|
--- |
|
datasets: |
|
- financial_phrasebank |
|
- clinc_oos |
|
- hate_speech_offensive |
|
tags: |
|
- finance |
|
language: |
|
- en |
|
--- |
|
# BERT Base Intent model |
|
This is a fine tuned model based on Bert-Base-Uncased model. This model is used to classify intent into 3 categories- Fintech, Out of Scope and Abusive. |
|
The base line model is a pretrained model on English language using a masked language modeling (MLM) objective. It was introduced in |
|
[this paper](https://arxiv.org/abs/1810.04805) and first released in |
|
[this repository](https://github.com/google-research/bert). |
|
## Training procedure |
|
|
|
### Training hyperparameters |
|
|
|
The following hyperparameters were used during training: |
|
- learning_rate: 2e-5 |
|
- num_epochs: 3 |
|
- weight_decay:0.01 |
|
|
|
### Training results |
|
|
|
| Training Loss | Epoch | Validation Loss | Accuracy | F1 | |
|
|:-------------:|:-----:|:----------------:|:---------------:|:--------:| |
|
| 0.114200 | 1.0 | 0.034498 | 0.991351 | 0.991346 | |
|
| 0.024100 | 2.0 | 0.037945 | 0.992349 | 0.992355 | |
|
| 0.009800 | 3.0 | 0.034846 | 0.993347 | 0.993345 | |
|
|
|
|
|
### Model Description |
|
|
|
The finetuned Hugging Face model is a variant of the BERT-base-uncased architecture, trained for intent classification |
|
with three labels: fintech, abusive, and out of scope. The model has undergone a fine-tuning process, where it has been |
|
trained on a large corpus of annotated data using a supervised learning approach. The objective of the model is to |
|
classify incoming text data into one of the three predefined classes based on the underlying intent of the text. |
|
|
|
The performance of the model was evaluated and it achieved high accuracy and F1 scores |
|
for all three classes. The model's high accuracy and robustness make it suitable for use in real-world applications, |
|
such as chatbots, customer service automation, and social media monitoring. |
|
Overall, the finetuned Hugging Face model provides an effective and reliable solution for intent classification |
|
with three labels: fintech, abusive, and out of scope. |
|
|
|
|
|
- **Developed by:** Jeswin MS, Venkatesh R, Kushal S Ballari |
|
- **Model type:** Intent Classification |
|
- **Language(s) (NLP):** English |
|
- **Finetuned from model:** Bert-base-uncased |
|
|
|
|