This section explains how Sentix figures out the Sentiment indicator and Sentiment Media Index, making it easy to understand the sense behind the calculations.
News sentiment is about the feelings, views, and attitudes shared in news stories or media. To interpret the sentiment in numbers, Sentix uses tools like artificial intelligence (AI), natural language processing (NLP), machine learning (ML), and text analysis.
To present the sentiment scoring process used by Sentix in a professional and structured format, including definitions for each part of the formula, we can outline it as follows:
Sentiment Score Calculation:
*Definitions:
- Total Negative Statements: The number of statements identified as negative by the algorithm within a specific news event.
- Total Positive Statements: The number of statements identified as positive by the algorithm within the same event.
- Total Statements: The sum of all statements (positive, negative, and neutral) analyzed in the event.
- Sentiment Score: A numerical representation of the overall sentiment, ranging from -10 (completely negative) to +10 (completely positive), with 0 indicating a neutral sentiment.
Calculation Process Decomposed:
- Raw Input Counting: Begin by tallying the total number of negative, positive, and overall statements within the event.
- Sentence-Level Prediction: Each sentence is analyzed for sentiment (positive, negative, or neutral) and assigned a confidence value (e.g., 99% confidence in a positive sentiment).
- Contextual Understanding and Model Confidence: The current Sentix model, trained extensively on crypto news, evaluates the context of each sentence, its relevance, and the confidence level of its prediction, highlighting key phrases and sentences influencing its decision.
- Summary Generation: Based on the predictions and confidence values, Sentix creates a summary for users, pointing out the significant phrases and sentences driving the sentiment score.
This structured approach not only quantifies sentiment in a clear, numerical format but also ensures that users have a comprehensive understanding of how each score is derived based on rigorous analysis and machine learning models specialized in the crypto news domain.