Release notes
Botminds V18.5
February 2023 release: corrective training for OCR, datasheet import from Excel formats, and pretrained OCR models added to AI pipelines in a click.
Botminds continues to keep up with its track record of multiple minor releases and a major release per quarter – every quarter since inception. We are happy to announce that this release of v18.5 comes with much sought after improvements and step changes – that have received great feedback from our users. Some of the key enhancements to the platform in this release are:
- Trainable OCR – corrective training for OCR in documents
- Data sheet import via excel– users can now import data from excel files (xlsx, xls, csv, tsv), in addition to SQL database
- Pretrained models– users can directly add these models to their AI pipeline
These key enhancements are elaborated further below.
Trainable OCR
The Botminds platform has proprietary Botminds OCR tool and third-party OCR tools (Google and Azure) integrated into the platform. In this release, the ability to do corrective training for Botminds OCR in documents and generate a model which will give more accurate OCR output has been introduced.
Further, a couple of improvements in OCR processed documents are also included – namely, correction in keyword highlight for OCR-processed documents; and more accurate drag selection in the documents.
Data sheet import via Excel
The ability to ingest structured data as input through SQL database (such as MySQL, Microsoft SQL) was introduced in v18. In the release v18.5, users can now import data from excel file formats (xlsx, xls, csv, tsv), configure columns, choose the index/ unique columns, update data types (e.g., date format) according to the data in the sheet.
The datasheet once imported can then be used to create a model and get predictions. Watch this space for further enhancements in data sheet capability which are lined up for upcoming releases.
Pretrained models
Pretrained models are models maintained by the Botminds team. Users can now directly add these pretrained models to their AI Pipeline – like they add custom models. The pretrained models introduced in this release are OCR-related – namely, Botminds OCR, Azure OCR, Google OCR.
This results in better accuracy of predictions and huge time-savings for users who now do not have to spend time or effort annotating the OCR output. Users can view the prediction results instantly. Only corrective training needs to be done over the pretrained models, if prediction quality is to be further improved.
Other key improvements and fixes in v18.5 are mainly in the areas of ingestion and RPA bots.
Ingestion
- Ingestion transparency - Users can view the progress of each document uploaded to the tool
- Customizable meta properties during ingestion; it is now possible to make meta property mandatory
RPA Bots
- UX improvements in import of RPA bots directly from external sources namely, Github and Azure DevOps
- Upgrade bots to latest repository changes – updates to bots get reflected automatically in the platform with a click of a button
- Public botstore is now available for users to access the bot library - universal access to the bots irrespective of their subscription.
- Bug fixes
Other improvements
- Project creation from templates
- UI improvement - Expose search & tab view for document set, menu re-categorization
- Expose training configurations to all users
- Product stability improvements
The updates in this release are in line with our laser-focus on enhancing the platform to provide increasingly better user experience and add capabilities that aid process transformation. As always, we are keen to hear your feedback. Reach out to your Botminds representative or contact us here or through community channels or LinkedIn.