Investigators

Need automation for investigation?

ElinarAI enables law enforcement, investigation, and intelligence communities to tap into unstructured data as never before. With ElinarAI intelligence professionals can automate analytics of large volumes differing unstructured data types and use results with more traditional analytics and export results into IBM i2®.

The possibilities of ElinarAI for Investigators are presented below. You will find more detailed technical information here.

ElinarAI for business link1Tech about ElinarAI link 2

ElinarAI for Investigators

ElinarAI allows investigators (for example Money Laundering, Tax Evasion, Ponzi-schemes, Fraud) to create “Investigation Specific AIs or ISAIs”. This allows automating scanning and entity recognition on a large scale with unmatched accuracy.

A common problem at the start of the investigation is that there is a lot of confiscated data to go through. For example, 200 TB. This will take humans years. What if you could train ISAI (Investigation Specific AI) at the beginning of the investigation and it would automatically scan trough all 200 TB of data, flag interesting data and extract all entities into an analytical platform. Imagine that instead of 200 TB there would be 400 Gb data that would have been pre-annotated and fed into a premier analytical platform? That would then allow you to export all entities into IBM i2®!

ElinarAI™ product logo

Talk more with our expert:
Elinar's CTO Ari Juntunen, specialist in augmenting and automating with Artificial Intelligence (AI)
Ari Juntunen
Chief Technology Officer
+358 40 524 4482

ElinarAI for Investigator accomplishes this using several profoundly powerful steps:

1. Capture

the expertise of your best investigators

Initially, we need to create a training data set. Let’s consider the Money Laundering case. At the start of the investigation, we use our best specialists to work through some of the material to be investigated. They will highlight interesting information (connections between organizations and individuals for example) and classify each document to “Mundane”, “Possibly Interesting” and “Crucial”. They need to spend some time on this. At the end, the result is the ISAI data set that captures their expertise and can then repeat the process tirelessly for all data.

2. Create

ElinarAI ISAI (IT task)

In order to get good results, there are a few important steps to do before AI can be taken into production. The first step is to create “Glorified Named Entity Recognizer” to support this investigation. This is done by analyzing the data set and identifying crucial terms. Then the data set is run trough ElinarNER and AI is trained using IBM Watson Machine Learning Accelerator. This step will output functional AI that can be deployed into production.

3. Scan

through all (200 TB) data

In this step we use a premier analytics solution from IBM called IBM Watson Explorer (WEX) scan trough all data. As WEX scans trough data it feeds the text from each document into ElinarAI ISAI and gets analyzed results back. WEX then combines traditional text analytics results that come out-of-the-box with WEX and ElinarAI results and allows the investigator to drill into the material with extreme speed and power. If data includes image-based documents (like scanned TIFF images or PDF with embedded TIFF images without text layer) IBM Datacap is used to transfer images into PDF with text layer so they can be analyzed.

4. Analyze

and gain huge benefits

ISAI augmented data has now been loaded into Watson Explorer (WEX). WEX analytical user interface allows the investigator to drill into data and relationships within data using all augmentations provided by ISAI. This allows them quickly to find and focus on real evidence that matters. This will cut downtime needed for investigative work significantly. When interesting data has been identified it can then be exported into IBM i2 for further analysis.

5. Build

organization-wide investigative AIs that grow and get more accurate

Data set from step 1 can be used as a basis for the next investigation if it is within a similar domain. Next time investigation specific training would be much shorter as the cases are similar. As an example Tax Office could have after a while extremely powerful AI to detect issues related to Money Laundering or Police organization could have similar AI for highlighting how Drug Money is moved around that can be used by any investigation related to Drug Money. By using this approach, it is possible to build extremely powerful assets that will grow over time to combat crime and fraud.