Bengali Name Extractor

keywords: Deep Learning, NLP, Token Classification, Name Entity Recognition

Brief Description

A deep learning based project to build a system which can extract person name from given text. The project was focused to build a robust model which can extract any person name which can be used in any call center and online voice based transaction systems.

As this is a name entity extraction task, it was handled as token classification task. In token classification process we can predict which token belongs to name entity class and extract names from this prediction. For the task, first I preprocessed the given data, making appropiate datasets for token classification modeling, then experimented with different huggingface models for the task. Then I train the models with these experimented parameters and build an end-to-end inference script.The inference script will load the best saved models (saved in training processe) and do prediction using the model and then post-process the model output for desire output format for given input.

Example -
input: আব্দুর রহিম নামের কাস্টমারকে একশ টাকা বাকি দিলাম
output: আব্দুর রহিম
input: অর্থনীতি ও আর্থসামাজিক বেশির ভাগ সূচকে বাংলাদেশ ছাড়িয়ে গেছে দক্ষিণ এশিয়াকে ।
output: None

For more details please visit the project repository

• Project Link : Bengali Name Extractor