Using Distributional Structure to Model Semant?

The ability to then do something generative. Obviously depends on your use case, right? If this is just some web ap or whatever may be, it just goes for the highest accuracy. So what i wanted to know was tet to give up as evidence that it is legitimate for us to use distributional structure to propagate our provecty knowledge. There's a learnable matta between these two. We weren't using it for its own sake. It was more like that was a convenient way. And we also tested out different types of models and found they were all very similar.