Browse Definitions :
natural language generation (NLG) sentiment analysis

named entity recognition (NER)

What is named entity recognition (NER)?

Named entity recognition (NER) is a natural language processing (NLP) method that extracts information from text. NER involves detecting and categorizing important information in text known as named entities. Named entities refer to the key subjects of a piece of text, such as names, locations, companies, events and products, as well as themes, topics, times, monetary values and percentages.

NER is also referred to as entity extraction, chunking and identification. It's used in many fields in artificial intelligence (AI), including machine learning (ML), deep learning and neural networks. NER is a key component of NLP systems, such as chatbots, sentiment analysis tools and search engines. It's used in healthcare, finance, human resources (HR), customer support, higher education and social media analysis.

What is the purpose of NER?

NER identifies, categorizes and extracts the most important pieces of information from unstructured text without requiring time-consuming human analysis. It's particularly useful for quickly extracting key information from large amounts of data because it automates the extraction process.

List of NLP elements
Natural language processing includes several handy elements, such as translation and information extraction.

NER delivers critical insights to organizations about their customers, products, competition and market trends. For example, companies use it to detect when they're mentioned in publications. Healthcare providers use it to extract key medical information from patient records.

As NER models improve their ability to correctly identify important information, they are helping improve AI systems in general. These systems are enhancing AI language comprehension capabilities in areas such as summarization and translation systems and the ability of AI systems to analyze text.

How does NER work?

NER uses algorithms that function based on grammar, statistical NLP models and predictive models. These algorithms are trained on data sets that people label with predefined named entity categories, such as people, locations, organizations, expressions, percentages and monetary values. Categories are identified with abbreviations; for example, LOC is used for location, PER for persons and ORG for organizations.

Example of named entity recognition use in text
Named entity recognition can identify and categorize key pieces of information in unstructured text.

Once an NER learning model has been trained on textual data and entity types, it automatically analyzes new unstructured text, categorizing named entities and semantic meaning based on its training. When the information category of a piece of text is recognized, an information extraction utility extracts the named entity's related information and constructs a machine-readable document that other tools can process to extract meaning.

What are the three types of NER?

The three most commonly used NER systems are the following:

  1. Supervised machine learning-based systems use ML models trained on texts humans have pre-labeled with named entity categories. Supervised machine learning approaches use algorithms such as conditional random fields and maximum entropy, two complex statistical language models. This method is effective for parsing semantic meanings and other complexities, though it requires large volumes of training data.
  2. Rules-based systems use rules to extract information. Rules can include capitalizations or titles, such as "Dr." This method requires a lot of human intervention to input, monitor and tweak the rules, and it might miss textual variations not included in its training annotations. It's thought that rules-based systems don't handle complexity as well as machine learning models.
  3. Dictionary-based systems use a dictionary with an extensive vocabulary and synonym collection to cross-check and identify named entities. This method might have trouble classifying named entities with variations in spellings.

There are also several emerging NER methods:

  • Unsupervised machine learning systems use ML systems not already pre-trained on annotated text data. Unsupervised learning models are thought to be capable of processing more complex tasks than supervised systems.
  • Bootstrapping systems, also known as self-supervised, predictively categorize named entities based on grammatical characteristics, such as capitalization, parts-of-speech tags and other pre-trained categories. A human then fine-tunes the bootstrap system, labeling the system's predictions as correct or incorrect and adding the correct ones to a new training set.
  • Neural network systems build an NER model using neural networks, bidirectional architecture learning models, such as Bidirectional Encoder Representations from Transformers, and encoding techniques. This approach minimizes human interaction.
  • Statistical systems use probabilistic models trained on textual patterns and relationships to predict named entities in new text data.
  • Semantic role labeling systems preprocesses an NER model with semantic learning techniques to teach it the context and relationships between categories.
  • Hybrid systems use aspects of multiple systems in a combined approach.
Comparison of machine learning models comparison
There are several different kinds of machine learning models, each offering unique advantages.

Who uses NER?

Various industries and applications use NER in different ways:

  • Chatbots. OpenAI's generative AI, ChatGPT, Google's Bard and other chatbots use NER models to identify relevant entities mentioned in user queries and conversations. This helps them understand the context of a user's question and improves their responses.
  • Customer support. Named entity recognition systems can organize customer feedback and complaints by product name and identify common or trending complaints about specific products or branch locations. This helps customer support teams prepare for incoming queries, respond faster and establish automated systems that route customers to relevant support desks and sections of FAQ pages.
  • Finance. NER can extract figures from private markets, loans and earnings reports, increasing the speed and accuracy of analyzing profitability and credit risk. NER can also extract names and companies mentioned in social media and other online posts, helping financial institutions monitor trends and developments that could affect stock prices.
  • Healthcare. NER tools can extract critical information from lab reports and patients' electronic health records, helping healthcare providers reduce workloads, analyze data faster and more accurately, and improve care.
  • Higher education. NER enables students, researchers and professors to quickly summarize volumes of papers and archival material, as well as find relevant subjects, topics and themes.
  • HR. These systems can streamline recruitment and hiring by summarizing applicants' resumes and extracting information, such as qualifications, education and references. NER can also filter employee complaints and queries to the relevant departments, helping to organize internal workflows.
  • News providers. News providers use NER to analyze the many articles and social media posts they need to review and to categorize the content into important information and trends. This helps them quickly understand and report on news and current events.
  • Recommendation engines. Many companies use NER to improve the relevancy of their recommendation engines. For instance, companies like Netflix use NER to analyze users' searches and viewing histories to provide personalized recommendations.
  • Search engines. NER is critical to search engines, identifying and categorizing subjects mentioned on the web and in searches. This helps search engines understand the relevancy of subjects to a user's search and provide users with accurate results.
  • Sentiment analysis. NER is a key component of sentiment analysis. It extracts product names, brands and other information mentioned in customer reviews, social media posts and other unstructured text. The sentiment analysis tool then analyzes the information to determine the author's sentiment. NER is also used to analyze employee sentiment in survey responses and complaints.

NER benefits and challenges

There are several benefits and challenges to NER.

Benefits of NER

Named entity recognition provides a range of advantages when used appropriately:

  • Automates the information extraction of large amounts of data.
  • Analyzes key information in unstructured text.
  • Facilitates the analysis of emerging trends.
  • Eliminates human error in analysis.
  • Is used in almost all industries.
  • Frees up time for employees to perform other tasks.
  • Improves the precision of NLP tasks and processes.

NER challenges

NER also comes with its own set of issues:

  • Has difficulty in analyzing lexical ambiguities, semantics and evolving usages of language in text.
  • Runs into problems with spelling variations.
  • Doesn't know all foreign words.
  • Can have issues with spoken word text, such as telephone conversations.
  • Leads to many state-of-the-art NER models reporting limited performance measures.
  • Can require large volumes of training data or a lot of human intervention.
  • Can be prone to bias in results if the ML algorithm has hidden bias.
Steps in the named entity recognition process
There are several steps involved with getting the NER process to produce analysis.

Natural Language Toolkit vs. SpaCy

Natural Language Toolkit (NLTK) and SpaCy are two NER programs with unique differences. NLTK is based on Python's NLP library and provides several algorithms. Often used for teaching NLP to beginners, as well as researchers building applications from the ground level, NLTK uses strings as inputs and outputs in preprocessing. It provides tokenization, stemming, part-of-speech tagging and parsing and can be trained on customized data.

SpaCy, on the other hand, is open source and uses a single stemmer algorithm well suited for concrete tasks. It is often used for building professional NLP applications and is object-oriented in preprocessing. SpaCy is also able to handle large data volumes, extract relationships between entities and offer support for word vectors. It is considered faster than NLTK.

Named entity recognition is a critical part of natural language processing. Learn how NLP benefits enterprise analytics.

This was last updated in March 2023

Continue Reading About named entity recognition (NER)

  • cloud security

    Cloud security, also known as 'cloud computing security,' is a set of policies, practices and controls deployed to protect ...

  • privacy impact assessment (PIA)

    A privacy impact assessment (PIA) is a method for identifying and assessing privacy risks throughout the development lifecycle of...

  • proof of concept (PoC) exploit

    A proof of concept (PoC) exploit is a nonharmful attack against a computer or network. PoC exploits are not meant to cause harm, ...

  • data collection

    Data collection is the process of gathering data for use in business decision-making, strategic planning, research and other ...

  • chief trust officer

    A chief trust officer (CTrO) in the IT industry is an executive job title given to the person responsible for building confidence...

  • green IT (green information technology)

    Green IT (green information technology) is the practice of creating and using environmentally sustainable computing resources.

  • diversity, equity and inclusion (DEI)

    Diversity, equity and inclusion is a term used to describe policies and programs that promote the representation and ...

  • ADP Mobile Solutions

    ADP Mobile Solutions is a self-service mobile app that enables employees to access work records such as pay, schedules, timecards...

  • director of employee engagement

    Director of employee engagement is one of the job titles for a human resources (HR) manager who is responsible for an ...

Customer Experience
  • contact center schedule adherence

    Contact center schedule adherence is a standard metric used in business contact centers to determine whether contact center ...

  • customer retention

    Customer retention is a metric that measures customer loyalty, or an organization's ability to retain customers over time.

  • virtual agent

    A virtual agent -- sometimes called an intelligent virtual agent (IVA) -- is a software program or cloud service that uses ...