Core Concepts
Last updated
Was this helpful?
Last updated
Was this helpful?
Project is a space for you to hold your training data, annotations and your model and its versions. Project is associated with a Project type . You will be asked to choose a type while creating. Why is this important? Jasper comes with a lot pf pretrained models to extract content from standard documents like your Bank Statements, W2, 1040’s, Resumes and more. Based on the Project type you chose, these models will get activated and linked to your project. For ex if you chose, HR as your project type, ML model for resume attribute extraction will get linked to your project.
To link the pre-trained machine learning model with your training document , as the case with project type, you can chose the document's type , to say if its a Bank Statement, Resume or W2 Forms.
You can always chose the 'Other' type. The point to note, for the type 'Other' the model need to be build from scratch as there is no pre-trained model attached
Data labelling or annotation is the process of labelling data to make it usable for machine learning. Data can be almost any kind of data that a human might understand ex, images, text. In the context of Jasper, we are talking only about images (Scanned Forms), text ( PDF).
Concept behind training an ML model involves providing an algorithm (learning algorithm) with training data. ML model is a binary object or the output that is created by the training process. Jasper makes this complex process easy for you. All the complexities behind algorithm teaching are hidden and abstracted. All you need to know is , annotate your data, provide the right labels for the annotated data. With a click of a button, ML models are built in background. You will be notified when the model is ready to use.
ML models can’t be moved to production use or plugged into your decision process, unless its properly tested. Jasper provides you a interface with in the studio to upload a new document which was not used in the training process and see how the model performs to extract the relevant content. Jasper shows the preview of the extracted content , with the confidence scores indicating how accurately Jasper was able to extract a value from the image or text.
Jasper makes the integration of the ML models with your enterprise workflow simple and straight forward. No need to write any code to expose the models as API endpoints. All you need to do is hit the Publish button, give the model a name and its ready to be consumed as REST API.
Pre-trained model is a model created by Jasper and provided to you out of the box. Instead of building a model from scratch to solve a similar problem, you could use this model trained on our dataset as a starting point. For example, If you are looking to extract a W2 form, Jasper has already build a model and made available as part of the studio. You could start using it right away, but you could always add more data and add additional labels to improve this model.
Project Type is an optional feature, depending on the license you are assigned, this option may or may not be visible
Document type is linked to Project Type. Example, for the project type 'Banking' , the default options you will get is W2, 1040, Payslip, Statement
