Quantcast
Channel: Electronic Discovery - E-Discovery Resources and Guides » recognition
Viewing all articles
Browse latest Browse all 2

Image Content Recognition Software for eDiscovery and Litigation Review – Electronic Discovery

0
0

Image Content Recognition Software for eDiscovery and Litigation Review

NEIL - Never Ending Image Learner - What if your Optical Character Recognition software was more like Image Content Recognition software? By Josh Headley, Discovery Engineer, D4 LLC

Merely a decade ago, most discovery projects involved storage warehouses and filing cabinets full of paper. Everyone cheered when Optical Character Recognition (OCR) showed up to the party, and the ensuing fracas lead to the conception of traditional keyword searching. Fast forward to 2014: even the most mundane eDiscovery project showcases a variety of modern technology – such as conceptual searching and predictive coding – designed to help legal teams find and digest information faster. There is, however, one stealthy category of evidence that eludes even the most well-planned, multimodal culling strategy: IMAGES.

Images come in many forms and from many sources: petrified PDFs with no electronically-extractable text; pictures from a Facebook profile; stock photos; screen captures; digital artwork; vector graphics; movie frames; and, those lovable company logos in JPG or GIF format in your e-mail signature. Although born from a variety of different sources, all share one commonality: they are completely undiscoverable using keywords, conceptual analytics, or predictive coding. To date, the legal industry has generally observed an unwritten gentlemen’s agreement that these elusive leopards of e-discovery can remain in the shadows.

Sure, you can OCR those image-only PDFs. You can run those little logo GIFs through OCR and, if the text is formatted cleanly and with sufficient contrast, you may get some searchable text. But what if your Optical Character Recognition software was more like Image Content Recognition software? What if the computer could record the fact that there is a beach, water, sunshine, swimmers, and a volleyball in that vacation selfie? The software could then begin to piece together common associations among items that frequently appear in the same scenes.

Enter NEIL.

Never Ending Image Learner, or NEIL, is a creation of Carnegie Mellon’s Robotics Institute. NEIL leverages technology and concepts from the growing field of computer vision to identify, classify, and associate objects in images. In the same way that conceptual analytics makes associations among textual attributes such as diction, frequency, order, and proximity, NEIL can form a conceptual mapping of digital objects in a 2-dimensional plane. Beaches are associated with oceans, volleyball can be played on a beach or in a gymnasium, and there are never many people on the beach unless the sun is shining.

The potential implications for the legal world are profound. A patent enforcer could run a query for images that contain their client’s brainwork, either alongside a competing product or possibly in the same picture as some manufacturing equipment. Counsel for pharmaceutical or medical device companies may query for pictures of the clients’ products being used improperly or in improper settings. Law enforcement may be able to quickly query surveillance images for a tall person wearing a red jacket holding a purse.

The possibilities for image analytics are limited only by the number of visible objects in the world.

 

Related Content

Admissibility of ESI in eDiscovery – A Facebook Tale

Is Social Media the New Email for eDiscovery?

Tips to Avoid Social Media and New Technology Blunders

 


Background:

Image Content Recognition Software for eDiscovery and Litigation Review
Source: original article
Author: d4admin
Categories: Electronic discovery, e-discovery, ediscovery

This e-discovery news is syndicated from e-discovery websites and blogs that make their feed available via RSS. Contact us to have your RSS feed added or removed.


Viewing all articles
Browse latest Browse all 2

Latest Images

Trending Articles





Latest Images