About This Hackathon

<h3>Welcome to Week 19 of the Weekly MachineHack Hackathon Series!</h3><p>This week’s challenge focuses on <strong>extracting key entities from product descriptions written in the Persian language</strong>. The task is to develop a solution that identifies the main product being discussed in the title and description, drawn from data provided by a mart. Participants must submit the extracted product corresponding to each entry in the test file.</p><h3><strong>Participation and Benefits</strong></h3><ul><li><strong>Skill Level</strong>:<br>Ideal for participants with experience in natural language processing (NLP), entity recognition, and working with multilingual text data.</li><li><strong>Community Engagement</strong>:<br>Join our Telegram group to collaborate with peers, ask questions, and share insights.</li><li><strong>Recognition</strong>:<br>All participants will receive a MachineHack certificate, and top performers will be highlighted on the leaderboard.</li></ul><h3><strong>Submission and Evaluation</strong></h3><ul><li><strong>Submission Format</strong>:<br>Submit the extracted product names in the provided Submission.csv file, ensuring that they align with the respective title/description in the test set.</li><li><strong>Evaluation Metric</strong>:<br>Submissions will be evaluated based on <strong>accuracy</strong> in correctly identifying the primary product.</li><li><strong>Leaderboard</strong>:<br>Track your ranking in real-time and compete for the top spot!</li></ul><h3><strong>How to Approach the Challenge</strong></h3><h4><strong>Data Preprocessing</strong></h4><ul><li><strong>Handle Text in Persian</strong>:<br>Ensure the dataset encoding supports Persian (UTF-8).</li><li><strong>Tokenization</strong>:<br>Use language-specific tokenizers such as Hazm or Parsivar to process Persian text effectively.</li><li><strong>Cleaning</strong>:<br>Remove irrelevant words, stopwords, and symbols that may not contribute to the product identification.</li></ul><h4><strong>Feature Engineering</strong></h4><ul><li><strong>Title and Description Alignment</strong>:<br>Focus on the alignment between the title and description for common product references.</li><li><strong>Keywords Extraction</strong>:<br>Apply techniques like TF-IDF or attention mechanisms to highlight key phrases.</li></ul><h4><strong>Modeling Techniques</strong></h4><ul><li><strong>Entity Recognition</strong>:<br>Utilize Named Entity Recognition (NER) models fine-tuned for Persian, such as ParsBERT.</li><li><strong>Transformers</strong>:<br>Experiment with transformer-based architectures like BERT or RoBERTa trained on Farsi data for accurate entity extraction.</li></ul><h4><strong>Validation and Tuning</strong></h4><ul><li><strong>Manual Validation</strong>:<br>Spot-check the extracted entities to ensure meaningful results.</li><li><strong>Hyperparameter Tuning</strong>:<br>Optimize model parameters using grid search or Bayesian optimization.</li></ul><h3><strong>Resources and Support</strong></h3><ul><li><strong>Starter Notebook</strong>:<br>A starter notebook will be available to help you begin with data exploration and model prototyping. Accessible for premium users.</li><li><strong>Expert Guidance</strong>:<br>A live walkthrough session will be held on <strong>9th December 2024</strong> at <strong>4:00 PM IST </strong>to provide tips and strategies for the challenge.</li></ul><h3><strong>Getting Started</strong></h3><ol><li><strong>Register Now</strong>:<br>Ensure you’re registered to receive updates and access challenge materials.</li><li><strong>Download the Dataset</strong>:<br>Access the dataset from the MachineHack platform and begin your analysis.</li><li><strong>Join the Community</strong>:<br>Collaborate and exchange ideas with other participants via our <a target="_blank" rel="noopener noreferrer" href="https://t.me/joinchat/NJLxnlWiz9lFnEJU20Sccw">Telegram group</a>.</li></ol><h3><strong>Support and Queries</strong></h3><p>For any questions, reach out to our support team at support@machinehack.com.</p><p>We’re excited to see your innovative solutions for this Persian-language NLP challenge! Best of luck!</p>

Welcome to Week 19 of the Weekly MachineHack Hackathon Series!

This week’s challenge focuses on extracting key entities from product descriptions written in the Persian language. The task is to develop a solution that identifies the main product being discussed in the title and description, drawn from data provided by a mart. Participants must submit the extracted product corresponding to each entry in the test file.

Participation and Benefits

Skill Level:
Ideal for participants with experience in natural language processing (NLP), entity recognition, and working with multilingual text data.
Community Engagement:
Join our Telegram group to collaborate with peers, ask questions, and share insights.
Recognition:
All participants will receive a MachineHack certificate, and top performers will be highlighted on the leaderboard.

Submission and Evaluation

Submission Format:
Submit the extracted product names in the provided Submission.csv file, ensuring that they align with the respective title/description in the test set.
Evaluation Metric:
Submissions will be evaluated based on accuracy in correctly identifying the primary product.
Leaderboard:
Track your ranking in real-time and compete for the top spot!

How to Approach the Challenge

Data Preprocessing

Handle Text in Persian:
Ensure the dataset encoding supports Persian (UTF-8).
Tokenization:
Use language-specific tokenizers such as Hazm or Parsivar to process Persian text effectively.
Cleaning:
Remove irrelevant words, stopwords, and symbols that may not contribute to the product identification.

Feature Engineering

Title and Description Alignment:
Focus on the alignment between the title and description for common product references.
Keywords Extraction:
Apply techniques like TF-IDF or attention mechanisms to highlight key phrases.

Modeling Techniques

Entity Recognition:
Utilize Named Entity Recognition (NER) models fine-tuned for Persian, such as ParsBERT.
Transformers:
Experiment with transformer-based architectures like BERT or RoBERTa trained on Farsi data for accurate entity extraction.

Validation and Tuning

Manual Validation:
Spot-check the extracted entities to ensure meaningful results.
Hyperparameter Tuning:
Optimize model parameters using grid search or Bayesian optimization.

Resources and Support

Starter Notebook:
A starter notebook will be available to help you begin with data exploration and model prototyping. Accessible for premium users.
Expert Guidance:
A live walkthrough session will be held on 9th December 2024 at 4:00 PM IST to provide tips and strategies for the challenge.

Getting Started

Register Now:
Ensure you’re registered to receive updates and access challenge materials.
Download the Dataset:
Access the dataset from the MachineHack platform and begin your analysis.
Join the Community:
Collaborate and exchange ideas with other participants via our Telegram group.

Support and Queries

For any questions, reach out to our support team at support@machinehack.com.

We’re excited to see your innovative solutions for this Persian-language NLP challenge! Best of luck!

EntityRecognitionChallenge