Most of the evidence is not in a database field
DetectX® Natural Language Processing transforms unstructured data and text into actionable insight, transcending the usual search capabilities. It works in tandem with the other engines to improve matching, detection and scoring, rather than running as an analysis of its own.
What NLP does here
Screening reads structured records. NLP reads everything else, scores what it finds, and hands the score to the engines that were already running.
From unstructured sources
Extract
Analyses textual data to extract the information that matters, enhancing the ability to identify risk factors from media and documents.
Web, databases and registers
Read
Any information source can be searched: Google, web media, and special sources such as LexisNexis, OpenCorporate and public registers, against predefined and custom keywords.
Risk and sentiment
Score
Evaluates text for potential risk, such as a compliance issue or a negative customer experience, and separately for sentiment. The resulting scores let users prioritise which risks to address first.
Sentiment over time
Sense
Reads customer feedback, social media interactions and other text sources to determine sentiment. Identifying the emotional response to a product or a service is also what lets trends in sentiment be recognised over time.
Automated content management
Qualify
Automates the qualification and scoring of content so it can be presented in dashboards, and keeps analysing and updating it, which is what makes the insight current rather than periodic.
Into the predictive models
Enrich
The sentiment and risk scores become variables in supervised learning, improving forecasts of customer behaviour, market trends and operational performance. Insight from the analysis then informs the refinement of those models in turn.
Where it is used
NLP is one engine on the DetectX® core. These solution pages depend on it.
Solution
Continuous analysis of the web, specialist databases and public registers, with each document scored.
Solution
Advanced text analytics behind the similarity score summary, and the negative-media check that runs beside the list check.
Solution
Sentiment and risk signals extracted from unstructured text enrich the client profile, so structured and unstructured data reach the same score.
Solution
Unstructured sources read alongside sanctions and provider lists.
Solution
Negative media read continuously alongside sanctions, ownership changes and political links.
Solution
Advanced data and text analytical methods applied to high-volume, high-speed monitoring.
Common questions
No. It is described as an integral part of the predictive analytics platform, and it works in tandem with the other engines to improve matching, detection and scoring rather than producing an analysis of its own.
Yes. Predefined and custom keywords combine with search inputs such as a name or a company, and which keywords fired is recorded with the document's score.
Three things are claimed. Detection improves, because scoring text for sentiment and risk gives the platform more to spot an anomaly or an emerging trend with. Risk management becomes proactive, because automated scoring surfaces a threat early enough to act on. And automating the reading of large volumes of text returns the hours a team spent on it, which is the least glamorous benefit and often the one people feel first.
The web-dev Adverse Media Search build states that analysis is not restricted to English and that results are delivered across multiple languages. No live page repeats it, so treat it as unconfirmed and settle the source set for your own deployment.
Natural Language Processing · DetectX®
See a document read, scored and filed.
Book a working session on your own sources. Bring your questions about coverage, keyword sets and how a scored document reaches the alert queue.
- Unstructured
- Media, documents and public registers, not only structured records.
- Two scores
- Risk and sentiment scored separately and both recorded.
- In tandem
- Works with the other engines to improve matching, detection and scoring.