Abbyy MailOrganizer

MailOrganizer is a cloud solution for intellectual processing of mail. Service is focused on the help to the person when processing unimportant letters in non-working or holiday time.

On April 1, 2016 the ABBYY company provided the product ABBYY MailOrganizer.

A system analyzes mail and announces important issues on which immediate reaction is required, can independently respond to a part of letters. These opportunities are implemented by means of ABBYY Compreno technology and to neural networks.

Representation of interaction of systems (2016)

ABBYY created MailOrganizer taking care of psychological health of workers in the world.

For training of technology the company assembled the body from 10 million letters in the English and Russian languages from different sources. After processing of the body the ABBYY Compreno engine, selected a recurrent neural network with two layers, an interface buried layer and a feedback. The technology considers background of the analysis and accumulates information for development of the correct hypothesis by means of the feedbacks going from exits of internal neurons, but not from an exit of network.

In the analysis of letters of MailOrganizer considers a terminological context, existence of different punctuation marks and their quantity, existence of grammatical, spelling and stylistic errors, the list of addressees and their order, both other semantic and syntax features. On the basis of analysis results one of categories is appropriated to each incoming letter: the general, postponed, conditional and important, important. The letters which are not requiring the urgent answer, a system will process independently:

• Will take the main thought from the general letters with a large number of addressees;
• Will postpone, creating a reminder in working time;
• Will automatically create the answer on the basis of the templates created by the user with preserving of tone and stylistics of the author.

Except templates of answers, the user can configure solution operating time, sounds of notifications, categories of mail, to correct scenarios the job got in the course of self-training of a system.

ABBYY MailOrganizer can be used for the analysis of the letters written in any language by means of technology of the analysis of understanding of the text in a natural language of ABBYY Compreno.

For April 1, 2016 in the analysis of a cluster from 10 million letters, indicators on quality are reached:

• 1 scenario – 3% of errors;
• 2 scenario – 4.5% of errors;
• 3 scenario – 1% of errors;
• 4 scenario – 8% of errors.

The indicator of a response was 146%, and response time – no more than 2.5 minutes.

The publication is prepared by the company on April 1, 2016.