Week 11: Final Results

As mentioned in the last post, there were a lot of frontend heavy tasks that awaited completion.

While two of the three Build views were completed, the pairwise comparison view was still in progress. This week, a push to complete that was established. Pairwise comparison did not come in without problems. Unlike the other two views, pairwise allowed for objects to be reused more than once. This leads to the problem of circular input. An example would be a user picking that MIT > WPI and then picking another pair that states that WPI > MIT. This leads to a need for loop detection, the algorithm for which was designed.

The output of the machine learning was just spit out in a json file to the screen. To make it more user interactive, we worked on outputting that json output to a table.

For better UX, instructions on how to use the different Build views were implemented.  

There was a bug in the search implementation where a searched objected appeared in the search even if it was being used in the comparison. 

When there were less than two outputs given to the backend, there would be an error. Therefore, a validator for the RANK! button was implemented such that the button is clickable only after users ranked at least two items.

The final website now looks like as follows:


Popular posts from this blog

Week 31: Fair Ranking Search

Week 32: Datasets

Week 26: Finishing the Paper