Frequently Asked Questions
AI Trainer can support both PDF and .docx document formats.
Your AI fields and documents are personal to your organization. Only other individuals within your organization will have access to the AI fields and documents.
AI Trainer can be used to build AI fields in other Latin script languages.
Your subscription to AI Trainer includes raw document storage up to 25GB.
There is no limit on how many AI fields can be trained, however, we recommend training no more than 5 to 10 AI fields on a set of documents at one time.
AI Trainer provides analytics for each AI field that you train that can be used to assess the AI field’s performance. The subject matter expert can review the AI Trainer-generated suggestions on unannotated documents to gauge accuracy of the AI fields’ performance. Once published to DocAI, AI fields can be further updated in AI Trainer and re-published to DocAI.
Training on top of a built-in AI field is currently not a function that exists in AI Trainer but is possible in DocAI with the training API. All AI fields that you build with AI Trainer will be entirely new AI fields.
Initiating Re-train on Available re-trains the field on all queued and previously trained annotations on documents that are still available in AI Trainer. Users may choose to delete documents that they no longer find relevant to that field and initiate Re-train on Available to effectively re-train the field quickly on only annotations that remain relevant to what they are trying to find. Zuva may, at times, in the future release updates to our behind the scenes ML strategies, users may choose to Re-train on Available to take advantage of these updated training strategies. You would only want to initiate Re-train on Available if you are dissatisfied with the performance of your field, as re-training may change the performance and is not a reversible action.