Conga AI Analyze leverages a core base of legal knowledge, machine learning technology, and rule-based criteria to accurately analyze and identify Contract document data as key legal clauses and field values. It is important to understand how each Conga AI Analyze component interacts and affects one another to maximize the product's effectiveness.
Review the sections below to follow and implement Conga AI Analyze best practices.
Do not oversaturate Blueprints with Clause and Field criteria.
The Contract Analyzer uses artificial intelligence and Blueprint criteria collaboratively to match contract document data with a Blueprint's Clauses and Fields. Each Clause and Field criteria acts as a rule during the Contract Analyzer's evaluation process. The Contract Analyzer first uses its artificial intelligence and machine learning technology to classify each paragraph in the document as a type of legal Clause (ex: Jurisdiction), as well as classifying each "named entity" as a type of Data Field (ex: Effective Date). The Clause and Field classifications are all attributed with a level of accuracy, or percentage of confidence that the classification is indeed correct. The accuracy and confidence increases as more samples are processed through the system (ex: Training Data).
After classifications are applied to each Clause and Field, the Blueprint criteria (rules) are used to "boost" or "attenuate" the confidence of the classifications. For example, the AI engine is 70% confident that a particular paragraph is the Limitation of Liability clause. Furthermore, there is criteria in the Blueprint regarding the Limitation of Liability clause that evaluates to TRUE (ex: a keyword of "limitation" was found). AI Analyze will now "boost" the confidence of the particular paragraph from 70% to 90% because there is Blueprint criteria that evaluates to TRUE.
Use criteria to strategically narrow down and identify data that correlates to Fields and Clauses, but do not oversaturate Fields and Clauses with criteria. If there are too many rules (criteria) for a Clause or Field, the Contract Analyzers artificial intelligence may inaccurately disregard data due to all criteria being unfulfilled. Too many rules for a Clause or Field also prevents machine learning from factoring into future analysis. If Blueprints exclusively use criteria to match data with Clauses and Fields, the criteria acts as a crutch and does not allow machine learning to occur. The machine learning component of Conga AI Analyze is critical towards customizing the product to individual customer use cases.
Use regex to define Clause and Field criteria.
Regex is the most effective method to define criteria in a Blueprint. Using only literal words and data values limits Clause and Field criteria functionality.
Select the correct data if the Contract Analyzer does not identify the Clause or Field, or if it selects the wrong Clause or Field.
If the Contract Analyzer does not find document text for a Clause or Field, ensure that you highlight the text within the document that matches the Clause of Field correctly and save. Likewise, if the Contract Analyzer matches a Clause or Field to document text incorrectly, highlight the correct text within the document and correlate it to the correct Clause or Field.
Adding and correcting Clauses and Fields within the Contract Analyzer is the feedback loop for online machine learning. Each time a user corrects the Contract Analyzer, the machine learning technology factors in the correction and becomes more accurate in future analysis. This process of initial correction is key for the Contract Analyzer's effectiveness and efficiency.