Report

Teaser, summary, work performed and final results

Periodic Reporting for period 1 - VIZLEGAL (Building the world\'s largest Application Programming Interface (API) for law, Vizlegal will change how legal information is utilised and understood by the legal industry.)

Teaser

1. Technical Report1a. Publishable SummaryContext and objectivesOur service is now live and available to try at www.vizlegal.comVizlegal has spent the six month duration of the SME-1 programme working on a number of aspects of the legal information industry, including; market...

Summary

1. Technical Report
1a. Publishable Summary

Context and objectives

Our service is now live and available to try at www.vizlegal.com

Vizlegal has spent the six month duration of the SME-1 programme working on a number of aspects of the legal information industry, including; market exploration, customer interviews, onboarding trial users, and gathering feedback. This process has led to building of three distinct prototypes layered on top of accumulated and structured legal data.

In our proposal we said: “Law is one of the last big industries yet to be seriously disrupted by technology. We believe that all legal information globally needs to be structured in order to create an entirely new legal informatics industry. We are coupling legal data with machine learning and artificial intelligence, specifically around solving classification and recommendation problems, leading to enormous opportunities for innovation and growth.”

We believe this still to be fundamentally the case. If anything our views have been re-enforced by our ability to prototype products using data that we ourselves have structured. One objective or concept of our original proposal - the mining of citations between court judgments to develop recommendation algorithms - remains an objective, but during the early part of SME-1 we discovered more fundamental user problems in the Irish, and other, legal systems.

We focussed on these problems (outlined below) in order to create solutions that we felt would more immediately address customer and market needs - and as a result we have garnered significant interest from both practising barristers and solicitors in Ireland. We have also built the infrastructure for graph citations recommendations (for CJEU case law, mapped in a graph database), that will allow us to productise a recommendation system.

Work performed

Work performed

Our work revolved three discrete strands of the proposed SAUL system, but before delving into those strands it’s important to lay out the foundations that were built to allow us to build three innovative and discrete product features.

Vizlegal has moved from a beta to a fully featured product in terms of several core feature:

We have built a production Extract/Transform/Load (ETL) system that can scale to address any of our future needs to extract data from public websites, and deal with any formats (PDF, Word, etc), including Optical Character Recognition (OCR) of scanned documents.
We have a fully featured API that can integrate with other systems.
We have built a fully featured Elastic cluster that allows us to index our growing document database
We have built a production web application for desktop, that can also be used as a Progressive Web Application (PWA) for iOS/Android on mobile or tablets.
We have ingested close to 1m documents from multiple courts, quasi-judicial tribunals and courts systems in four jurisdictions. These documents have been parsed into highly structured data, and are fully indexed

These core features were either built or improved upon during the course of SME-1. This included a switch to a new frontend framework that allowed us to adapt for (4), based on user feedback requiring a greater focus on mobile accessibility than we had originally envisaged.

The resulting production product now addresses a number of fundamental problems for both barristers and solicitors, particularly in the Irish market. But we see this needs as being addressable across multiple and larger markets in similar contexts to the Irish one.

Core features:

During the course of SME-1 we improved or built from scratch new UI elements for each of these core features. Because of a focus on mobile accessibility in SME-1, all of these features have also become easily accessible on phones and tablets.

Search; Suggestions; Filters. This feature is a core need for users to either search the documents in our database quickly, or find a document they know we have quickly (suggestions); and rapidly zone in their searches through a process of filtering by jurisdiction, court etc.
Advanced search. Build a query as fine tuned as possible, before searching and filtering.
Bookmarks. Users can bookmark a judgment for example, and tag that judgment with keywords.
Alerts. Users can create alerts such as “alert me when a new judgment appears that contains the phrase “freedom of information”, specified by the user for any or all courts.
Case tracking. In the Irish High Court context, a feature that allows litigators to track the progress of their case over time, and be alerted when changes occur on the public file. This can act as an early warning for litigators, while also assisting with case management.

Rapid SAUL prototypes:

Each of these prototypes were built during the course of SME-1.

Consolidated Court Rules and Practice Directions (PDs) (Version 1) (Fully featured)

In the Irish context, a consistent pain outlined during the course of user interviews was a frustration around court rules and PDs. In order for any practitioner, both solicitors and barristers, to engage with the courts system they must rely on the rules of the courts set by statute.

Rules are amended on average every 3-4 weeks by new statutes, along with brand new rules and PDs being introduced. This has a direct effect on practitioners, as it affects how they carry out their work. Existing solutions are deemed inadequate - either books, or an out-of-date official website. There are risks, including reputational and practice risks, to a practitioner using a rule that is out of date.

To address this we hand-consolidated all the rules of Irish courts going back to their origins in 1986. To do this we also built a bespoke Git-based document versioning system, which allows us to “version” the laws over time. We improved the aesth

Final results

Results achieved

We have built three key elements to the overall SAUL system that formed a part of the overall SME-1 proposal:

A legal API; an ETL system; an Elastic search system; a web and mobile UI
A GraphDB that links the citations of all CJEU decisions to each other
Discrete SAUL prototypes: Git-consolidated court rules for two jurisdictions; law firm analytics; judge profiles for the Irish jurisdiction

Our results have been demonstrated to active practitioners, including four paying customers who utilise our case tracking core feature on a daily basis. We outline here briefly the results achieved from the building of this technology:

API usage: During the course of SME-1 one research Machine Learning (ML) Masters student interested in building algorithms that could predict the outcome of cases of the European Court of Human Rights (ECHR) that we have ingested and parsed from their official website; and one data platform based in London that resells access to our structured corpus of England and Wales court judgments as far back as 2004. We are in talks for a third potential API customer/integration with a news company based in Ireland.
Application usage: Customers using and paying for the core application.
Court rules: Following demonstration of our consolidated court rules and PDs feature (for mobile and desktop) and judge profiles before members of the Bar Council of Ireland (the representative body for all barristers in Ireland), we have moved towards subsequent demos to further members, which has resulted in an active trial both with the Bar Library, and with members of the Bar Council Library Committee, who are deliberating on signing up all barristers in Ireland to the service. As we had assumed based on user interviews mobile access combined with easy of use and consolidation of rules were Unique Selling Points (USPs). We have also been approached by several law firms who see value for their litigation teams (and estimate the product would save 100s of man hours per month).
Law firm analytics: We have demonstrated this feature to a number of firms, including the second largest firm in Ireland by revenue. The team considered this both to be novel and useful in the context of competitive intelligence. They saw further opportunity to augment the data with other data they have identified. The idea of using the tool in the context of potential mergers and acquisitions was also broached, since client identification was a USP.
Judge profiles: Barristers have shown exceptional interest in this feature. One trial user indicated it would give him a “definite edge” as it allowed queries on phones while moving between courts before judges with which the barrister may not be familiar. They also have suggested features for future versions of the product.

Website & more info

More info: https://www.vizlegal.com.