File size: 4,954 Bytes
0d254bc c97f929 0d254bc |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 |
---
license: mit
language:
- en
base_model:
- distilbert/distilbert-base-uncased
pipeline_tag: text-classification
---
# Topic Classifier
This repository contains the Topic Classifier model developed by DAXA.AI. The Topic Classifier is a machine learning model designed to categorize text documents across various domains, such as corporate documents, financial texts, harmful content, and medical documents.
## Model Details
### Model Description
The Topic Classifier is a BERT-based model, fine-tuned from the `distilbert-base-uncased` model. It is intended for categorizing text into specific topics, including "CORPORATE_DOCUMENTS," "FINANCIAL," "HARMFUL," and "MEDICAL." This model streamlines text classification tasks across multiple sectors, making it suitable for various business use cases.
- **Developed by:** DAXA.AI
- **Funded by:** Open Source
- **Model type:** Text classification
- **Language(s):** English
- **License:** MIT
- **Fine-tuned from:** `distilbert-base-uncased`
### Model Sources
- **Repository:** [https://huggingface.co/daxa-ai/topic-classifier](https://huggingface.co/daxa-ai/Topic-Classifier-2)
- **Demo:** [https://huggingface.co/spaces/daxa-ai/Topic-Classifier-2](https://huggingface.co/spaces/daxa-ai/Topic-Classifier-2)
## Usage
### How to Get Started with the Model
To use the Topic Classifier in your Python project, you can follow the steps below:
```python
# Import necessary libraries
from transformers import AutoTokenizer, AutoModelForSequenceClassification
import torch
import joblib
from huggingface_hub import hf_hub_url, cached_download
# Load the tokenizer and model
tokenizer = AutoTokenizer.from_pretrained("daxa-ai/topic-classifier")
model = AutoModelForSequenceClassification.from_pretrained("daxa-ai/topic-classifier")
# Example text
text = "Please enter your text here."
encoded_input = tokenizer(text, return_tensors='pt')
output = model(**encoded_input)
# Apply softmax to the logits
probabilities = torch.nn.functional.softmax(output.logits, dim=-1)
# Get the predicted label
predicted_label = torch.argmax(probabilities, dim=-1)
# URL of your Hugging Face model repository
REPO_NAME = "daxa-ai/topic-classifier"
# Path to the label encoder file in the repository
LABEL_ENCODER_FILE = "label_encoder.joblib"
# Construct the URL to the label encoder file
url = hf_hub_url(REPO_NAME, filename=LABEL_ENCODER_FILE)
# Download and cache the label encoder file
filename = cached_download(url)
# Load the label encoder
label_encoder = joblib.load(filename)
# Decode the predicted label
decoded_label = label_encoder.inverse_transform(predicted_label.numpy())
print(decoded_label)
```
## Training Details
### Training Data
The training dataset consists of 29,286 entries, categorized into four distinct labels. The distribution of these labels is presented below:
| Document Type | Instances |
| ------------------- | --------- |
| CORPORATE_DOCUMENTS | 17,649 |
| FINANCIAL | 3,385 |
| HARMFUL | 2,388 |
| MEDICAL | 5,864 |
### Evaluation
#### Testing Data & Metrics
The model was evaluated on a dataset consisting of 4,565 entries. The distribution of labels in the evaluation set is shown below:
| Document Type | Instances |
| ------------------- | --------- |
| CORPORATE_DOCUMENTS | 3,051 |
| FINANCIAL | 409 |
| HARMFUL | 246 |
| MEDICAL | 859 |
The evaluation metrics include precision, recall, and F1-score, calculated for each label:
| Document Type | Precision | Recall | F1-Score | Support |
| ------------------- | --------- | ------ | -------- | ------- |
| CORPORATE_DOCUMENTS | 1.00 | 1.00 | 1.00 | 3,051 |
| FINANCIAL | 0.95 | 0.96 | 0.96 | 409 |
| HARMFUL | 0.95 | 0.95 | 0.95 | 246 |
| MEDICAL | 0.99 | 1.00 | 0.99 | 859 |
| Accuracy | | | 0.99 | 4,565 |
| Macro Avg | 0.97 | 0.98 | 0.97 | 4,565 |
| Weighted Avg | 0.99 | 0.99 | 0.99 | 4,565 |
#### Test Data Evaluation Results
The model's evaluation results are as follows:
- **Evaluation Loss:** 0.0233
- **Accuracy:** 0.9908
- **Precision:** 0.9909
- **Recall:** 0.9908
- **F1-Score:** 0.9908
- **Evaluation Runtime:** 30.1149 seconds
- **Evaluation Samples Per Second:** 151.586
- **Evaluation Steps Per Second:** 2.391
## Conclusion
The Topic Classifier achieves high accuracy, precision, recall, and F1-score, making it a reliable model for categorizing text across the domains of corporate documents, financial content, harmful content, and medical texts. The model is optimized for immediate deployment and works efficiently in real-world applications.
For more information or to try the model yourself, check out the public space [here](https://huggingface.co/spaces/daxa-ai/Topic-Classifier-2). |