `י` |
| באוויסטזיין | **`באוויסטז-י-ין`** | 7.5 | `י` |
| באַוווּסט | **`באַוווּ-ס-ט`** | 7.5 | `ס` |
| אייברשטען | **`אייברש-ט-ען`** | 7.5 | `ט` |
| בעסמעדרעש | **`בעסמעדר-ע-ש`** | 7.5 | `ע` |
| איוואניטש | **`אי-ווא-ניטש`** | 6.0 | `ניטש` |
| מיטאַרבעטערס | **`מיטאַרבעט-ער-ס`** | 6.0 | `מיטאַרבעט` |
| דרייווערס | **`דרייוו-ער-ס`** | 6.0 | `דרייוו` |
| דעמאלסטיקער | **`דעמאלסט-יק-ער`** | 6.0 | `דעמאלסט` |
| בלייבנדיק | **`בלייב-נד-יק`** | 6.0 | `בלייב` |
| גלייבנדיג | **`גלייב-נד-יג`** | 6.0 | `גלייב` |
| שרייבנדיק | **`שרייב-נד-יק`** | 6.0 | `שרייב` |

### 6.6 Linguistic Interpretation

> **Automated Insight:**
The language Yiddish 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 (4.55x) |
| N-gram | **2-gram** | Lowest perplexity (275) |
| Markov | **Context-4** | Highest predictability (94.8%) |
| 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 05:37:12*