Ramgopal Prajapat:

Learnings and Views

Real world Applications of Named entity recognition (NER)

By: Ram on Dec 17, 2022

Summary

Named entity recognition (NER) is a subfield of natural language processing (NLP).  In this blog, we will discuss the following

  • Named entity recognition -NER Overview
  • Applications and Use-cases of NER or NER Projects

In the next blog, we will show you end to end steps complete NER Project using Python.

 

Named Entity Recognition (NER): Overview

 Named entity recognition (NER) is a subfield of natural language processing (NLP).  And it involves in identifying “Entities” from a text corpus/body or document. Once the “Entities” are identified, they are classified into a few categories such as Person, Place, Time Period etc.

Ms. Gopinath previously served as the Chief Economist of the Fund from 2019-22.”

In this sentence, 3 entities are identified and then “Ms. Gopinath” will be classified as “Person”, “Chief Economist” as “Designation” and “2019-22” as Time Period.

So, the Named entities can be specific words or phrases that refer to a person, place, organization, or other real-world entity.

There are several approaches available for ER, including rule-based, dictionary-based, and machine learning-based methods. Machine learning-based methods are commonly used. There are pre-trained models available in the Python Packages and can be done with a few lines of code.

 

Applications and Use-cases of NER

NER systems are used in a wide variety of applications, including information extraction, question answering, and text summarization. They are also used across industries such as finance, ecommerce, healthcare, and media. We will give a real word use-cases and applications of NER for some of these industries.

Some of the real-world applications of NER are detailed out below and definitely not an exhaustive list.

 

  1. Financial Services or Finance

Credit Risk Modelling and NER

For estimating risk involving granting credit line to corporate client is one of the key ML applications in Financial Services. The modelling is called Probability of Default Model or Application Scorecard.  Data augmentation is key to improve the accuracy of these models. The financial clients or bank use external data – new or social media data about the existing and new clients. These data sources are not structured data. Named entity recognition (NER) models helps in linking the news and posts to the corporate clients of the banks and financial services. In addition to NER, other of natural language processing (NLP) methods or techniques are used to extract additional information for example sentiment scores from the reviews etc.

 

  1. Media and New

News Article Classification and Recommendations

Reader engagement is critical for News Portal or Media companies. They use Named Entity Recognition to scan news articles and extract major people, organizations, and places referred in the articles.  The identified NER then will be used for organizing the articles and referring/recommending related news articles.  If reading articles about Mr Narendra Modi – PM India, it will show the articles which referred Mr Modi inside.

 

Sales Lead Generation for TV Channels

Some of the start-ups release TV commercials either when they get funding (probably Series B or later) or they have decided to go pan India.  Some of these inputs are available in the news articles. The news articles can be extracted from multiple sources and then NER Model can help in identifying the companies that are looking to create TV commercials. This is a great sales lead pipeline for the Media company’s sales team to engage with the entities or organizations identified by NER.

  1. Ecommerce

Search is critical part of the customer journey on an ecommerce platform.  The ecommerce customers use search functionality to reach to the products of their choice quickly. When customers are search, it is important to extract intent and information correctly to show the products they are looking for. For example, when customer is using “adidas running shoes under 2000”, the NER can help identify that “Adidas” is a brand and must be used for filtering the products only for Brand – Adidas.

 

Graphical user interface, website

Description automatically generated

 

Concluding Thoughts

NER Methods use text as input and identify entities such as Person, Place, Organization or Time Period. The methods can be custom trained to identify domain specific entities such designations of people or Brand names.  NER has wide range of applications and all of us can further augment the list.  You can also think on how NER can be useful in your context or industry. Please share if you come across any interesting NER use-case or want to discuss anything with us. Please reach out.

In the next blog, we will use Python for performing NER.

1 comments

Biswajit Pal Dec-2022

Very well-structured blog. It answers the question of "Where NER can be used" very proficiently.

Leave a comment