Natural Language Processing reads what is written about your clients and turns it into evidence an analyst can use. It searches news and other online sources, recognises the people and companies an article is about, and works out whether what it says is a risk. Each finding is scored and attached to the client’s alert with its source, so the reviewer sees the evidence, not a pile of links.
The risk is often written down, not filled in.
Fields and codes tell a rule what it was told to look for. The rest sits in free text: news articles, payment references, address lines, in whatever language they were written. Natural Language Processing reads that text, finds the people, companies and places in it, and weighs how serious it is.
Zurich prosecutors have opened an investigation into Rheintal Logistics SA over suspected money laundering linked to shipments routed through Rotterdam. The company’s beneficial owner, Lukas Vogel, denied wrongdoing in a statement on Tuesday.
Nothing in this article is in a field.
Illustrative article. Highlight colours follow the markup used in DetectX® media search.
From a search to a scored, readable result.
Every document a media search returns goes through the same five steps. Click through them to see what happens to one article.
Search the web and news sources for a person or company, by country, mode, language and date range.
Zurich prosecutors have opened an investigation into Rheintal Logistics SA over suspected money laundering linked to shipments routed through Rotterdam. The company’s beneficial owner, Lukas Vogel, denied wrongdoing in a statement on Tuesday.
Zurich prosecutors have opened an investigation into Rheintal Logistics SA over suspected money laundering linked to shipments routed through Rotterdam. The company’s beneficial owner, Lukas Vogel, denied wrongdoing in a statement on Tuesday.
Illustrative search, articles and scores. Highlight colours follow DetectX® media search.
Every document gets a risk score and a sentiment score.
The score draws on what the text says and who it names: how serious the words are, the tone, and whether the people or organisations in it already appear in the risk database. When an analyst publishes the articles that matter, the alert carries an overall risk score based on them.
From the 2 articles published to this alert
Illustrative articles, scores and alert. Click a result to read its article.
NLP reads and scores. Your team decides.
Every finding comes with its source and a score an analyst can follow, nothing counts until a person publishes it, and the language model can run where your data already lives.
Illustrative articles, scores and alert.
NLP scores every article it finds. An analyst reads them and publishes the ones that matter, and only those count towards the alert’s risk score.
Each published article stays attached to the alert with its source and date, so anyone reviewing the case can read exactly what the score was based on.
Zurich prosecutors have opened an investigation into Rheintal Logistics SA over suspected money laundering linked to shipments routed through Rotterdam.
The language model is modular. It can run on your own infrastructure, so articles are read and scored inside your environment, not on a third-party service.
Every article gets a risk score and a sentiment score. Both come from what the text says and who it names, so an analyst can see why an article scored the way it did.
One media search, read by Natural Language Processing.
The same reading on the DetectX® MediaSearch screen, as an analyst runs it. Every step is yours to drive: nothing advances on its own.
Illustrative exampleInvented Names & Data
e.martiRules work with every capability.
The Natural Language Processing decides what happens. The other capabilities give it more to decide on. Pick one to see what it does.
What compliance teams usually ask before they choose Natural Language Processing.
News articles and other online sources returned by a media search. It works on unstructured text, the free-written content a rule can't read. Structured data such as records, fields and transactions is prepared by Extraction & Transformation.
Articles are read in the language they're written in. The examples on this page show English, German and French; ask us about the languages you need.
It recognises the people, organisations and places named in the text, then checks whether they match the name searched and whether they already appear in the risk database. An article that only mentions a similar name is scored lower.
Every article gets a risk score and a sentiment score, based on how serious the words are, the tone, and who it names. Only articles an analyst publishes count towards the alert's overall risk score.
No. Natural Language Processing is the capability that reads and scores the text. Adverse Media Search is the solution built on it, which runs the searches, manages the results and puts them on the alert.
Yes. The language model is modular, so it can be updated or replaced without changing how results reach the alert, and it can run on your own infrastructure.
Where NLP does its daily work: names searched across news and online sources, every article read, scored and placed on the alert.
See Adverse Media SearchSolutionThird Party Due DiligenceSuppliers, agents and partners checked before you sign: the same reading and scoring applied to the companies you do business with.
See Third Party Due DiligencePlatformName Screening IntelligenceHow names are matched to the right person, so an article about someone with a similar name doesn't count against your client.
See Name Screening IntelligenceBook a working session on your own sources. Bring your questions about coverage, keyword sets and how a scored document reaches the alert queue.
