|
--- |
|
license: mit |
|
language: |
|
- en |
|
pipeline_tag: text-classification |
|
--- |
|
|
|
# EmoBERTv2 Model |
|
This Model Card is a work in progress and will be completed in the future (dataset upload pending, etc) |
|
## Model Description |
|
EmoBERTv2 is a emotion text classification model trained on a large dataset of english social media posts. The model is fine-tuned |
|
from "prajjwal1-bert-tiny" EmoBERTv2 can be used for either further fine-tuning, or for usage in real-time emotion prediction applications |
|
|
|
|
|
## Datasets |
|
This model was trained on the [Dataset Name] dataset, which is an aggregation of many datasets through relabling and data subsetting. The |
|
dataset has 9 labels: joy, sad, love, anger, disgust, surprise, neutral, fear, and worry |
|
|
|
## Training Procedure |
|
EmoBERTv2 was fine-tuned from [Base Model Name] with specific hyperparameters [List Hyperparameters]. Training involved [X] epochs, using a learning rate of [Y]. |
|
|
|
## Intended Use |
|
This model is intended for emotion classification in [specific domains or general use]. It should be used as a tool for [Specify Applications]. |
|
|
|
## Performance |
|
EmoBERTv2 demonstrates an accuracy of 86.17% on the [Test Dataset Name]Test set. For detailed performance metrics, refer to [Link to Performance Metrics]. |
|
|
|
## Bias and Fairness |
|
While efforts have been made to reduce bias, users should be aware of potential biases in the data. It is advisable to test the model in specific contexts. |
|
|
|
## Licensing and Usage |
|
EmoBERTv2 is released under the MIT License and can be freely used as outlined in the license. |
|
|
|
## Other Model Variations |
|
Additional variations of EmoBERTv2 include [List Variations]. These variations offer different trade-offs in terms of size, speed, and performance. |
|
|
