On usual procedure we need to write the step definition and data storage classes wrt PM model. For each of the feature steps we need to map that to relative stepdefinition i.e method which performs action.
This eventually manual and time consuming effort.
Solution is to automate cucumber test script generation using NLP technique. Tensor flow uses features files and converts the features files to required step definition methods(Java).Separate set of classes will be generated as per the application UI pages. Object repository will be created dynamically when user supplies the steps in feature file.
Tool scans the html page and retrieves all required locators using intelligent locators generating strategy(to achieve this we need to supply strong data as part of feature files)
This eventually leads to no code automation using cucumber with web driver.