Task: Text sentiment analysis

[fabiocat93]

Description

Extracting sentiment (polarity) and emotional category from text

Tasks

• [ ] Sentiment analysis
• [ ] Emotional analysis

Freeform Notes

No response

[adi611]

Hi @fabiocat93 - I noticed the task is unassigned and would like to take a shot at it. Let me know if that works!

[fabiocat93]

Hi @adi611 - Thanks for offering to take this on! I've assigned the task to you. Below are the details and the checklist to ensure everything meets the project requirements.

Description

Your task involves implementing two main functionalities:

1. Sentiment Analysis: Determine the sentiment polarity of each text in the list, providing both a numerical score (ranging from -1 to 1) and a corresponding label ("negative", "neutral", or "positive").

2. Emotional Analysis: Identify the emotional content of each text, returning a dictionary of scores (probability for each possible emotion) and the dominant emotion as a string.

Tasks

• [ ] Implement Sentiment Analysis:

  • Accepts pieces_of_text: List[str], device: DeviceType, model: SenselabModel, and **args as inputs.
  • Returns a typed Dict[str, Any] with the keys:
    • score: A float between -1 and 1 indicating the sentiment polarity.
    • label: A string ("negative", "neutral", "positive") representing the sentiment category.
  • Edge Cases:
    • Handle empty strings or strings without any meaningful content gracefully.
    • Ensure proper handling of multilingual text or non-standard characters.
  • Unit Tests:
    • Ensure high coverage with meaningful tests, including edge cases like empty strings, very long text, or ambiguous sentiments.
  • Documentation:
    • Include Google-style docstrings for all functions and methods.
  • Pre-commit Hooks:
    • Ensure all code passes pre-commit checks (e.g., linting, formatting).

• [ ] Implement Emotional Analysis:

  • Accepts pieces_of_text: List[str], device: DeviceType, model: SenselabModel, and **args as inputs.
  • Returns a typed Dict[str, Any] with the keys:
    • scores: A dictionary with emotions as keys and their respective probabilities as values.
    • dominant_emotion: A string representing the most likely emotion in the text.
  • Edge Cases:
    • Consider texts with mixed emotions and ensure the dominant emotion is consistently identified.
    • Handle cases where the model might return similar probabilities for multiple emotions.
  • Unit Tests:
    • Cover diverse scenarios, including texts with clear single emotions and those with complex emotional tones.
  • Documentation:
    • Include Google-style docstrings for all functions and methods.
  • Pre-commit Hooks:
    • Ensure all code passes pre-commit checks (e.g., linting, formatting).

Acceptability Criteria

• Typed Python Code: Ensure all code is fully typed, including inputs and return types.
• Pre-commit Hooks: All code must pass pre-commit hooks without errors.
• Good Unit Test Coverage: Aim for at least 90% coverage, ensuring all edge cases are tested.
• Google Style Docstrings: All functions and methods should have clear and concise docstrings in Google style.
• Expected Outputs:
  • For Sentiment Analysis:
  • score: Float between -1 and 1.
  • label: String ("negative", "neutral", "positive").
  • For Emotional Analysis:
  • scores: Dictionary with probabilities for each emotion.
  • dominant_emotion: String of the most probable emotion.
• Edge Cases: Ensure proper handling of special cases, such as empty strings, very short or long texts, mixed-language input, etc.

I am happy to discuss this further if you feel it's needed. Thanks again for taking this on, and looking forward to your implementation!

[adi611]

Hey @fabiocat93, I have made some changes and added tests, but the commit needs me to run pytest on all the test cases, and it's taking too long at the test_synthesize_texts_with_mars5_model. I checked the hardware requirements for the model and it’s way too much for my machine to handle. Any workaround?

[adi611]

Hey @fabiocat93, I have made some changes and added tests, but the commit needs me to run pytest on all the test cases, and it's taking too long at the test_synthesize_texts_with_mars5_model. I checked the hardware requirements for the model and it’s way too much for my machine to handle. Any workaround?

using SKIP=pytest as a solution for now.