Add files using upload-large-folder tool

20250701/af/README.md

@@ -0,0 +1,252 @@
+# Wikipedia AF Dataset (20250701)
+
+## Overview
+
+This dataset contains processed Wikipedia articles for the **af** language, extracted from the Wikipedia dump dated **20250701**. The dataset has been processed through a comprehensive 9-stage pipeline that includes text normalization, tokenization, n-gram analysis, article scoring, and representative sampling.
+
+## Dataset Statistics
+
+## Dataset Structure
+
+This dataset is organized into the following components:
+
+### 📰 Full Articles Dataset
+- **Complete Dataset**: All Wikipedia articles in multiple Parquet files for HuggingFace compatibility
+- **Location**: `/dataset/full_part_XXX.parquet`
+- **Schema**: id, title, url, text, namespace, raw_mediawiki
+- **Optimization**: Small row groups (1000 rows) for efficient multi-reading
+
+### 🤖 Tokenizer Models
+- **Location**: `/models/tokenizer/`
+- **Multiple Sizes**: 8k, 16k, 32k, and 64k vocabulary sizes
+- **SentencePiece Models**: Trained subword tokenizers for different use cases
+- **Vocabulary Files**: Complete vocabularies with token mappings for each size
+
+### 📊 Subword N-gram Models (Parquet Format)
+- **Location**: `/models/subword_ngram/`
+- **2-gram Model**: Subword bigram frequencies and IDF scores in Parquet format
+- **3-gram Model**: Subword trigram frequencies and IDF scores in Parquet format
+- **4-gram Model**: Subword 4-gram frequencies and IDF scores in Parquet format
+- **Top N-grams**: Most frequent subword n-grams in separate Parquet files
+- **Optimization**: Small row groups for efficient querying and multi-reading
+
+### 📝 Word N-gram Models (Parquet Format)
+- **Location**: `/models/word_ngram/`
+- **2-gram Model**: Word-level bigram frequencies and IDF scores in Parquet format
+- **3-gram Model**: Word-level trigram frequencies and IDF scores in Parquet format
+- **4-gram Model**: Word-level 4-gram frequencies and IDF scores in Parquet format
+- **Top N-grams**: Most frequent word n-grams in separate Parquet files
+- **Tokenization**: Simple whitespace and punctuation-based word splitting
+
+### 🔗 Subword Markov Chain Models (Parquet Format)
+- **Location**: `/models/subword_markov/`
+- **2-gram Context**: Subword transition probabilities for text generation
+- **3-gram Context**: Higher-order subword context for better text generation
+- **4-gram Context**: Maximum subword context for sophisticated text generation
+- **Schema**: context (JSON), next_token, probability, context_count
+
+### 🔗 Word Markov Chain Models (Parquet Format)
+- **Location**: `/models/word_markov/`
+- **2-gram Context**: Word-level transition probabilities for text generation
+- **3-gram Context**: Higher-order word context for better text generation
+- **4-gram Context**: Maximum word context for sophisticated text generation
+- **Schema**: context (JSON), next_token, probability, context_count
+
+### 📚 Vocabulary Models
+- **Location**: `/models/vocabulary/`
+- **Language Dictionary**: Vocabulous-based language detection model
+- **Word-Language Frequencies**: Statistical language identification data
+
+### 📈 Statistics & Reports
+- **Comprehensive Statistics**: Detailed corpus analysis in JSON format
+- **Human-Readable Summary**: Key statistics and insights
+
+### 🎯 Representative Sample Datasets
+- **Location**: `/dataset/`
+- **Sample Sizes**: 1k.parquet, 5k.parquet, 10k.parquet (only created if enough articles available)
+- **Coverage-Optimized**: Samples maximize n-gram coverage of the full corpus
+- **Schema**: id, title, url, text, tokens (JSON), scores (JSON), features (JSON), individual score columns
+- **Optimization**: Small row groups for efficient filtering and analysis
+- **Note**: Sample sizes larger than available articles are automatically skipped
+
+## Processing Pipeline
+
+This dataset was created using a 9-stage processing pipeline:
+
+1. **Data Acquisition**: Download and parse Wikipedia XML dumps
+2. **Text Normalization**: Clean and normalize text using unscript
+3. **Tokenizer Training**: Train SentencePiece subword tokenizers
+4. **Dictionary Building**: Build language detection models with vocabulous
+5. **N-gram Analysis**: Generate comprehensive n-gram models
+6. **Article Scoring**: Score articles for representativeness and quality
+7. **Sample Generation**: Create coverage-optimized representative samples
+8. **Statistics Generation**: Generate comprehensive corpus statistics
+9. **Publication**: Upload all artifacts to Hugging Face Hub
+
+## Usage Examples
+
+### Loading Full Articles Dump
+
+```python
+import pandas as pd
+
+# Load all articles
+df = pd.read_parquet('articles/af_articles.parquet')
+
+# Access article data
+print(f"Total articles: {len(df)}")
+print(df[['id', 'title', 'text']].head())
+```
+
+### Loading the Tokenizer
+
+```python
+import sentencepiece as spm
+
+# Load the trained tokenizer
+sp = spm.SentencePieceProcessor()
+sp.load('af_tokenizer.model')
+
+# Tokenize text
+text = "Your text here"
+tokens = sp.encode_as_pieces(text)
+token_ids = sp.encode_as_ids(text)
+```
+
+### Loading N-gram Models (Parquet Format)
+
+```python
+import pandas as pd
+import json
+
+# Load subword 2-gram model
+subword_df = pd.read_parquet('models/subword_ngram/af_2gram_model.parquet')
+
+# Access subword n-gram data
+print("Subword 2-grams:")
+for _, row in subword_df.head().iterrows():
+    ngram = json.loads(row['ngram'])  # Convert back from JSON
+    frequency = row['frequency']
+    idf_score = row['idf_score']
+    print(f"N-gram: {ngram}, Freq: {frequency}, IDF: {idf_score:.3f}")
+
+# Load word 2-gram model
+word_df = pd.read_parquet('models/word_ngram/af_2gram_model.parquet')
+
+# Access word n-gram data
+print("\nWord 2-grams:")
+for _, row in word_df.head().iterrows():
+    ngram = json.loads(row['ngram'])  # Convert back from JSON
+    frequency = row['frequency']
+    idf_score = row['idf_score']
+    print(f"N-gram: {ngram}, Freq: {frequency}, IDF: {idf_score:.3f}")
+
+# Load top subword n-grams
+subword_top_df = pd.read_parquet('models/subword_ngram/af_2gram_top1000.parquet')
+print("\nTop 10 subword bigrams:")
+print(subword_top_df.head(10))
+
+# Load top word n-grams
+word_top_df = pd.read_parquet('models/word_ngram/af_2gram_top1000.parquet')
+print("\nTop 10 word bigrams:")
+print(word_top_df.head(10))
+```
+
+### Loading Markov Chain Models (Parquet Format)
+
+```python
+import pandas as pd
+import json
+
+# Load subword Markov chain (2-gram context)
+subword_markov_df = pd.read_parquet('models/subword_markov/af_markov1_transitions.parquet')
+
+# Access subword Markov transitions
+print("Subword Markov transitions:")
+for _, row in subword_markov_df.head().iterrows():
+    context = json.loads(row['context'])  # Convert back from JSON
+    next_token = row['next_token']
+    probability = row['probability']
+    context_count = row['context_count']
+    print(f"Context: {context} -> Next: '{next_token}' (p={probability:.3f}, count={context_count})")
+
+# Load word Markov chain (2-gram context)
+word_markov_df = pd.read_parquet('models/word_markov/af_markov1_transitions.parquet')
+
+# Access word Markov transitions
+print("\nWord Markov transitions:")
+for _, row in word_markov_df.head().iterrows():
+    context = json.loads(row['context'])  # Convert back from JSON
+    next_token = row['next_token']
+    probability = row['probability']
+    context_count = row['context_count']
+    print(f"Context: {context} -> Next: '{next_token}' (p={probability:.3f}, count={context_count})")
+```
+
+### Loading Sample Datasets (Parquet Format)
+
+```python
+import pandas as pd
+import json
+
+# Load sample dataset
+sample_df = pd.read_parquet('samples/af_sample_1000.parquet')
+
+# Access article data
+for _, row in sample_df.head().iterrows():
+    tokens = json.loads(row['tokens'])  # Convert back from JSON
+    scores = json.loads(row['scores'])  # Convert back from JSON
+    print(f"Title: {row['title']}")
+    print(f"Tokens: {len(tokens)}")
+    print(f"Scores: {scores}")
+```
+
+### Loading Dictionary Models
+
+```python
+from vocabulous import Vocabulous
+
+# Load language detection model
+model = Vocabulous.load('af_dictionary.json')
+
+# Detect language of text
+text = "Your text here"
+detected_lang = model.detect_language(text)
+```
+
+## Data Quality
+
+- **Source**: Official Wikipedia dumps from Wikimedia Foundation
+- **Processing Date**: 2025-07-31T16:50:25.741062
+- **Pipeline Version**: 1.0.0
+- **Memory Constraints**: Processed within 32GB RAM limits
+- **Quality Assurance**: Multi-stage validation and error handling
+
+## Citation
+
+If you use this dataset in your research, please cite:
+
+```bibtex
+@dataset{wikipedia_af_20250701,
+  title={Wikipedia AF Dataset},
+  author={Wikipedia Monthly Data Processing Pipeline},
+  year={2025},
+  url={https://huggingface.co/datasets/omarkamali/wikipedia-monthly-testing},
+  note={Processed from Wikipedia dump 20250701}
+}
+```
+
+## License
+
+This dataset is released under the same license as Wikipedia content:
+- **Text**: [Creative Commons Attribution-ShareAlike 3.0 Unported License](https://creativecommons.org/licenses/by-sa/3.0/)
+- **Code/Models**: [MIT License](https://opensource.org/licenses/MIT)
+
+## Contact
+
+For questions or issues with this dataset, please open an issue in the repository or contact the maintainers.
+
+---
+
+*Generated automatically by the Wikipedia Monthly Data Processing Pipeline*
+*Processing completed: 2025-07-31T16:50:25.741062*

20250701/af/dataset/10000k.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fcef67660aeeeaf2e0d854d9724664658337a0db84912abb57ee3b1a13cd5996
+size 2255984

20250701/af/dataset/1000k.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d93cc156cf7b53e0f8b612cc81eb0574e528ad614b6f5541ef8ad6385f129726
+size 447476

20250701/af/dataset/5000k.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:67a12e7179faa482e2c0d279e2013b3479b8fbd9a447ef2394f3d46e3c295b05
+size 1395097

20250701/af/dataset/full/full_000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bf27980260c14b4669d73b6c39d52f5fcc54a4ee353c8049a0268de288a996fe
+size 29493901

20250701/af/dataset/full/full_001.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0ed1501ed4333bc40e16bbb546c84f26b46d234e605161cd9289b3df33518c8f
+size 31171738

20250701/af/dataset/full/full_002.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7bacd92631add76f72b3a96bd2d31aa788080eb6423fd451a2b00e43ea1597cd
+size 22028302

20250701/af/dataset/full/full_003.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4648d5d1362765bf3e4768a31b19ff3fe7ccddef67b1a7163992ae349cde91b3
+size 16156360

20250701/af/dataset/full/full_004.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1c57c60ab18eb472387e7f2ce5b245d9cdb214848cff74434e6adad29158baf3
+size 14276973

20250701/af/dataset/full/full_005.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e6be742a869c219955bdc0424caadd74fc5094e2e1d5e1a3481da95b45c3a47e
+size 14233167

20250701/af/dataset/full/full_007.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4713bb6e1f4824d1f5a239cb24384055b1f63d8b792ad9d42a8ced8b10cfcb60
+size 19741456

20250701/af/dataset/full/full_008.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a001bbb8ae24de08c7ca252f3cfdea90802edc9d29c4c8b94af4f8a743fa00ce
+size 23841162

20250701/af/dataset/full/full_009.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:37869e7e2c03f59fe42933f79604260b46534b6ad27e1bd4ac564c93ceb8091c
+size 16703061

20250701/af/dataset/full/full_010.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f5201a5bfad7346b6f75363cadce3bb03e8d8e18ce2bad87f2689c0d701384ed
+size 4909680

20250701/af/dataset/full/full_011.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7b3b0ca71968164d73ea9f6faac5bef097a114d744d27f04009cd87ae1e07094
+size 3815531

20250701/af/dataset/full/full_012.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:adc290382d49bb7e09cb1e55ccd7db33625724e88d9a0f520983d28568691ca3
+size 8103349

20250701/af/dataset/full/full_013.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cdbe39ba98e070622ea259d58f1519d275c8886960bef4290e2ff1afd3d13f72
+size 8073159

20250701/af/dataset/full/full_014.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8fb7c1f929e042575a1f66872286740674b05fd5525b023e007a5dc2284ff50d
+size 7657276

20250701/af/dataset/full/full_015.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e312778fb027aa9d5c31940d034da277149c1e41ffda8242727c43ca95c6dcd4
+size 9908320

20250701/af/dataset/full/full_016.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:990d58fd7f023ebc263d8eda01965d131df849c95dc66163002ca10f01dd03de
+size 11055131

20250701/af/dataset/full/full_017.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ae04db9958d75f0772516b73cc22bc30b3b357c4624f8a53bde9fd5acfea0d9c
+size 13054392

20250701/af/dataset/full/full_018.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0bb48caa6763bc4b5aeda2f33e3b372dac24fb89d7d8613e21ae3a73c56e4ee5
+size 13224173

20250701/af/dataset/full/full_019.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:856140c15142aa9a2056c2886ca0cd7ca6c65be8eaa3110fa2d9aaded64c2b17
+size 9828853

20250701/af/dataset/full/full_020.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e0f8108768a2923e4ea4b169806b7080b003885004d0bbfcee92985d0480cc0c
+size 10830389

20250701/af/dataset/full/full_021.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e4a79fbad99fd61bbf27654c5dfdd222e76fba87723f2d2d3ce33d5c3bcd831e
+size 9520160

20250701/af/dataset/full/full_022.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a114828d963dc6af0ca2547e3c07c8ef87c730dabfa2961c8b430abf1dc3b075
+size 9333909

20250701/af/dataset/full/full_024.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c0b4212987e0dd18023fd848252a14fd532b26c6ac793e6f0bc01df2a460ccdd
+size 9761710

20250701/af/dataset/train/train_010.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8fd555e32a4c7afa127c4df791f5f987dd82142be2d5f8eb615be44ab3c71394
+size 1506782

20250701/af/dataset/train/train_011.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:75749f9eea7914f3f5e41e26170d1d407d7000e7e8bb6c30f472639ec2ce9629
+size 1018886

20250701/af/dataset/train/train_018.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5404f96ab560aed4ba2d9126e2310e448415feac96ee6946f12cbf6a6df0d597
+size 4806670

20250701/af/metadata.json

@@ -0,0 +1,457 @@
+{
+  "total_articles": 124877,
+  "dump_date": "20250701",
+  "processing_completed_at": "2025-07-31T16:50:25.741062",
+  "stage_metadata": {
+    "1": {
+      "dump_path": "temp_20250731_145350/afwiki.xml.bz2",
+      "dump_date": "20250701",
+      "article_count": 124877,
+      "dump_metadata": {
+        "schema_version": "0.11",
+        "case_sensitive": "first-letter",
+        "sitename": "Wikipedia",
+        "base_url": "https://af.wikipedia.org/wiki/Tuisblad"
+      },
+      "temp_dir": "temp_20250731_145350/af",
+      "articles_parquet_files": [
+        "temp_20250731_145350/af/af_articles_part_000.parquet",
+        "temp_20250731_145350/af/af_articles_part_001.parquet",
+        "temp_20250731_145350/af/af_articles_part_002.parquet"
+      ],
+      "completed_at": "2025-07-31T14:57:44.819267",
+      "duration": 232.6967751979828
+    },
+    "2": {
+      "total_articles": 124877,
+      "majority_script": "Latn",
+      "script_distribution": {
+        "Latn": 124877
+      },
+      "normalized_articles_files": [
+        "normalized_articles/af/af_normalized_articles.parquet"
+      ],
+      "completed_at": "2025-07-31T15:26:41.943544",
+      "duration": 1737.2102828025818
+    },
+    "3": {
+      "total_articles": 124877,
+      "trained_models": {
+        "8k": {
+          "vocab_size": 8000,
+          "model_path": "models/af/tokenizer/af_tokenizer_8k.model",
+          "model_prefix": "models/af/tokenizer/af_tokenizer_8k"
+        },
+        "16k": {
+          "vocab_size": 16000,
+          "model_path": "models/af/tokenizer/af_tokenizer_16k.model",
+          "model_prefix": "models/af/tokenizer/af_tokenizer_16k"
+        },
+        "32k": {
+          "vocab_size": 32000,
+          "model_path": "models/af/tokenizer/af_tokenizer_32k.model",
+          "model_prefix": "models/af/tokenizer/af_tokenizer_32k"
+        },
+        "64k": {
+          "vocab_size": 64000,
+          "model_path": "models/af/tokenizer/af_tokenizer_64k.model",
+          "model_prefix": "models/af/tokenizer/af_tokenizer_64k"
+        }
+      },
+      "primary_model_path": "models/af/tokenizer/af_tokenizer_64k.model",
+      "model_path": "models/af/tokenizer/af_tokenizer_64k.model",
+      "training_text_size": 204924573,
+      "sample_tokens": [
+        "▁afrika",
+        "▁px",
+        "▁ligging",
+        "▁van",
+        "▁afrika",
+        "▁op",
+        "▁n",
+        "▁we",
+        "▁reldkaart",
+        "▁oppervlakte",
+        "▁km",
+        "▁de",
+        "▁bevolking",
+        "▁miljard",
+        "▁de",
+        "▁bevolkings",
+        "digtheid",
+        "▁km",
+        "▁lande",
+        "▁en"
+      ],
+      "sample_ids": [
+        234,
+        387,
+        3767,
+        31,
+        234,
+        86,
+        18,
+        221,
+        61247,
+        1921,
+        588,
+        75,
+        1106,
+        4654,
+        75,
+        2973,
+        3542,
+        588,
+        1784,
+        36
+      ],
+      "completed_at": "2025-07-31T15:29:18.234762",
+      "duration": 156.2300901412964
+    },
+    "4": {
+      "total_articles": 124877,
+      "training_samples": 49935,
+      "eval_samples": 999,
+      "model_path": "models/af/af_dictionary.json",
+      "dictionary_size": 484292,
+      "word_freq_size": 15051,
+      "final_accuracy": 1.0,
+      "final_f1": 1.0,
+      "avg_confidence": 0.9844343164510623,
+      "high_confidence_articles": 124660,
+      "high_confidence_ratio": 0.9982622900934519,
+      "training_cycles": 3,
+      "completed_at": "2025-07-31T16:02:59.733485",
+      "duration": 2022.2301919460297
+    },
+    "5": {
+      "subword_ngram_models": {
+        "2": {
+          "unique_ngrams": 988607,
+          "total_ngrams": 43074327,
+          "min_frequency": 5,
+          "model_file": "models/af/subword_ngram/af_2gram_model.parquet",
+          "metadata_file": "models/af/subword_ngram/af_2gram_metadata.json",
+          "top_ngrams_file": "models/af/subword_ngram/af_2gram_top1000.parquet"
+        },
+        "3": {
+          "unique_ngrams": 1933919,
+          "total_ngrams": 42949586,
+          "min_frequency": 3,
+          "model_file": "models/af/subword_ngram/af_3gram_model.parquet",
+          "metadata_file": "models/af/subword_ngram/af_3gram_metadata.json",
+          "top_ngrams_file": "models/af/subword_ngram/af_3gram_top1000.parquet"
+        },
+        "4": {
+          "unique_ngrams": 3175168,
+          "total_ngrams": 42824846,
+          "min_frequency": 2,
+          "model_file": "models/af/subword_ngram/af_4gram_model.parquet",
+          "metadata_file": "models/af/subword_ngram/af_4gram_metadata.json",
+          "top_ngrams_file": "models/af/subword_ngram/af_4gram_top1000.parquet"
+        }
+      },
+      "subword_markov_chains": {
+        "1": {
+          "context_size": 1,
+          "unique_contexts": 63465,
+          "total_transitions": 43074327,
+          "transitions_file": "models/af/subword_markov/af_markov1_transitions.parquet",
+          "metadata_file": "models/af/subword_markov/af_markov1_metadata.json"
+        },
+        "2": {
+          "context_size": 2,
+          "unique_contexts": 7502582,
+          "total_transitions": 42949586,
+          "transitions_file": "models/af/subword_markov/af_markov2_transitions.parquet",
+          "metadata_file": "models/af/subword_markov/af_markov2_metadata.json"
+        },
+        "3": {
+          "context_size": 3,
+          "unique_contexts": 21032987,
+          "total_transitions": 42824846,
+          "transitions_file": "models/af/subword_markov/af_markov3_transitions.parquet",
+          "metadata_file": "models/af/subword_markov/af_markov3_metadata.json"
+        }
+      },
+      "word_ngram_models": {
+        "2": {
+          "unique_ngrams": 1312897,
+          "total_ngrams": 38487033,
+          "min_frequency": 3,
+          "model_file": "models/af/word_ngram/af_2gram_model.parquet",
+          "metadata_file": "models/af/word_ngram/af_2gram_metadata.json",
+          "top_ngrams_file": "models/af/word_ngram/af_2gram_top1000.parquet"
+        },
+        "3": {
+          "unique_ngrams": 3125001,
+          "total_ngrams": 38362292,
+          "min_frequency": 2,
+          "model_file": "models/af/word_ngram/af_3gram_model.parquet",
+          "metadata_file": "models/af/word_ngram/af_3gram_metadata.json",
+          "top_ngrams_file": "models/af/word_ngram/af_3gram_top1000.parquet"
+        },
+        "4": {
+          "unique_ngrams": 2527228,
+          "total_ngrams": 38237555,
+          "min_frequency": 2,
+          "model_file": "models/af/word_ngram/af_4gram_model.parquet",
+          "metadata_file": "models/af/word_ngram/af_4gram_metadata.json",
+          "top_ngrams_file": "models/af/word_ngram/af_4gram_top1000.parquet"
+        }
+      },
+      "word_markov_chains": {
+        "1": {
+          "context_size": 1,
+          "unique_contexts": 823324,
+          "total_transitions": 38487033,
+          "transitions_file": "models/af/word_markov/af_markov1_transitions.parquet",
+          "metadata_file": "models/af/word_markov/af_markov1_metadata.json"
+        },
+        "2": {
+          "context_size": 2,
+          "unique_contexts": 8320124,
+          "total_transitions": 38362292,
+          "transitions_file": "models/af/word_markov/af_markov2_transitions.parquet",
+          "metadata_file": "models/af/word_markov/af_markov2_metadata.json"
+        },
+        "3": {
+          "context_size": 3,
+          "unique_contexts": 19534054,
+          "total_transitions": 38237555,
+          "transitions_file": "models/af/word_markov/af_markov3_transitions.parquet",
+          "metadata_file": "models/af/word_markov/af_markov3_metadata.json"
+        }
+      },
+      "subword_totals": {
+        "total_unique_ngrams": 6097694,
+        "total_ngrams": 128848759,
+        "total_markov_contexts": 28599034,
+        "total_transitions": 128848759
+      },
+      "word_totals": {
+        "total_unique_ngrams": 6965126,
+        "total_ngrams": 115086880,
+        "total_markov_contexts": 28677502,
+        "total_transitions": 115086880
+      },
+      "saved_files": {
+        "subword_2gram": "models/af/subword_ngram/af_2gram_model.parquet",
+        "subword_2gram_metadata": "models/af/subword_ngram/af_2gram_metadata.json",
+        "subword_2gram_top": "models/af/subword_ngram/af_2gram_top1000.parquet",
+        "subword_3gram": "models/af/subword_ngram/af_3gram_model.parquet",
+        "subword_3gram_metadata": "models/af/subword_ngram/af_3gram_metadata.json",
+        "subword_3gram_top": "models/af/subword_ngram/af_3gram_top1000.parquet",
+        "subword_4gram": "models/af/subword_ngram/af_4gram_model.parquet",
+        "subword_4gram_metadata": "models/af/subword_ngram/af_4gram_metadata.json",
+        "subword_4gram_top": "models/af/subword_ngram/af_4gram_top1000.parquet",
+        "subword_markov1": "models/af/subword_markov/af_markov1_transitions.parquet",
+        "subword_markov1_metadata": "models/af/subword_markov/af_markov1_metadata.json",
+        "subword_markov2": "models/af/subword_markov/af_markov2_transitions.parquet",
+        "subword_markov2_metadata": "models/af/subword_markov/af_markov2_metadata.json",
+        "subword_markov3": "models/af/subword_markov/af_markov3_transitions.parquet",
+        "subword_markov3_metadata": "models/af/subword_markov/af_markov3_metadata.json",
+        "word_2gram": "models/af/word_ngram/af_2gram_model.parquet",
+        "word_2gram_metadata": "models/af/word_ngram/af_2gram_metadata.json",
+        "word_2gram_top": "models/af/word_ngram/af_2gram_top1000.parquet",
+        "word_3gram": "models/af/word_ngram/af_3gram_model.parquet",
+        "word_3gram_metadata": "models/af/word_ngram/af_3gram_metadata.json",
+        "word_3gram_top": "models/af/word_ngram/af_3gram_top1000.parquet",
+        "word_4gram": "models/af/word_ngram/af_4gram_model.parquet",
+        "word_4gram_metadata": "models/af/word_ngram/af_4gram_metadata.json",
+        "word_4gram_top": "models/af/word_ngram/af_4gram_top1000.parquet",
+        "word_markov1": "models/af/word_markov/af_markov1_transitions.parquet",
+        "word_markov1_metadata": "models/af/word_markov/af_markov1_metadata.json",
+        "word_markov2": "models/af/word_markov/af_markov2_transitions.parquet",
+        "word_markov2_metadata": "models/af/word_markov/af_markov2_metadata.json",
+        "word_markov3": "models/af/word_markov/af_markov3_transitions.parquet",
+        "word_markov3_metadata": "models/af/word_markov/af_markov3_metadata.json"
+      },
+      "completed_at": "2025-07-31T16:44:59.441004",
+      "duration": 2536.534306049347
+    },
+    "6": {
+      "total_articles": 124877,
+      "output_files": [
+        "scored_articles/af/af_scored_articles_part_000.parquet",
+        "scored_articles/af/af_scored_articles_part_001.parquet",
+        "scored_articles/af/af_scored_articles_part_002.parquet",
+        "scored_articles/af/af_scored_articles_part_003.parquet",
+        "scored_articles/af/af_scored_articles_part_004.parquet"
+      ],
+      "score_statistics": {
+        "mean_tfidf": {
+          "mean": 0.05124321167467498,
+          "min": 0.0005799509719341689,
+          "max": 6.829809753400186,
+          "count": 124877
+        },
+        "overall_novelty": {
+          "mean": 0.0,
+          "min": 0.0,
+          "max": 0.0,
+          "count": 124877
+        },
+        "hapax_legomena_ratio": {
+          "mean": 0.72602454849505,
+          "min": 0.0,
+          "max": 1.0,
+          "count": 124877
+        },
+        "total_tokens": {
+          "mean": 345.9340310865892,
+          "min": 1,
+          "max": 39463,
+          "count": 124877
+        },
+        "lexical_diversity": {
+          "mean": 0.9843732208519338,
+          "min": 0.0,
+          "max": 1.0,
+          "count": 124877
+        },
+        "unique_tokens": {
+          "mean": 156.5526237818013,
+          "min": 1,
+          "max": 6644,
+          "count": 124877
+        },
+        "moving_average_ttr": {
+          "mean": 0.7046797727093798,
+          "min": 0.02930232558139535,
+          "max": 1.0,
+          "count": 124877
+        },
+        "type_token_ratio": {
+          "mean": 0.6290323290908232,
+          "min": 0.06140350877192982,
+          "max": 1.0,
+          "count": 124877
+        },
+        "max_tfidf": {
+          "mean": 0.24166506861657502,
+          "min": 0.013544769437258434,
+          "max": 6.829809753400186,
+          "count": 124877
+        },
+        "overall_representativeness": {
+          "mean": 0.0,
+          "min": 0.0,
+          "max": 0.0,
+          "count": 124877
+        }
+      },
+      "ngram_models_used": [],
+      "batch_size": 10000,
+      "completed_at": "2025-07-31T16:49:33.700147",
+      "duration": 257.318391084671
+    },
+    "7": {
+      "samples": {
+        "full": [
+          "samples/af/af_sample_fullk_part_000.parquet",
+          "samples/af/af_sample_fullk_part_001.parquet",
+          "samples/af/af_sample_fullk_part_002.parquet",
+          "samples/af/af_sample_fullk_part_003.parquet",
+          "samples/af/af_sample_fullk_part_004.parquet",
+          "samples/af/af_sample_fullk_part_005.parquet",
+          "samples/af/af_sample_fullk_part_006.parquet",
+          "samples/af/af_sample_fullk_part_007.parquet",
+          "samples/af/af_sample_fullk_part_008.parquet",
+          "samples/af/af_sample_fullk_part_009.parquet",
+          "samples/af/af_sample_fullk_part_010.parquet",
+          "samples/af/af_sample_fullk_part_011.parquet",
+          "samples/af/af_sample_fullk_part_012.parquet",
+          "samples/af/af_sample_fullk_part_013.parquet",
+          "samples/af/af_sample_fullk_part_014.parquet",
+          "samples/af/af_sample_fullk_part_015.parquet",
+          "samples/af/af_sample_fullk_part_016.parquet",
+          "samples/af/af_sample_fullk_part_017.parquet",
+          "samples/af/af_sample_fullk_part_018.parquet",
+          "samples/af/af_sample_fullk_part_019.parquet",
+          "samples/af/af_sample_fullk_part_020.parquet",
+          "samples/af/af_sample_fullk_part_021.parquet",
+          "samples/af/af_sample_fullk_part_022.parquet",
+          "samples/af/af_sample_fullk_part_023.parquet",
+          "samples/af/af_sample_fullk_part_024.parquet"
+        ],
+        "train": [
+          "samples/af/af_train_part_000.parquet",
+          "samples/af/af_train_part_001.parquet",
+          "samples/af/af_train_part_002.parquet",
+          "samples/af/af_train_part_003.parquet",
+          "samples/af/af_train_part_004.parquet",
+          "samples/af/af_train_part_005.parquet",
+          "samples/af/af_train_part_006.parquet",
+          "samples/af/af_train_part_007.parquet",
+          "samples/af/af_train_part_008.parquet",
+          "samples/af/af_train_part_009.parquet",
+          "samples/af/af_train_part_010.parquet",
+          "samples/af/af_train_part_011.parquet",
+          "samples/af/af_train_part_012.parquet",
+          "samples/af/af_train_part_013.parquet",
+          "samples/af/af_train_part_014.parquet",
+          "samples/af/af_train_part_015.parquet",
+          "samples/af/af_train_part_016.parquet",
+          "samples/af/af_train_part_017.parquet",
+          "samples/af/af_train_part_018.parquet",
+          "samples/af/af_train_part_019.parquet",
+          "samples/af/af_train_part_020.parquet",
+          "samples/af/af_train_part_021.parquet",
+          "samples/af/af_train_part_022.parquet",
+          "samples/af/af_train_part_023.parquet",
+          "samples/af/af_train_part_024.parquet"
+        ],
+        "1000k": [
+          "samples/af/af_sample_1000k_part_000.parquet"
+        ],
+        "5000k": [
+          "samples/af/af_sample_5000k_part_000.parquet"
+        ],
+        "10000k": [
+          "samples/af/af_sample_10000k_part_000.parquet"
+        ]
+      },
+      "sample_counts": {
+        "1000": 50,
+        "5000": 250,
+        "10000": 500,
+        "full": 124877,
+        "train": 124877
+      },
+      "total_articles_processed": 124877,
+      "processing_completed_at": "2025-07-31T16:49:52.374138",
+      "duration": 18.365309953689575
+    },
+    "8": {
+      "statistics_report": "statistics/af/af_statistics_report.json",
+      "corpus_statistics": {
+        "total_articles": 124877,
+        "total_tokens": 43199204,
+        "vocabulary_size": 63466,
+        "zipf_adherence": "excellent"
+      },
+      "sample_coverage": {
+        "full": {
+          "overall_coverage": 0.0,
+          "vocabulary_coverage": 1.0
+        },
+        "1000": {
+          "overall_coverage": 0.0,
+          "vocabulary_coverage": 0.15731257681278166
+        },
+        "5000": {
+          "overall_coverage": 0.0,
+          "vocabulary_coverage": 0.31859578356915513
+        },
+        "10000": {
+          "overall_coverage": 0.0,
+          "vocabulary_coverage": 0.41441086566035357
+        }
+      },
+      "processing_completed_at": "2025-07-31T16:50:24.744598",
+      "duration": 32.78457522392273
+    }
+  },
+  "pipeline_version": "1.0.0"
+}

20250701/af/models/subword_markov/af_markov1_metadata.json

@@ -0,0 +1,8 @@
+{
+  "context_size": 1,
+  "unique_contexts": 63465,
+  "total_transitions": 43074327,
+  "language_code": "af",
+  "model_type": "subword",
+  "created_at": "2025-07-31T16:40:37.063668"
+}

20250701/af/models/subword_markov/af_markov2_metadata.json

@@ -0,0 +1,8 @@
+{
+  "context_size": 2,
+  "unique_contexts": 7502582,
+  "total_transitions": 42949586,
+  "language_code": "af",
+  "model_type": "subword",
+  "created_at": "2025-07-31T16:41:22.070623"
+}

20250701/af/models/subword_markov/af_markov3_metadata.json

@@ -0,0 +1,8 @@
+{
+  "context_size": 3,
+  "unique_contexts": 21032987,
+  "total_transitions": 42824846,
+  "language_code": "af",
+  "model_type": "subword",
+  "created_at": "2025-07-31T16:42:36.692158"
+}

20250701/af/models/subword_ngram/af_2gram_metadata.json

@@ -0,0 +1,9 @@
+{
+  "n": 2,
+  "total_ngrams": 43074327,
+  "unique_ngrams": 988607,
+  "min_frequency": 5,
+  "language_code": "af",
+  "model_type": "subword",
+  "created_at": "2025-07-31T16:40:08.319325"
+}

20250701/af/models/subword_ngram/af_3gram_metadata.json

@@ -0,0 +1,9 @@
+{
+  "n": 3,
+  "total_ngrams": 42949586,
+  "unique_ngrams": 1933919,
+  "min_frequency": 3,
+  "language_code": "af",
+  "model_type": "subword",
+  "created_at": "2025-07-31T16:40:13.616357"
+}

20250701/af/models/subword_ngram/af_4gram_metadata.json

@@ -0,0 +1,9 @@
+{
+  "n": 4,
+  "total_ngrams": 42824846,
+  "unique_ngrams": 3175168,
+  "min_frequency": 2,
+  "language_code": "af",
+  "model_type": "subword",
+  "created_at": "2025-07-31T16:40:22.238911"
+}

20250701/af/models/vocabulary/af_dictionary_metadata.json

@@ -0,0 +1,40 @@
+{
+  "language_code": "af",
+  "total_articles": 124877,
+  "training_samples": 49935,
+  "eval_samples": 999,
+  "dictionary_size": 484292,
+  "word_freq_size": 15051,
+  "training_report": {
+    "cycle_reports": [
+      {
+        "f1": 1.0,
+        "accuracy": 1.0,
+        "precision": 1.0,
+        "recall": 1.0,
+        "confusion": 0,
+        "confidence_margin": 1.0,
+        "confusion_matrix": [],
+        "total_samples": 49558,
+        "removed_samples": 0
+      },
+      {
+        "f1": 1.0,
+        "accuracy": 1.0,
+        "precision": 1.0,
+        "recall": 1.0,
+        "confusion": 0,
+        "confidence_margin": 1.0,
+        "confusion_matrix": [],
+        "total_samples": 49558,
+        "removed_samples": 0
+      }
+    ]
+  },
+  "final_accuracy": 1.0,
+  "final_f1": 1.0,
+  "confidence_threshold": 0.5,
+  "confidence_margin": 0.3,
+  "training_cycles": 3,
+  "completed_at": "2025-07-31T16:02:59.004063"
+}

20250701/af/models/word_markov/af_markov1_metadata.json

@@ -0,0 +1,8 @@
+{
+  "context_size": 1,
+  "unique_contexts": 823324,
+  "total_transitions": 38487033,
+  "language_code": "af",
+  "model_type": "word",
+  "created_at": "2025-07-31T16:43:12.538748"
+}

20250701/af/models/word_markov/af_markov2_metadata.json

@@ -0,0 +1,8 @@
+{
+  "context_size": 2,
+  "unique_contexts": 8320124,
+  "total_transitions": 38362292,
+  "language_code": "af",
+  "model_type": "word",
+  "created_at": "2025-07-31T16:43:52.328902"
+}

20250701/af/models/word_markov/af_markov3_metadata.json

@@ -0,0 +1,8 @@
+{
+  "context_size": 3,
+  "unique_contexts": 19534054,
+  "total_transitions": 38237555,
+  "language_code": "af",
+  "model_type": "word",
+  "created_at": "2025-07-31T16:44:56.903211"
+}

20250701/af/models/word_ngram/af_2gram_metadata.json

@@ -0,0 +1,9 @@
+{
+  "n": 2,
+  "total_ngrams": 38487033,
+  "unique_ngrams": 1312897,
+  "min_frequency": 3,
+  "language_code": "af",
+  "model_type": "word",
+  "created_at": "2025-07-31T16:42:42.416761"
+}

20250701/af/models/word_ngram/af_3gram_metadata.json

@@ -0,0 +1,9 @@
+{
+  "n": 3,
+  "total_ngrams": 38362292,
+  "unique_ngrams": 3125001,
+  "min_frequency": 2,
+  "language_code": "af",
+  "model_type": "word",
+  "created_at": "2025-07-31T16:42:49.829509"
+}

20250701/af/models/word_ngram/af_3gram_model.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4fa503a945674b3707acc6224b165da87aefe600dd2e2f5d54ad8ce57c2b0ce3
+size 54238615

20250701/af/models/word_ngram/af_4gram_metadata.json

@@ -0,0 +1,9 @@
+{
+  "n": 4,
+  "total_ngrams": 38237555,
+  "unique_ngrams": 2527228,
+  "min_frequency": 2,
+  "language_code": "af",
+  "model_type": "word",
+  "created_at": "2025-07-31T16:42:57.012278"
+}

20250701/af/models/word_ngram/af_4gram_model.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3e04ee684ca4c3c2ac49756b1cb67b308fb9a13c457b3af6eef5b13311ce3379
+size 47134196