Upload folder using huggingface_hub

README.md

@@ -6,6 +6,7 @@ task_categories:
 - text-retrieval
 - question-answering
 - text-generation
+- text-classification
 size_categories:
 - 1K<n<10K
 tags:
@@ -15,9 +16,10 @@ tags:
 - rag
 - knowledge-base
 - production-ready
-- cleaned
 - high-quality
-pretty_name: BYU-Idaho Web Content (Production)
+- categorized
+- scored
+pretty_name: BYU-Idaho Web Content (Enhanced)
 dataset_info:
   features:
   - name: index
@@ -32,68 +34,89 @@ dataset_info:
     dtype: string
   - name: content
     dtype: string
+  - name: category
+    dtype: string
+  - name: content_type
+    dtype: string
+  - name: quality_score
+    dtype: int64
+  - name: reading_level
+    dtype: float64
   splits:
   - name: train
-    num_examples: 2666
+    num_examples: 2448
 ---
 
-# BYU-Idaho Web Content Dataset (Production)
+# BYU-Idaho Web Content Dataset (Enhanced)
 
-**Production-ready**, high-quality web content from BYU-Idaho's official website. Extensively cleaned and filtered for optimal RAG, search, and training applications.
+**Ultra-high-quality**, enriched web content from BYU-Idaho with category classification, quality scoring, and advanced filtering. Enterprise-ready for production RAG, search, and AI applications.
 
 ## Dataset Description
 
-**Records:** 2,666 production-quality pages
+**Records:** 2,448 ultra-high-quality pages
 **Source:** byui.edu and subdomains
-**Format:** Markdown-formatted text
-**Version:** 2.0.0 (Production)
-**Quality:** 34.9% of raw data filtered out for quality
+**Format:** Markdown with enriched metadata
+**Version:** 3.0.0 (Enhanced)
+**Quality:** 40.2% of raw data filtered, avg score 91.5/100
 **Last Updated:** December 2025
 
+## Key Features
+
+✨ **Category Classification** - 14 categories (Academics, Admissions, Student Life, etc.)
+✨ **Content Type Detection** - 7 types (informational, guide, FAQ, procedure, policy, etc.)
+✨ **Quality Scoring** - 0-100 score based on multiple quality factors
+✨ **Reading Level** - Flesch-Kincaid grade level for each page
+✨ **Advanced Filtering** - Faculty bios, repetitive lists, low-quality content removed
+
 ## Quality Assurance
 
-### Comprehensive Filtering Applied
-
-**Content Quality Filters:**
-- ✅ Minimum 200 characters per page (no stubs or redirects)
-- ✅ 0% empty titles (100% have titles)
-- ✅ 0% empty topics (100% have H1 headings)
-- ✅ HTTP 200 only (no 404 errors)
-- ✅ Unique content via SHA256 deduplication
-
-**Temporal Content Removed:**
-- ❌ Academic calendar pages (`/academic-calendar/`)
-- ❌ Financial deadline pages (`/financial-deadlines/`)
-- ❌ Event pages with dates in URLs (`/2024-`, `/2025-`)
-- ❌ Job postings (`/help-wanted/`, `/job-postings/`)
-- ❌ Speeches archive (`/speeches/`)
-- ❌ Newsroom articles (`/newsroom/`)
-
-**Repetitive Content Removed:**
-- ❌ "Last Day to Withdraw" variations (100+ duplicates)
-- ❌ "Late Charge" notices
-- ❌ Deadline pages
-- ❌ Calendar event duplicates
+### Enhanced Filtering Pipeline
+
+**Stage 1: Base Filters** (from v2.0)
+- ❌ Academic calendars and financial deadlines
+- ❌ Temporal content (events with dates in URLs)
+- ❌ Job postings and newsroom articles
+- ❌ 404 errors and very short pages (<200 chars)
+
+**Stage 2: Advanced Filters** (NEW in v3.0)
+- ❌ **Faculty bio stubs** - 82 removed (short CV-only pages)
+- ❌ **Repetitive list pages** - 149 removed (>65% word repetition)
+- ❌ **Low-quality content** - Pages scoring <40/100 removed
+
+**Stage 3: Quality Scoring** (NEW in v3.0)
+- ✅ Content length optimization (penalty for too short/long)
+- ✅ Metadata completeness check
+- ✅ Structural quality (headers, lists, sentences)
+- ✅ Uniqueness ratio (penalize repetitive text)
+- ✅ **Average score: 91.5/100** (range: 50-100)
 
 ### Cleaning Statistics
 
 | Stage | Count | Removed | % Removed |
 |-------|-------|---------|-----------|
 | Original crawled pages | 4,097 | - | - |
-| After deduplication | 3,440 | 657 | 16.0% |
-| After quality filters | **2,666** | 774 | 18.9% |
-| **Total removed** | - | **1,431** | **34.9%** |
+| After v2.0 filters | 2,679 | 1,418 | 34.6% |
+| Removed faculty bios | 2,597 | 82 | 2.0% |
+| Removed repetitive lists | 2,448 | 149 | 3.6% |
+| **Final enhanced dataset** | **2,448** | **1,649** | **40.2%** |
 
 ## Dataset Structure
 
 ### Fields
 
-- **index** (`int64`): Sequential identifier (1-2666)
+**Core Fields:**
+- **index** (`int64`): Sequential identifier (1-2448)
 - **url** (`string`): Source URL
-- **title** (`string`): Cleaned page title (no duplicate suffixes)
+- **title** (`string`): Cleaned page title
 - **topic** (`string`): Main heading (H1)
 - **meta_description** (`string`): SEO description
-- **content** (`string`): Full page content in Markdown (avg 2,526 chars)
+- **content** (`string`): Full page content in Markdown (avg 2,230 chars)
+
+**NEW Enrichment Fields:**
+- **category** (`string`): Primary category (14 options)
+- **content_type** (`string`): Content type (7 options)
+- **quality_score** (`int64`): Quality score 0-100 (avg: 91.5)
+- **reading_level** (`float64`): Flesch-Kincaid grade level (avg: 14.4)
 
 ### Example
 
@@ -103,146 +126,199 @@ dataset_info:
   "url": "https://www.byui.edu/",
   "title": "Home - BYU-Idaho",
   "topic": "Christ-Centered. Student-Focused.",
-  "meta_description": "Welcome to BYU-Idaho's home page! Learn about BYU-Idaho's mission...",
-  "content": "# Christ-Centered. Student-Focused.\n\nBYU Idaho Icons\n\nCelebrating 25 Years..."
+  "meta_description": "Welcome to BYU-Idaho's home page...",
+  "content": "# Christ-Centered. Student-Focused.\n\nBYU Idaho Icons...",
+  "category": "About BYU-Idaho",
+  "content_type": "overview",
+  "quality_score": 95,
+  "reading_level": 12.3
 }
 ```
 
+## Categories (14 total)
+
+| Category | Count | % | Description |
+|----------|-------|---|-------------|
+| **General** | 1,951 | 79.7% | General university content |
+| **Academics** | 151 | 6.2% | Programs, majors, degrees |
+| **Employment** | 113 | 4.6% | HR, jobs, faculty resources |
+| **Student Life** | 62 | 2.5% | Housing, activities, campus life |
+| **Financial Aid** | 54 | 2.2% | Scholarships, grants, loans |
+| **Career Services** | 41 | 1.7% | Internships, job placement |
+| **Admissions** | 37 | 1.5% | Application, requirements |
+| **Registration & Records** | 12 | 0.5% | Registration, transcripts |
+| **About BYU-Idaho** | 11 | 0.4% | Mission, history, leadership |
+| **Academic Support** | 8 | 0.3% | Tutoring, success services |
+| **Library** | 3 | 0.1% | Library services |
+| **Athletics & Recreation** | 3 | 0.1% | Sports, fitness |
+| **International** | 1 | 0.0% | Study abroad, int'l students |
+| **Policies & Compliance** | 1 | 0.0% | Rules, compliance |
+
+## Content Types (7 total)
+
+| Type | Count | % | Description |
+|------|-------|---|-------------|
+| **informational** | 2,142 | 87.5% | General information pages |
+| **contact** | 125 | 5.1% | Contact information |
+| **guide** | 94 | 3.8% | Long-form guides (>5000 chars, 3+ sections) |
+| **faq** | 35 | 1.4% | Frequently asked questions |
+| **procedure** | 28 | 1.1% | Step-by-step procedures |
+| **overview** | 14 | 0.6% | Overview/about pages |
+| **policy** | 10 | 0.4% | Policy documents |
+
+## Quality Metrics
+
+| Metric | Value | Notes |
+|--------|-------|-------|
+| **Total pages** | 2,448 | 40.2% filtered from raw |
+| **Avg quality score** | **91.5/100** | Range: 50-100 |
+| **Min quality score** | 50 | No low-quality pages |
+| **Avg reading level** | 14.4 grade | College-appropriate |
+| **Avg content length** | 2,230 chars | Substantial content |
+| **Empty titles** | 0% | 100% complete |
+| **Empty topics** | 0% | 100% complete |
+
 ## Usage
 
-### Load the Dataset
+### Basic Loading
 
 ```python
 from datasets import load_dataset
 
-dataset = load_dataset("BYU-Idaho/Web-Content")
-print(f"Production records: {len(dataset['train'])}")  # 2,666
+dataset = load_dataset("BYU-Idaho/Web-Content")['train']
+print(f"Total pages: {len(dataset)}")  # 2,448
+```
+
+### Filter by Category
+
+```python
+# Get all Admissions pages
+admissions = dataset.filter(lambda x: x['category'] == 'Admissions')
+
+# Get all Academic pages
+academics = dataset.filter(lambda x: x['category'] == 'Academics')
 ```
 
-### Semantic Search Example
+### Filter by Quality Score
 
 ```python
-# Filter by topic area
-scholarships = dataset['train'].filter(
-    lambda x: 'scholarship' in x['content'].lower()
-)
-
-# Find admissions content
-admissions = dataset['train'].filter(
-    lambda x: 'admissions.byui.edu' in x['url']
-)
+# Get only highest quality pages (95+)
+top_quality = dataset.filter(lambda x: x['quality_score'] >= 95)
+
+# Get pages suitable for general audiences (lower reading level)
+accessible = dataset.filter(lambda x: x['reading_level'] <= 12.0)
+```
+
+### Filter by Content Type
+
+```python
+# Get all guides
+guides = dataset.filter(lambda x: x['content_type'] == 'guide')
+
+# Get all FAQs
+faqs = dataset.filter(lambda x: x['content_type'] == 'faq')
+
+# Get procedural content
+procedures = dataset.filter(lambda x: x['content_type'] == 'procedure')
 ```
 
-### RAG Application
+### RAG Application with Filtering
 
 ```python
 from datasets import load_dataset
 from sentence_transformers import SentenceTransformer
 
-# Load dataset
+# Load high-quality pages only
 ds = load_dataset("BYU-Idaho/Web-Content")['train']
+high_quality = ds.filter(lambda x: x['quality_score'] >= 80)
 
-# Generate embeddings
+# Generate embeddings for retrieval
 model = SentenceTransformer('all-MiniLM-L6-v2')
-embeddings = model.encode(ds['content'])
+embeddings = model.encode(high_quality['content'])
 
-# Use for retrieval-augmented generation
+# Use for context retrieval in RAG
 ```
 
-## Content Categories
-
-The dataset includes high-quality evergreen content about:
-
-- **Academics** - Programs, majors, degrees, courses
-- **Admissions** - Application process, requirements, transfer credits
-- **Student Life** - Housing, activities, campus resources
-- **Financial Aid** - Scholarships, grants, loans (general info only)
-- **Campus Services** - Libraries, tutoring, career services
-- **About BYU-Idaho** - Mission, history, leadership
-- **Policies** - Academic policies, student conduct
-- **Research** - Faculty research, student opportunities
-
-## Excluded Content
-
-**Temporal content** (changes frequently):
-- Academic calendars and deadlines
-- Event announcements
-- News articles
-- Job postings
-- Speeches archive
-
-**Low-value content:**
-- 404 error pages
-- Redirect pages
-- Stub pages (<200 chars)
-- Duplicate content (same text, different URLs)
-- Empty pages
-
-**Media and navigation:**
-- Images, videos, PDFs
-- Navigation menus
-- UI elements
-- External links
-
-## Data Collection & Processing
-
-**Crawler:** Custom Python crawler (Playwright + httpx)
-**Scope:** Internal domain only (`*.byui.edu`)
-**Method:** Robots.txt compliant, sitemap-based discovery
-**Processing:** HTML → Markdown with noise removal
-**Normalization:** Full ASCII (Unicode quotes removed)
-**Deduplication:** SHA256 content hashing
-**Filtering:** Multi-stage quality pipeline
+### Category-Specific RAG
 
-## Quality Metrics
+```python
+# Build domain-specific retrievers
+admissions_data = ds.filter(lambda x: x['category'] == 'Admissions')
+financial_data = ds.filter(lambda x: x['category'] == 'Financial Aid')
 
-| Metric | Value | Notes |
-|--------|-------|-------|
-| Total pages | 2,666 | 34.9% filtered from raw data |
-| Avg content length | 2,526 chars | Up from 2,016 in raw data |
-| Empty titles | 0% | 100% have titles |
-| Empty topics | 0% | 100% have H1 headings |
-| Empty descriptions | 0.64% | 17 pages only |
-| Min content length | 200 chars | Enforced minimum |
-| Unique content | 100% | SHA256 deduplicated |
+# Route queries to appropriate category retriever
+```
 
 ## Use Cases
 
-✅ **Retrieval-Augmented Generation (RAG)**
-High-quality context for AI assistants answering BYU-Idaho questions
+### ✅ Retrieval-Augmented Generation (RAG)
+- **Category-aware routing** - Direct queries to relevant content
+- **Quality filtering** - Use only high-scoring pages for responses
+- **Content-type matching** - Return FAQs for questions, guides for how-to queries
 
-✅ **Semantic Search**
-Building internal search engines with embeddings
+### ✅ Semantic Search
+- **Multi-field search** - Search by category, content type, quality
+- **Reading level filtering** - Adjust complexity for audience
+- **Quality-ranked results** - Surface best content first
 
-✅ **Knowledge Base**
-Foundation for institutional knowledge systems
+### ✅ Chatbot Training
+- **FAQ extraction** - 35 pre-labeled FAQ pages
+- **Procedure training** - 28 step-by-step procedures
+- **Category-based intents** - Map user queries to 14 categories
 
-✅ **Fine-tuning**
-Training data for domain-specific language models
+### ✅ Content Analysis
+- **Quality assessment** - Identify low-performing pages
+- **Category gaps** - Find underrepresented topics
+- **Reading level analysis** - Optimize content accessibility
 
-✅ **Content Analysis**
-Studying university web content and information architecture
-
-✅ **Chatbot Training**
-Building conversational AI for student services
-
-## Advantages Over Raw Data
-
-1. **35% smaller** - Only high-quality evergreen content
-2. **Higher information density** - Avg 2,526 chars vs 2,016
-3. **No temporal noise** - No outdated calendars or deadlines
-4. **No duplicates** - Unique content only
-5. **Complete records** - 0% empty titles/topics
-6. **Better RAG performance** - Relevant answers, no calendar spam
-7. **Reduced hallucination** - No contradictory deadline dates
+### ✅ Fine-tuning
+- **High-quality training data** - Avg 91.5/100 score
+- **Diverse categories** - Balanced domain coverage
+- **Structured content** - Headers, lists, clear organization
 
 ## Version History
 
+**v3.0.0 (Enhanced)** - December 2025
+- Added category classification (14 categories)
+- Added content type detection (7 types)
+- Added quality scoring (0-100)
+- Added reading level calculation
+- Removed 82 faculty bio stubs
+- Removed 149 repetitive list pages
+- 2,448 ultra-high-quality pages
+
 **v2.0.0 (Production)** - December 2025
 - Initial production release
 - 2,666 high-quality pages
-- Comprehensive quality filtering
-- 34.9% of raw data removed
+- 34.9% filtering applied
+
+**v1.0.0 (Deduplicated)** - December 2025
+- 3,442 deduplicated pages
+- Basic quality filtering
+
+## Advantages Over Previous Versions
+
+| Feature | v1.0 | v2.0 | v3.0 (Enhanced) |
+|---------|------|------|-----------------|
+| **Records** | 3,442 | 2,666 | **2,448** |
+| **Avg quality** | Unknown | Unknown | **91.5/100** |
+| **Categories** | ❌ | ❌ | **✅ 14 categories** |
+| **Content types** | ❌ | ❌ | **✅ 7 types** |
+| **Quality scores** | ❌ | ❌ | **✅ 0-100 scale** |
+| **Reading levels** | ❌ | ❌ | **✅ Flesch-Kincaid** |
+| **Faculty bios removed** | ❌ | ❌ | **✅ 82 removed** |
+| **List pages removed** | ❌ | ❌ | **✅ 149 removed** |
+
+## Quality Guarantee
+
+Every page in this dataset:
+- ✅ Scores ≥50/100 on quality metrics
+- ✅ Contains ≥200 characters of content
+- ✅ Has both title and topic (H1)
+- ✅ Is unique (SHA256 deduplicated)
+- ✅ Is evergreen (no temporal content)
+- ✅ Is categorized and typed
+- ✅ Has calculated reading level
 
 ## License
 
@@ -251,11 +327,11 @@ Internal BYU-Idaho use. Contact: ai-team@byui.edu
 ## Citation
 
 ```bibtex
-@misc{byui-web-content-production-2025,
-  title={BYU-Idaho Web Content Dataset (Production)},
+@misc{byui-web-content-enhanced-2025,
+  title={BYU-Idaho Web Content Dataset (Enhanced)},
   author={BYU-Idaho AI Team},
   year={2025},
-  version={2.0.0},
+  version={3.0.0},
   publisher={Hugging Face},
   howpublished={\url{https://huggingface.co/datasets/BYU-Idaho/Web-Content}}
 }
@@ -268,4 +344,4 @@ Internal BYU-Idaho use. Contact: ai-team@byui.edu
 
 ---
 
-**Quality Guarantee:** This dataset has been extensively cleaned and validated for production use in RAG, search, and training applications. All temporal, duplicate, and low-quality content has been systematically removed.
+**Enterprise-Ready:** This dataset features advanced enrichment and filtering for production AI applications. With category classification, quality scoring, and reading levels, it's optimized for intelligent routing, quality filtering, and audience-appropriate responses.

data-00000-of-00001.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:606e36ec666e67993845c172b60fd957076d85e302afd6f101fcf87ad8c68640
-size 3564447
+oid sha256:c71864a83f8feee2fe049847b136dd8bdee96bb3fd475d48a07893bb2ed23b44
+size 3036330

dataset_info.json

@@ -1,7 +1,7 @@
 {
-  "dataset_name": "byui-web-content-production",
-  "version": "2.0.0",
-  "description": "Production-quality BYU-Idaho web content (cleaned and filtered)",
+  "dataset_name": "byui-web-content-enhanced",
+  "version": "3.0.0",
+  "description": "Ultra-high-quality BYU-Idaho web content with category classification and quality scores",
   "features": {
     "index": {
       "dtype": "int64",
@@ -26,13 +26,22 @@
     "content": {
       "dtype": "string",
       "_type": "Value"
-    }
-  },
-  "splits": {
-    "train": {
-      "name": "train",
-      "num_bytes": 7496056,
-      "num_examples": 2666
+    },
+    "category": {
+      "dtype": "string",
+      "_type": "Value"
+    },
+    "content_type": {
+      "dtype": "string",
+      "_type": "Value"
+    },
+    "quality_score": {
+      "dtype": "int64",
+      "_type": "Value"
+    },
+    "reading_level": {
+      "dtype": "float64",
+      "_type": "Value"
     }
   }
 }

state.json

@@ -4,6 +4,6 @@
       "filename": "data-00000-of-00001.parquet"
     }
   ],
-  "_fingerprint": "byui_web_content_production_v2",
+  "_fingerprint": "byui_web_content_enhanced_v3",
   "_format_type": "parquet"
 }