ThoughtRiver Blog | Latest in AI Legal Tech

Lexible - breaking down the complexity that lawyers face

Written by Dom Hudson | Jun 26, 2019 10:31:13 AM

Wording in contractual documents can be difficult for even a human to interpret! At ThoughtRiver, we have designed a system called Lexible to begin to break down this complexity that lawyers face on a daily basis.

Lexible is an ontology of "properties" which encapsulates complex relationships present in an agreement. It includes thousands of individual questions that together build up a granular picture of legal obligations, indemnities and rights that may lie within a document. ThoughtRiver datapoints are not isolated units; but instead work as part of an ecosystem - together building to form an understanding of the contract.



In a contract, relevant points of meaning can be distributed across large amounts of context. Caveats are often grammatically disjunct from influenced clauses. In the opposite extreme, hugely impactful points of meaning can be localised to a single word or implied through use of a reciprocal term.

The biggest part of data-science is effective evaluation of your data and one of the easiest ways to do this is visually.

Below is a diagram of the property "op-confidentiality-prior". This Lexible datapoint is asking the question "Does the agreement specify that your confidential information includes information disclosed to the counterparty prior to the date of this agreement?". It follows a number of more general questions regarding confidentiality.



I have used the unsupervised machine-learning algorithm tSNE to project the training data for this property down to a three-dimensional representation. The original data was over one-thousand dimensions. The beauty of the tSNE algorithm is that it retains complex and nuanced relationships between dimensions, whilst allowing us to visualise the data in a much more manageable space.

For more information on how our pre-screening software can help your business be more efficient, click on the demo link below or contact us and we will be in touch!