Gen-Z Slang Sentiment RoBERTa
This model is a fine-tuned version of cardiffnlp/twitter-roberta-base-sentiment designed to solve the "Literal Trap" in sentiment analysis.
While standard models often flag Gen-Z hyperbolic slang (e.g., "I'm dead", "this is insane") as negative due to their literal dictionary definitions, this model recognizes them as high-intensity positive expressions when used in a conversational context.
Model Description
- Developed by: Adil
- Model type: RoBERTa-base
- Language(s): English
- Finetuned from model: cardiffnlp/twitter-roberta-base-sentiment
- Domain: Social Media DMs, Informal Chat, Gen-Z Slang
Intended Uses & Limitations
Primary Use Case
Ideal for lead qualification and sentiment tracking in informal DM environments (Instagram, TikTok, WhatsApp) where users speak naturally and use high-energy slang.
Out-of-Scope
This model is not intended for formal document analysis, legal texts, or clinical psychology sentiment where words like "insane" or "dead" retain their literal, serious meanings.
Training Data
The model was fine-tuned on a Contrastive Dataset of 5,000+ examples. The data features "Semantic Pairs" to teach the model nuance:
- Slang Positive: "omg im dead 💀" -> Positive
- Literal Negative: "the patient is dead" -> Negative
- Slang Positive: "that's sick!!" -> Positive
- Literal Negative: "I feel very sick" -> Negative
How to Get Started
You can use this model directly with the Hugging Face pipeline:
from transformers import pipeline
classifier = pipeline("sentiment-analysis", model="ghcvbn/sentiment-with-genz-slang-words-false-negative")
# Test a slang phrase
result = classifier("this is insane, im terrified and im obsessed")
print(result)
# Expected: [{'label': 'positive', 'score': 0.98}]