Develop an API that analyzes sentiment of user's text input

[harshita-srivastava-yral]

Objective:
Develop an API that analyzes user text input, identifies sentiment, suggests tone-down versions of aggressive or overly strong input, and filters out spammy, promotional, or fraudulent content. This API should be reusable for various use cases such as comments, token descriptions, or any other scenario where users can input text.

---

Core Requirements:

1. Sentiment Analysis:
   • Classify user input into sentiment categories (e.g., Positive, Neutral, Negative, Aggressive) DS can use their own ways of classification.
   • Provide a confidence score for the classification.
2. User Suggestions:
   • Generate alternative text suggestions for inputs flagged as aggressive or overly negative.
   • Ensure suggestions retain the original intent while softening tone.
   • Incorporate user feedback to refine suggestions.
3. Spam and Fraud Detection:
   • Identify and flag spammy, promotional, or fraudulent content.
   • Filter and remove inappropriate content programmatically.
4. API Features:
   • Accept text input of up to 30 words.
   • Provide responses for:
     • Sentiment analysis results.
     • Suggestions for alternative text.
     • Flags for spam or fraudulent content.
   • Ensure flexibility for integration into multiple use cases.

---

System Expectations:

• High accuracy for sentiment analysis and spam detection.
• Minimal latency to ensure real-time usability.
• Adherence to ethical AI guidelines and data privacy standards.

---

Deliverable:
A robust, reusable API for sentiment analysis, text moderation, and spam detection, capable of supporting diverse text-based functionalities across the platform.

Notion link: https://www.notion.so/gobazzinga/Sentiment-Analysis-for-user-s-text-input-17bdfecd572680beb023d83d5ee77a21

[harshita-srivastava-yral]

• This is to be picked post the Yral video moderation on NSFW

[siyara-m-yral]

• @harshita-srivastava-yral to update issue and assign to DS team
• @sagar-sarkale-yral and @jay-dhanwant-yral to work on implementation details

[siyara-m-yral]

• Researching over the new models
• Would likely complete by Friday

[harshita-srivastava-yral]

• Done with the exploration and result
• Demo can be done today

[harshita-srivastava-yral]

• We need to setup meeting with for the demo