Without using specifics, you will need to create an spatial relation between the two datasets. One method to do this is to create a buffer around the features of dataset1 of a set tolerance and compare the output with dataset2. Relations you should look for are:
- Dataset2 within Dataset1 - this will give you features that are very similar or follow the same path
- Dataset2 intersects Dataset1 - this will give you those features that deviates from the path or is possibly new
- Dataset2 not intersects Dataset1 - this will give you those features that are completely new.
The buffer can be used to manage tolerance to deviation between two datasets. A larger defined buffer will be more forgiving of larger path deviations.
An alternative is to use a similar method using the path endpoints of both datasets, where you create a buffer of dataset1 and if your spatial 'within' relation identifies those paths that are similar. Those endpoints that are not with in tolerance could be considered new paths.
Hope this give you some ideas.