Making Documents Intelligent: How Metadata Powers Market Discovery

By Cristina Belchin, Head of News & Research at EMIS

Metadata turns a pile of documents into a map of a market by tagging each document with the entities it describes, companies, people, industries, countries and events. At EMIS, we are working to expand this approach with richer entity coverage, including people and executives, products, events to create more comprehensive connections across content and link entities to one another.

Every day our content operations at EMIS process a large volume of news, filings, reports and data from emerging markets. On its own, each document answers one narrow question. Connected through metadata, they start to answer the questions our clients actually ask: what matters in this market, how are the entities linked, and what is changing?

What is metadata in market intelligence?

In today’s data flooded environment, raw information is useless without context, which is why metadata serves as the key differentiator for market intelligence. By providing a structured architecture of labels, such as language, timestamps, geographic location, identifiers and relationships, metadata transforms fragmented data points into actionable strategy.

It acts as the connection between information streams, ensuring that modern analysts accurately interpret the shifting competitor behaviours and consumer trends behind them.

Without this critical layer of contextual data, the market signals can be misunderstood.

A news article about a bank raising capital in Brazil is as text, a few hundred words. With metadata, it becomes a set of precise connections: „a specific bank with a unique identifier“, „the banking industry“, „Brazil“ and „a date“.

Additional enrichment, including more detailed events identification, extraction of quoted executives and executive-to-company relationships, is an area of ongoing development.

What does good metadata look like?

A well-designed metadata helps the platforms bring together information from different sources, uncover connections that might otherwise be missed and make unstructured content ready for advanced analytics and AI applications.

There are several key building blocks that make this possible:

1. Standardised entity mapping: Making sure companies and organisations are correctly identified, even when they are mentioned in different ways. They are mapped afterwards to the global identifiers such as ISIN, CUSIP and LEI, which helps avoid confusion and creates consistent links across data sources.
2. Cross-industry taxonomies: Frameworks such as NAICS help create common classifications, making it easier to link information across markets, companies, supply chains and external data providers.
3. High-precision timestamping: Knowing exactly when information became available matters, especially in fast-moving financial markets. Standardised ISO 8601 timestamps help establish a consistent timeline and connect new information with market events.
4. Data lineage and provenance: Knowing where information comes from and how it has been processed is essential. Tracking original sources, ingestion times, document identifiers and transformation history make data more transparent, traceable and reliable.
5. Sentiment and semantic enrichment (in development): Looking beyond keywords to understand the meaning and tone of financial content. We are exploring capabilities such as sentiment analysis to help turn textual information into measurable signals that can be used in analytics and predictive models.
6. Combining automation with human expertise: Automated extraction provides the scale needed to process large volumes of data, while human review helps address complex cases, resolve ambiguities, and ensure accuracy across languages and markets.
7. Keeping metadata continuously updated: Without ongoing maintenance, metadata gradually becomes outdated, reducing its reliability and turning a current information map into a historical record.

Which entities does metadata connect?

Five entity types can play an important role in market research. While company, industry and country metadata are already widely used, richer event, people and product metadata is an area we are continuing to develop.

Entity	Question it answers	What tagging captures
Companies	Who is acting?	A unique identifier, legal name and aliases, parent and subsidiaries
People	Who is deciding?	Potential metadata attributes include role, employer, tenure, board seats and links to other companies. We are currently working on expanding and maintaining this type of metadata.
Industries	Where does it fit?	A consistent classification (for example NAICS), primary and secondary activities
Countries	Where does it apply?	Country of operation, of incorporation, and of the event itself
Events	What happened?	Event type (M&A, IPO, default, regulation, leadership change), date, parties involved. Currently also under development

Companies are the anchor of most market maps. The hardest part is not tagging a name, but resolving it: recognising that a transliterated name, a local-language name and a ticker all refer to the same legal entity.

Industries provide a common language for comparison. Shared classifications bring related companies and sectors together, ensuring that information can be consistently discovered, analysed, and compared, regardless of the terminology used in individual reports

Countries add the macro context: regulation, currency, political risk. In emerging markets, where a company’s home country and operating country often differ, which is an important signal for the analysis.

More detailed events (under development). It will help to identify events specific to companies and improve signals around companies’ activities.

People/executive. Shared board memberships can reveal important corporate connections. For example, an executive who sits on three boards links three companies together, yet that relationship often remains hidden unless it is intentionally recorded and mapped.

How does metadata turn documents into a map?

When each article, report and filing is tagged to shared entities, independent documents become connected through the companies, people, sectors, countries and events they have in common.

Each tag is a connecting point to every other piece of information bearing the same tag. Multiply this across thousands of documents and the result is a multidimensional network that a researcher can navigate.

This changes the research workflow in three practical ways:

• From searching to navigating. Analysts can follow connections between companies, people, industries, and events rather than continuously rebuilding search queries for each new research question.

• From reading to comparing. Consistent metadata and tagging make it possible to compare companies, sectors, regions, or countries side by side, revealing patterns and differences faster.
• From snapshots to timelines. Event-based metadata provides a historical view of how companies, industries, and markets evolve over time, enabling researchers to track change rather than assess only the current state.

Why is metadata essential for AI-powered research?

AI research tools are only as reliable as the structure underneath them. A large language model can summarise a document well, but, for instance, it cannot reliably know that two differently spelled names refer to one company.

Well-structured metadata gives AI systems three things they need:

• Grounding. Answers can be traced back to specific, identified entities and sources rather than inferred from loose text.
• Precision. A question about “mining companies in Peru with leadership changes this year” becomes a filter across tagged entities and events.
• Context. Linked entities let a system bring in related information: the parent company, the regulator, the peer group.

The same logic applies to answer engines and AI search. When content is clearly structured around named entities and explicit relationships, it is easier to retrieve, cite and trust.

Why does metadata matter more in emerging markets?

Metadata matters more in emerging markets because the information is more fragmented, more multilingual and less standardised than in developed markets. The same company may appear under a Cyrillic, Latin and local-script name. Disclosure rules vary by country. Many important companies are private and rarely covered by global media.

In that environment, a simple text search misses as much as it finds. Consistent entity tagging is what lets a Turkish-language news item, a Polish company filing in the local language and an English-language industry report all point to the same business.

This is the core of our metadata enrichment work at EMIS. Collecting content is only the first step. The lasting value is in making that content connect.

---

By Cristina Belchin, Head of News & Research at EMIS, ISI Markets

Cristina Belchin is the Head of News & Research at EMIS, where she leads the teams that collect, process, curate, produce, structure and enrich news and research across the emerging markets. Based in Sofia, she has more than a decade of experience in document integration, metadata enrichment and ML-driven tagging, complemented by nearly 10 years of prior experience as a business journalist working for international business newswire.

---

Learn more about EMIS, leading provider of reliable industry and company intelligence across emerging markets, and explore our recently launched AI assistant ASkISI.