Inference on Unseen Data Related to Conflicting/Duplicate Requirements Identification:

• Users can upload a CSV file of test requirements in a format (a pair of requirements text with a label) similar to benchmark datasets (see download section).


• On successful activation of processing command, the exploratory data analysis engine will process the data shortly to identify conflicting or duplicate requirements against the given pair of requirements texts.


• Users can download the result file after data processing by clicking the button.

Training the Model from Scratch

• Users need to provide a CSV file containing conflicting/duplicate requirements in a format similar to benchmark datasets (see download section).


• Users have the freedom to select similarity methods and text embedding models.


• Users can choose the size of the LLM vector to be projected using PCA.


• Users can choose the number of folds for data splitting, model layers, activation functions, epochs, batch size, and learning rate.


• Before starting the training process, users must sign up, preferably using an organizational email account, and provide the required data and purpose of experimentation.


• After approval, users will gain access for one-time training.


• At the end of training, users can download performance-related artifacts to analyze the model behavior.