What Goes Into a Foreign-Language Media Research

“Do you cover foreign-languages?” is a client question we receive fairly often. Unfortunately for the person asking, they sometimes get a thirty-minute explanation of the nuances of work involved. This is an attempt to distill some of that down just to the high level details.

The expectation for users is to enter a name (any way they want) and get relevant information. In practice, doing that reliably is difficult even within a single language, and more so across languages.

A provider might answer with a list of languages it covers, but having access to a source and being able to consistently find the right information in it from user search inputs are very different things.

The name you enter may not be the name in the source

Consider a search entered in Chinese as “公司名稱”. Local reporting may use the company’s Chinese name, while international reporting uses an English company name. Someone searching in English still expects to find the Chinese articles, and someone entering the Chinese name may want the English reporting too.

Connecting those names takes a lot of work. To make that information searchable, names need to be processed as structured data, preserving the original script and adding transliterated versions. Searches can then check both forms using fuzzy matching, which allows for differences in spelling without requiring an exact match. Having relevant information in a database is only useful if the name entered by the user can lead to it.

The search terms have to change too

For adverse media, a live web search might combine a name with terms like “fraud,” “launder,” or “crime.” That makes sense for an English-language search on John Smith. Apply those same English terms to a company name entered in another language, though, and they may do little to help find relevant information.

Each language needs search terms that reflect how the issues are described locally. A direct translation can be grammatically correct and still miss the terminology used in local news and other public sources. The terms also need to fit the category being searched, since a word can have different meanings depending on the context.

We have worked with clients to build and refine these search terms over years of actual research, drawing on experience running millions of searches a month. That experience and client feedback help us adjust terms that miss relevant information or return too much unrelated material.

Finding a match is only part of the job

The results still need to be reviewed to determine whether they concern the right subject and contain something relevant. An article can mention a company and fraud without accusing that company of anything. It might be the victim, a supplier, or simply mentioned in passing. Translating the article into English does not resolve those questions by itself.

The relevant details then have to be presented in a language the user can work with. That means preserving who did what, when it happened, and whether the source describes an allegation, an investigation, or an established outcome.

The above only scratches the surface, but these are the kinds of problems we work on at Threat.Digital every day. Foreign-language coverage depends on how well those parts work together. Information can be missed because a source was unavailable, a name was written differently, or a search used the wrong terms. Even when a source is found, the result is only useful if it has been matched to the right subject and interpreted correctly.

There is unlikely to be a point where all of this is solved. But experience running real searches, reviewing the gaps, and working with clients gives us a way to keep improving coverage