From a 50,000-line chat export to a clear case picture with i2 TextChart

Few things scare an analyst more than a freshly minted folder containing six months of group chats and a deadline of yesterday. i2 TextChart will not investigate the case for you — but it will get you from text to entity-level structure inside a single afternoon.

What TextChart actually does

TextChart is an extension to IBM i2 Analyst's Notebook that runs Named Entity Recognition over a body of unstructured text and instantly populates a chart with entities (people, organisations, phone numbers, accounts, locations) and the links between them. Think of it as the bridge between an evidence dump and the analytical surface where you actually do the work.

A workflow that scales

1. Normalise the source data. Convert exports to plain text or RTF; strip system messages, app metadata and emoji noise.
2. Run TextChart with a domain-specific dictionary. We maintain dictionaries per case category (drug trafficking, financial crime, CSE) and they make a measurable difference to extraction quality.
3. Triage the chart. Merge duplicate entities, prune obvious noise, then group by role.
4. Layer in structured sources. Telecom CDRs and bank statements come in next; TextChart's entities act as the anchor for the rest.
5. Iterate with the lead investigator. A 30-minute review with the case lead is the highest-leverage thing an analyst can do at this point.

Common mistakes

The two we see most often: blindly trusting the auto-extracted entity types (always merge by role manually) and importing too much at once. A single chart with 18,000 nodes is unreadable; eight smaller charts linked through a top-level summary work far better.

When to bring in our team

If the dataset crosses a million records, includes voice transcripts, or needs to be married with banking data and OSINT, our analytics consultants will sit alongside your unit for the duration of the case. We work under NDA, embed in your tooling, and leave the methodology behind.