📰 TruStorE™: Sifting Fact From Fiction

NLP classifier for misinformation detection. Powered by Truth Over Tone Technology™, TruStorE™ blends linguistic heuristics, sentiment drift analysis, and Word Pair Logic™ to flag manipulative tone in news articles, a tell-tale sign of fake news as emotions are elicited to replace facts. Built for reproducibility, modular deployment, and impact.

---

---

🧪 Skills Demonstrated

• Modular NLP pipeline design with multilingual scaling
• Truth Over Tone Technology™ for tone and bias detection
• Word Pair Logic™ for linguistic signal extraction
• Strategic sampling to bypass NLTK constraints
• Visual clarity through histogram plots
• Artifact-first branding and footer evolution

---

🧰 Tech Stack

Layer	Tools Used
ETL & Wrangling	pandas, numpy, google.colab
Preprocessing	nltk, strategic sampling to bypass NLTK constraints
Modeling	sentiment_analysis, ngram_analysis, td-idf vectorization, Manipulative Tactic Detector™
Feature Extraction	Word Pair Logic™
Certification	TrueStorE™ Certification Engine
Visualization	matplotlib, seaborn, histogram plots

---

📦 Dependencies: See requirements.txt for install stack

---

📁 Repo Structure

TruStorE/
├── modules/
│   ├── etl.py                  # Data loading and cleaning
│   ├── preprocessing.py        # Labeling, tokenization, sampling
│   ├── tone_detector.py        # Manipulative Tactic Detector™
│   ├── sentiment_analysis.py   # Sentiment scoring and hypothesis testing
│   ├── word_pair_logic.py      # Linguistic signal extraction
│   ├── certification_engine.py # Final decision logic
│   └── viz.py                  # Histogram plotting utilities
├── notebook/
│   └── TruStorE_classifier.ipynb
├── README.md
├── LICENSE
├── requirements.txt
└── .gitignore
 

---

⚙️ How It Works

• Truth Over Tone Technology™
  Flags tonal asymmetry and sentiment drift using a proprietary blend of heuristics and NLP scoring logic.

• Word Pair Logic™
  Extracts linguistic signals from bigram patterns using SME-informed heuristics. Visualized via histogram plots to highlight frequency and semantic asymmetry.

• Sentiment Analysis
  Tests emotional polarity hypotheses. Visualized with histogram distributions and mean score overlays.

• TrueStorE™ Certification Engine
  Final decision logic that certifies article integrity based on cumulative linguistic, tonal, and semantic signals.
  Think of it as the final stamp—deployable, explainable, and recruiter-facing.

• Strategic Sampling
  Bypasses NLTK constraints by curating datasets that preserve semantic diversity and tone variance.

---

📥 Data Provenance

This classifier was trained and tested on a curated dataset of true and fake news articles.
The CSVs were going to be included in /data/ for reproducibility until I realized they exceeded GitHub's upload limit: See Dataset Downloads below

Source: Provided as part of a Ground.News simulation task.
No proprietary data used. All preprocessing and labeling logic is visible in the notebook.

---

📐 TF-IDF: Textbook vs Codebook Logic

📚 Textbook Definition
TF-IDF (Term Frequency–Inverse Document Frequency) evaluates word importance across documents.

1. Term Frequency (TF):
   TF(t) = (# of times term t appears in a document) / (total terms in the document)

2. Inverse Document Frequency (IDF):
   IDF(t) = log(total documents / documents containing term t)

3. TF-IDF Score:
   TF-IDF(t, d) = TF(t, d) × IDF(t)

💻 Codebook Logic in TruStorE™
In this classifier, TF-IDF is signal extraction—layered with heuristics and tone detection.

1. Strategic Sampling:
   Curated multilingual payloads preserve tone variance and semantic diversity.

2. Weighted TF-IDF:
   Terms weighted by emotional polarity and drift, not just frequency.

3. Word Pair Logic™ Overlay:
   Bigram patterns extracted post-vectorization to flag manipulative phrasing.

4. Sentiment Drift Index™ Integration:
   TF-IDF vectors cross-referenced with tonal asymmetry.

5. Certification Engine Input:
   Final TF-IDF vectors feed into TruStorE™ Certification Engine for integrity scoring.

From textbook to codebook, TF-IDF becomes a linguistic scalpel—powered by Truth Over Tone Technology™.

---

📦 Dataset Downloads

Due to GitHub’s upload limit and Google Drive’s preview threshold, both datasets are hosted externally. Download directly below:

• ✅ True_stories.csv
• ❌ Fake_stories.csv

---

✅ Deployment Strategy

• Notebook-first walkthrough is ready for Jupyter download
• Modular logic blocks are structured for RESTful API deployment
• Multilingual pipeline supports expansion to non-English datasets
• Dashboard wiring with recruiter-facing metrics and footer lore is next on deck

---

👣 Next Steps for Real-World Integration

• 🧠 Save and version the trained model
• 🔌 Create an API endpoint for real-time classification
• 🌐 Integrate with Ground News website backend
• 📱 Extend functionality into the mobile app interface
• 🌍 Add multilingual translation tech for global scalability
• 📊 Leverage balanced F1 harmonics (0.96 sweep) for EVALS benchmarking

This isn’t just a model—it’s the TruStorE™ Certification Engine, powered by Word Pair Logic™, Sentiment Drift Index™, and Truth Over Tone Technology™.

---

🧾 License

MIT © 2025 Garrick Piñón. TruStorE™, Word Pair Logic™, Sentiment Drift Index™, and Truth Over Tone Technology™ are trademarked methodologies.
Commercial use requires attribution and explicit permission.

---

🏷️ Loyalty Badges – Faithful to the Footer

If you’re reading this, you’re not just curious—you’re committed. These badges are for you. Let's be friends.

---