Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/gaizkiaadeline/wine-variety-classifier-using-distilbert
Classifying Wine Varieties Based on The Provided Wine Expert Descriptions in The Wine Reviews Dataset using DistilBERT as the Language Model
https://github.com/gaizkiaadeline/wine-variety-classifier-using-distilbert
distilbert fine-tuned text-classification text-mining
Last synced: 1 day ago
JSON representation
Classifying Wine Varieties Based on The Provided Wine Expert Descriptions in The Wine Reviews Dataset using DistilBERT as the Language Model
- Host: GitHub
- URL: https://github.com/gaizkiaadeline/wine-variety-classifier-using-distilbert
- Owner: gaizkiaadeline
- Created: 2024-09-30T07:24:09.000Z (about 1 month ago)
- Default Branch: main
- Last Pushed: 2024-10-22T03:03:46.000Z (17 days ago)
- Last Synced: 2024-10-23T04:36:26.061Z (16 days ago)
- Topics: distilbert, fine-tuned, text-classification, text-mining
- Language: Jupyter Notebook
- Homepage:
- Size: 445 KB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Classifying Wine Varieties Based on The Provided Wine Expert Descriptions in The Wine Reviews Dataset using DistilBERT as the Language Model
# Project Description
This project implements a wine variety classification system based on textual descriptions provided by wine experts. Using DistilBERT, a lighter and faster version of BERT, the model aims to classify wine varieties based on detailed descriptions of their taste, aroma, and other characteristics.
# Problem Statement:
With the vast diversity in wine types, distinguishing between varieties based on expert descriptions can be challenging for consumers. To aid in wine selection, this project seeks to build a model that can automatically classify wine varieties based on the text reviews associated with each wine.
# Methodology:
**Dataset:**
- The dataset used in this project is sourced from Kaggle’s Wine Reviews dataset, which contains detailed reviews of different wines.
- After performing Exploratory Data Analysis (EDA), the focus was narrowed down to the four most frequent wine varieties in the dataset.
**Preprocessing:**
- Only the relevant columns, "description" (containing the text) and "variety" (representing the wine type), were retained.
- Tokenization was done using the tokenizer from the DistilBERT model to convert text data into input IDs and attention masks.
**Modeling:**
- DistilBERT (uncased) was used as the base model for text classification, with the classification head fine-tuned to output predictions for wine varieties.
- The model was fine-tuned and evaluated using hyperparameters such as learning rate, batch size, dropout rate, and the number of epochs.
**Parameter Tuning:**
Hyperparameters were tuned to find the optimal values for model performance, with manual adjustments made to the learning rate, epoch, dropout, and other model parameters.
**Results:**
- The model achieved an accuracy of 91%, indicating strong performance in classifying wines based on the expert descriptions.
- Performance across classes was consistent, with the Chardonnay class showing the best results in terms of precision, recall, and F1 score.
# Conclusion:
The use of DistilBERT proved effective for the task of classifying wine varieties based on textual descriptions, offering a smaller, faster model with a performance level close to the original BERT model.
Link Dataset : https://www.kaggle.com/datasets/zynicide/wine-reviews
