TI Automate

Automation of document processing and indexing is critical to modern clinical studies, given the sheer volume of documentation associated with even a single trial. However, it can be difficult to be certain that an automated process can easily handle discrepancies and judgment calls that a human document reviewer would handle.

With TI v10.5, the Trial Interactive machine learning capability will award a confidence score for those documents processed using TI Automate, the machine learning and metadata extraction service. The confidence score is calculated using an algorithm that assigns weight to (a) the success of document classification and (b) the success in the extraction of metadata fields.

Suppose the indexing and extraction of a particular document exceeds a configurable confidence threshold. In that case, the document will move automatically to the next workflow stage, skipping the indexing stage completely and moving on to the review stage.

With TI v10.5, users of TI Automate can easily select which documents need to be reviewed personally and for which documents the system was able to correctly and confidently perform indexing.