An upcoming revolution for exploring mailboxes

Being the most important communication channel, email is the most common form of information collected during discovery in litigation.

If traditional ediscovery solutions were of some help, they demonstrate significant limitations in terms of deployment efforts and relevance.

Forget them! Recent developments in artificial intelligence open brand-new perspectives for exploring/reviewing mailboxes and for overcoming limitations of the traditional solutions.
Learn how to implement them as soon as possible.

The Next Generation Of Artificial Intelligence
Self-Supervised learning allows quicker deployment – leaving away the bias and constraints fixed by labelling and training

In a recent LinkedIn post, we shared a very interesting article from Forbes “The Next Generation Of Artificial Intelligence”, in which different emerging areas in artificial intelligence, expected to redefine the field, are commented.

The first one being the “unsupervised learning” – or in more precise terms used by Yann LeCun the “self-supervised learning” – announced to overcome the limitations of the widely spread supervised learning approach which implies an expensive and cumbersome labelling processes and a potential bias associated to human supervision.

Amongst sectors that will significantly be impacted by this AI area, the legal sector is holding a particular place: self-supervised learning indeed brings a very efficient support in the investigation of the exponentially increasing information to be reviewed, including emails, and opens new perspectives. Already now!

Why?

Simply because it allows quicker deployment – leaving away the bias and constraints fixed by labelling and training – and faster, deeper, more flexible and more accurate researches – integrating the contextual dimension and the ability to use natural language (words used in requests still remain key but more important is the context they describe by interacting with each other).

Priceless when you need to explore mailboxes: less efforts for more relevant results and open-ended exploration. This explains why solutions built on self-supervised learning are taking the lead in e-discovery (and supervised solutions are becoming legacy).

All of this explains why, at EisphorIA from the very beginning, we have developed in-house, our own self-supervised algorithms which provided significant gains in accuracy vs. supervised models, and we will certainly not deviate from this cap!

The benefits of adopting EisphorIA for mailboxes exploration

We use our solution to explore our own emails internally!


At EisphorIA, we use our solution to explore our own emails internally given the significant advantages it provides compared to Gmail, a.o.:

#

Advantage

What does that mean concretely?

1

Proven greater relevance in results (incl. ranking and ‘numbering’ of results in function of relevance)

On average -50% less investigation time and frustration

2

Full flexibility in the manner searches are conducted

Initiation of requests by keywords or natural language at our choice, without any kind of constraints

3

Ability to parameter the searches with intelligent filters

Narrow-down the searches in selected subsets of emails/attachments and/or focus on emails/attachments showing specific criteria (e.g. name, concept, reference,…)

4

Decomposition of the chain of emails and attachments (with of course, the ability to recompose the chain if needed)

Treatment of each email and attachment as separate documents when processing the request

5

Automated OCR-isation of any textual attachment in an image format

Exploration of the full content of the dataset with no burden linked to scans or other type of static textual documentation

6

Ability to group/identify emails or attachments presenting a very high degree of similarity

Direct identification of redundancies and successive versions of attachments,…

7

Flagging of the most important sections, in light of the request, within a document

Direct identification of most relevant portion within voluminous documents’ set

8

Ability to initiate search from relevant sections of document

Quick overview and review of emails / attachments dealing with the same subject-matter

9

Activation of translation modules

Possibility to discover critical information formulated in other languages

10

Recognition of all dates and other named-entities

Ability to work with timelines to focus on specific time-zones. Possibility of email threading. Possibility of pseudonymisation

11

Ability to proceed, highlight, annotate and flag documents

Optimal archiving of work done on e-mails and attachments during the review

12

Ability to organize the “watch” of the information

Identification of new uploaded documents that could be relevant considering previous key search requests


Those are massive benefits... that can be obtained today by deploying our solution on any kind of mailboxes environment.


Last but not least…

The other good news is that the deployment itself is very simple, not requiring any kind of effort for both the set-up and its use. As a child’s play!


Indeed, a closed and highly secured environment (on-premise or in the cloud) allows at any time to upload (or connect) without any difficulty the selected mailboxes: a few minutes of processing and the mailboxes are ready to be reviewed with all our functionalities.


Curious to know more about this? We would be very happy to connect with you.