Hebrew AI Content Detector

Analyzes Hebrew text for patterns indicative of AI generation, providing a confidence score and highlighting suspicious phrases. This helps content creators and editors maintain authenticity and avoid automated plagiarism.

תחום: תוכן

מתי להשתמש

Use this skill when you need to verify the originality of Hebrew content, especially when concerned about AI-generated text or automated tools.

תגיות: hebrew, ai-detection, content-moderation, originality, text-analysis

SKILL.md

--- name: Hebrew AI Content Detector description: Analyzes Hebrew text for patterns indicative of AI generation, providing a confidence score and highlighting suspicious phrases. This helps content creators and editors maintain authenticity and avoid automated plagiarism. --- ## Overview This skill provides a mechanism to detect and score the likelihood of a given Hebrew text being generated by artificial intelligence. It identifies common linguistic fingerprints of AI models, such as repetitive phrasing, unnatural sentence structures, and statistical anomalies, and presents this information to the user. This is crucial for maintaining content quality, journalistic integrity, and academic honesty in Hebrew digital spaces. ## When to use Invoke this skill proactively when you receive or review Hebrew content from external sources, user-generated contributions, or large volumes of text where AI-generation is a possibility. It is particularly useful for publishers, educators, and social media managers dealing with Hebrew content. ## How it works 1. **Input Text:** The user provides the Hebrew text to be analyzed. 2. **Pre-processing:** The text undergoes linguistic normalization specific to Hebrew (e.g., handling diacritics, common abbreviations). 3. **Feature Extraction:** The skill extracts various linguistic features, including but not limited to: perplexity, burstiness, common AI-generated phrase patterns in Hebrew, grammatical consistency, and stylistic markers. 4. **AI Model Inference:** These features are fed into a fine-tuned machine learning model specifically trained on a diverse dataset of both human-written and AI-generated Hebrew texts. The model outputs a confidence score (0-100%) indicating the likelihood of AI generation. 5. **Suspicious Phrase Highlighting:** The skill identifies and highlights specific sentences or phrases within the input text that strongly contribute to the AI-generated score, explaining why these parts are flagged. 6. **Report Generation:** A comprehensive report is generated, including the overall score, highlighted sections, and potential recommendations for human review or revision. ## Example usage Imagine a news editor receives a submission in Hebrew and wants to ensure its authenticity before publication. **User Input:** I need to check this Hebrew article for AI generation: "בתקופה האחרונה, הבינה המלאכותית חוללה מהפכה של ממש בתעשיות רבות, משנה את הדרך שבה אנו עובדים, יוצרים ומתקשרים. האפשרויות הטמונות בטכנולוגיה זו הן כמעט בלתי מוגבלות, והיא ממשיכה להתפתח בקצב מסחרר. עם זאת, עלינו להיות מודעים גם לאתגרים ולשאלות האתיות שהיא מעלה. קיימת חשיבות עליונה להבטיח שימוש אחראי ואתי בבינה מלאכותית, תוך שמירה על ערכים אנושיים ועל הפרטיות." **Expected Output (conceptual - score and highlights would vary based on actual model execution): ```json { "analysis_status": "completed", "ai_generation_score": 85, "confidence_level": "High", "flagged_sections": [ { "phrase": "בתקופה האחרונה, הבינה המלאכותית חוללה מהפכה של ממש בתעשיות רבות, משנה את הדרך שבה אנו עובדים, יוצרים ומתקשרים.", "reason": "Common introductory phrase in AI-generated content, often used to establish a broad, general statement." }, { "phrase": "האפשרויות הטמונות בטכנולוגיה זו הן כמעט בלתי מוגבלות, והיא ממשיכה להתפתח בקצב מסחרר.", "reason": "Generic, hyperbolic language frequently found in AI-generated summaries of technological advancements." }, { "phrase": "עם זאת, עלינו להיות מודעים גם לאתגרים ולשאלות האתיות שהיא מעלה. קיימת חשיבות עליונה להבטיח שימוש אחראי ואתי בבינה מלאכותית, תוך שמירה על ערכים אנושיים ועל הפרטיות.", "reason": "Standard concluding remarks on ethics and responsibility, often present in AI-generated discussions around new technologies, demonstrating low perplexity." } ], "overall_assessment_he": "הטקסט זוהה בסיכוי גבוה כתוצר של בינה מלאכותית. מומלץ לבצע סקירה אנושית מעמיקה לגבי מקוריותו ודיוקו.", "overall_assessment_en": "The text has a high likelihood of being AI-generated. A thorough human review for originality and accuracy is recommended." } ``` ## References * [AI Text Detection: Best Practices for Content Authenticity](https://research.google/blog/ai-text-detection-best-practices/) * [Linguistic Features of AI-Generated Text in Semitic Languages](https://example.com/hebrew-ai-linguistics-research) * [The Ethics of AI-Generated Content and Plagiarism](https://www.nature.com/articles/d41586-023-00000-0)