Entity tagging in news data is the process of identifying and labelling the people, organisations, and locations mentioned in an article. Each entity is then linked to a canonical identifier so downstream systems can match coverage to known entities without manual disambiguation.
When an article mentions “Apple,” entity tagging determines whether it refers to Apple Inc., a different company with a similar name, or something else entirely — and attaches the appropriate identifier so the reference is unambiguous in every language and source the article appears in.
What is an entity in a news feed?
In a structured news feed, an entity is any named real-world object that appears in article content and can be identified and classified. The three primary entity types are:
- Organisations — companies, government bodies, regulatory agencies, NGOs, and other named institutions.
- People — named individuals including executives, politicians, public figures, and other persons of interest.
- Locations — countries, cities, regions, and named geographic features.
Some providers also tag additional entity types such as financial instruments, products, events, and topics, though coverage of these varies significantly between providers.
How does entity tagging work?
Entity tagging typically involves two stages: recognition and resolution.
Recognition is the process of identifying that a string of text refers to a named entity. Named Entity Recognition (NER) models scan article text and identify spans of text that refer to people, organisations, or locations. This is straightforward when an entity is referred to by its full official name. It becomes significantly harder when entities are referred to by abbreviations, nicknames, former names, or local-language variants.
Resolution is the process of determining which specific entity a recognised mention refers to, and linking it to a canonical record. Resolution is where most of the complexity sits. “Samsung” in an article could refer to Samsung Electronics, Samsung C&T, Samsung Life Insurance, or Samsung Heavy Industries — all distinct legal entities. Resolution determines which one is meant in context and attaches the appropriate identifier.
The quality of entity tagging in a news feed depends primarily on the quality of resolution, not recognition. Recognition is a solved problem for clean English-language text. Resolution across multiple languages, naming conventions, and entity types is where providers diverge significantly.
What identifiers are attached to tagged entities?
A well-structured news feed attaches multiple identifiers to each entity tag, covering different downstream use cases:
- LEI — for matching to compliance and financial systems that use Legal Entity Identifiers.
- FIGI — for matching to financial instruments in trading and investment platforms.
- PermID — for connecting to Refinitiv/LSEG data infrastructure.
- Wikidata ID — for cross-language entity disambiguation and knowledge graph enrichment.
Having multiple identifiers attached to the same entity means each downstream system can use the identifier type it already works with, without running a separate mapping step.
Why does entity tagging matter for platform builders?
Without entity tagging, filtering a news feed by company or person requires keyword matching — searching for every name variant, abbreviation, ticker, and translation of the entity you are monitoring. Keyword matching produces false positives (articles about a different entity with the same name) and false negatives (articles that refer to the entity by a variant you did not anticipate).
Entity-tagged news data replaces keyword matching with identifier matching. A compliance platform monitoring a counterparty queries by LEI rather than by name. A trading system routing news to a portfolio queries by FIGI. An intelligence platform tracking a named individual queries by Wikidata ID. In each case, the entity tag makes the match precise and consistent across all sources and languages in the feed.
This matters most for multilingual pipelines. A company mentioned in Japanese, Arabic, and Serbian will appear under different scripts, transliterations, and naming conventions. Entity resolution maps all of those variants to the same canonical identifier, so the monitoring system sees one entity across all three languages rather than three unconnected strings of text.
Opoint pre-attaches entity tags with LEI, FIGI, PermID, and Wikidata identifiers to articles across 250,000+ sources in 135 languages, resolved at the point of indexing. See how the enrichment layer works →
Trusted by
FAQ
What is the difference between entity tagging and keyword matching?
Keyword matching searches for specific strings of text in article content. Entity tagging identifies which real-world entity a mention refers to and attaches a canonical identifier, regardless of how the entity is named in the text. Keyword matching misses name variants and produces false positives when the same string refers to different entities. Entity tagging is more precise and consistent, particularly across languages and large source volumes.
What is Named Entity Recognition (NER)?
Named Entity Recognition is a natural language processing technique that identifies spans of text referring to named entities — people, organisations, locations, and other defined types. NER is the first stage of entity tagging: it finds the mentions. A separate resolution step then determines which specific entity each mention refers to and links it to a canonical identifier. NER alone is not sufficient for professional news data use cases; resolution quality is what determines whether entity tags are accurate and usable.
How are entities resolved across languages?
Entity resolution across languages requires mapping name variants, transliterations, and local-language forms to a canonical record. A company may appear under its legal name in one language, a localised trading name in another, and a transliterated version in a third. Resolution models trained on multilingual data and linked to identifier registries such as GLEIF (for LEI) and Wikidata can map these variants to the same entity. The quality of cross-language resolution is one of the most significant differentiators between news data providers.
Can entity tags include subsidiaries and related entities?
It depends on the provider and the identifier type. LEI Level 2 data includes ownership relationships, meaning a parent company's LEI can identify coverage of its subsidiaries when those subsidiaries have their own LEIs. Wikidata also includes organisational relationships. The extent to which these relationships are surfaced in entity tags varies by provider and should be confirmed during evaluation if subsidiary monitoring is a requirement.
What entity types are typically available in a news feed?
Organisations, people, and locations are standard across most structured news feeds. Some providers also tag financial instruments, products, events, and topics as entity types. Coverage of non-standard entity types varies significantly and should be tested against your specific use case during evaluation rather than assumed from a provider's feature list.
