ReMatch & Obé - machine learning message matching framework and app

Description

ABSTRACT
Symphony doesn't have a built in ability to match messages based on machine learning principles and provide matches to a business-side application

PROPOSAL
ReMatch is a machine learning message matching framework which uses Machine Learning and other techniques to match messages based on classification and provides matching messages via an API. Obé is an app built to demonstrate the framework by providing a web UI and pre-canned classifications for messages which would be interesting to compliance departments such as Bribes, Over-gifting, Threats, etc. We plan to open source both the framework and the application

BACKGROUND
N/A

RATIONALE
Many companies have built in systems to perform matching on other correspondence types such as email, but typically don't have a similar system for instant messaging. It also may be cost-prohibitive for smaller companies to implement such a system as they are often commercial and expensive. They may also have a very steep learning curve. The framework is simple to provision and train, and new classifications can be added by the end-user

CORE DEVELOPERS
Core Committers: Greg Olmstead <greg.olmstead@rbccm.com>, David Leung <david.l.leung@rbccm.com>, Amit Sharma <amit.sharma2@rbccm.com>, Burcu Kilic <burcu.kilic@rbccm.com>

INITIAL GOALS
We have some parts of the system which need to have a more complete implementation, but out of the box, the system has the capability to match on trained data. Currently, the system works via a bot, but once Symphony rolls out On Behalf Of authentication, or the compliance group has access to the Firehose API, this will be usable across the pod. The feedback feature will also provide supervised learning to the machine learning algorithm, but for now it does allow flagging and dismissing of the messages in the UI for demonstrative purposes.

Assignee

Gabriele Columbro

Reporter

Greg Olmstead

Due date

None

Priority

Medium

Labels

Configure