The first stage of creating good weak labels is to explore your data, to understand it and come up with ideas for labelling rules. We've built in some powerful features that make data exploration much easier:
At the top of the screen is a search bar that lets you search across your corpus of documents and highlights occurrences of words.
You can use the filter at the top of the screen to narrow the data displayed to look at only one label at a time or the results of one labelling function at a time.
If you click on any of the rows in your data-points table it will expand that datapoint and show you the results of any labelling functions or manual annotation on that individual document. This is particularly useful when you have long documents like contracts.
The interface is split into two sections. On the right-hand side you see your data set and have the ability to explore it. On the left hand side you have a summary of the tags you're trying to extract and interfaces to write labelling functions. The impacts of the rules you write on the left are then displayed on the right.