Data and research intelligence

Early signals from research data

Connect companies, people and research through source-linked keywords and graph-based retrieval.

The problem

Company and researcher profiles relied on outdated keywords from conventional text extraction, across a corpus of more than 7 TB from many sources. Better keywords from patents and publications had to fit the existing database and backend.

What we built

Data pipelines and a backend that connect companies, people and research, using topic modelling, entity matching and graph-based retrieval.

Where it runs

In the customer environment, with a data warehouse, scheduled pipelines and a knowledge graph. Data refreshes weekly.

Operational outcome

Document-linked research retrieval operating across more than 7 TB.

In production in the client’s product, refreshed weekly.

Industry: investment intelligence. Client details and metrics are shared after an NDA, to protect our clients’ interests.

How it works

From source to a reviewed answer, in three steps.

01
Papers and patents
Research documents refreshed through scheduled pipelines.
02
Connect the records
Extract keywords, match entities and build graph-based retrieval.
03
Analysts explore
Follow signals and links in the profiles they already use.