What’s This? Introduction to Named Entity Recognition

Last Updated August 18, 2023

named entity recognition

Data collection + annotation tasks like Named Entity Recognition results in smart linguistic datasets with a focus on quality control.

Your AI innovation is only as good as the data it learns from. And that data is only useful if it goes through a meaningful annotation process.

So, in this article, we’re going to get right into the thick of one specific task that really demonstrates  how our team is enriching AI.

What is Named Entity Recognition?

Named Entity Recognition (NER) annotation is a process in natural language processing (NLP) and machine learning (ML) where names of people, organizations, locations, dates, and other types of significant terms, are identified and labeled within a text or a body of texts.

The goal of NER annotation is to teach a machine learning model to recognize and classify these named entities correctly.

NER annotation involves manually labeling entities within a given text dataset, indicating the type of each entity.

For instance, consider the sentence: “Apple Inc. was founded by Steve Jobs on April 1, 1976, in Cupertino, California.”

NER annotation for this sentence would involve labeling “Apple Inc.” as an organization, “Steve Jobs” as a person, “April 1, 1976” as a date, and “Cupertino, California” as a location.

We’ll talk more later about how we do it, but first …

Why is Named Entity Recognition Important?

Here’s several reasons why you need to invest in NER.

Information Extraction

NER allows machines to extract structured information from unstructured text. By identifying named entities, NER enables systems to better understand and process the content of text data.

Search and Retrieval

NER enhances search engines by enabling more accurate indexing and retrieval of documents. Users can search for specific entities, such as finding all articles mentioning a particular person or organization.

Question Answering

NER aids in question answering systems, where the system needs to identify relevant entities to answer user questions accurately.

Text Summarization

NER helps in generating concise summaries of texts by identifying and retaining only the most important entities, improving the overall quality of summaries.

Relation Extraction

NER plays a role in relation extraction, where we identify the relationships between different entities. For example, NER can help identify that “Steve Jobs” founded “Apple Inc.”

Sentiment Analysis

Identifying named entities can also be important for sentiment analysis. The sentiment towards a particular entity can impact the overall sentiment of a text.

Data Enrichment

NER can be used to enrich datasets by extracting structured information from raw text data. This is particularly valuable in building knowledge graphs and databases.

Language Understanding

NER is a crucial step in enhancing a machine’s understanding of language, enabling it to comprehend the context and significance of different entities in a text.


NER can aid in machine translation by preserving the identities of important entities across languages.

A Glimpse into the Named Entity Recognition Process

Let’s say you have translated content and need to identify the named entities. You’ll be using these categories:

NER Categories

  1. Person (PER): Individuals, groups of people, nicknames, fictional characters, animal names, etc.
  2. Location (LOC): Physical locations such as countries, cities, lakes, buildings, planets, geographic coordinates, streets, etc.
  3. Organization (ORG): Governments (i.e., geopolitical organizations), companies, sports teams, religions, etc.
  4. Brand (BRAND): Organization, group, or producer of a specific commercial item or line of products.
  5. Commercial Item (COMM): iPhone, Corolla LX, Barbie, etc. (any non-generic purchasable product).
  6. Title (TITLE:OTHER): Name of any creation or creative work of art not captured by the categories Movie, Music, Book, Software, or Game.
  7. Movie (TITLE:MOVIE): Name of a movie, whether full name, nickname, or subtitle.
  8. Music (TITLE:MUSIC): Name of a song, whether full or partial, as well as a collection of individual music creations, such as an album or anthology.
  9. Book (TITLE:BOOK): Name of a book, whether professionally or self-published.
  10. Software (TITLE:SOFT): Name of an officially released software product.
  11. Game (TITLE:GAME): Name of a game, whether video game, board game, common game, or sport.
  12. Personal Title (PERSON:TITLE): Official titles and honorifics such as President, PhD, Dr., etc.
  13. Event (EVENT): Festival, concert, election, war, conference, etc.

NER Models

The existing named entity translations should belong to one of the four modes listed below:

  • Transliteration: meets all three conditions:
    • The source entity is in the script of the source language, and
    • Its translation is in the script of the target language, and
    • The aim of the translation is to preserve phonetics
  • Copy-through: The entity is directly copied from source to target
  • Translation: The entity is translated from the source language to the target language
  • Mixed: The entity translation is a mix of the above modes.

How It Comes Together

So, tag the source sentence (src_sentence_marked) and the target sentence (trg_sentence_marked):

a. with numbered named entity tags so that multiple entities can be identified.

<NE1>…</NE1> ..m <NE2> … </NE2> …

b. and named entity translation mode which can be one of the following:

  1. Transliteration: <transliteration> … </transliteration>
  2. Translation: <translation> … </translation>
  3. Copy-through: <copy-through> … </copy-through>
  4. Mixed: <mixed> … </mixed>

Here’s an example of transliteration + translation

  1. Source (EN): Today, the <NE1>Vatican Palaces</NE1> encompass a floor area of about 162,000 square meters (1,744,000 square feet).
  2. Target (JA): 現在、<NE1><mixed>バチカン宮殿</mixed></NE1>の床面積は約162,000平方メートル(1,744,000平方フィート)です。

Overall, NER annotation and its subsequent application in machine learning models contribute to improving various NLP tasks, making text processing more accurate, efficient, and semantically meaningful.

Get to Know our Data Annotation Services

As innovators in the data collection space, we offer flexible, customizable data services that evolve with your needs.

Render your data meaningful and train your algorithm free from biases with our labeling and classification services for text, speech, image, and video data.

So, contact us today to learn more.

Related Posts

Summa Linguae uses cookies to allow us to better understand how the site is used. By continuing to use this site, you consent to this policy.

Learn More