🧠 MLM-Based Symbolic Classification

Analyze text using masked language modeling to predict symbolic roles at specific positions.

Input Text

Masking Strategy

Top Predictions per Position

1 10

Symbolic Roles to Consider

<subject> <subject1> <subject2> <pose> <emotion> <surface> <lighting> <material> <accessory> <footwear> <upper_body_clothing> <hair_style> <hair_length> <headwear> <texture> <pattern> <grid> <zone> <offset> <object_left> <object_right> <relation> <intent> <style> <fabric> <jewelry>

Analysis Summary

Positions Analyzed

Best Prediction

Detailed Results

Input Text

Mask Positions (comma-separated)

Specify token positions to mask (0-based indexing)

Symbolic Roles

<subject> <subject1> <subject2> <pose> <emotion> <surface> <lighting> <material> <accessory> <footwear> <upper_body_clothing> <hair_style> <hair_length> <headwear> <texture> <pattern> <grid> <zone> <offset> <object_left> <object_right> <relation> <intent> <style> <fabric> <jewelry>

Manual Analysis Results

Summary

Positions

Text to Inspect

Token Breakdown

Shows how text is tokenized with position indices

🖼️ Test with Training-Style Patterns

The model was trained to predict descriptive words AFTER symbolic tokens.

Test with patterns like:

<subject> a young woman wearing elegant dress
<lighting> soft natural illumination on the scene
<emotion> happy expression while posing confidently

Training-Style Example

Training-Style Example

Training-Style Example

Training-Style Example

Training-Style Example

Training-Style Example

Training-Style Example

Training-Style Example

·

Built with Gradio logo