Uh oh!
There was an error while loading. Please reload this page.
Add components of paths mentioned in comments as features of the component model - #6505
Add components of paths mentioned in comments as features of the component model#6505marco-c wants to merge 2 commits into
Conversation
suhaibmujahid
left a comment
There was a problem hiding this comment.
The test task is failing.
| # In case of stack traces, the position is important (e.g. component of the first path is more indicative than component of the last path, | ||
| # which tends to be a "main"-like entry point common to most stack traces). We weight each mention by the reciprocal of its rank among all | ||
| # paths mentioned in the comments, and sum the weights of repeated mentions of the same component. | ||
There was a problem hiding this comment.
What if we make the component and the position as one feature, and then the model will learn the importance based on the position, which could learn to detect outlier patterns.
There was a problem hiding this comment.
What do you mean by "one feature"? Do you mean one-hot COMPONENT-POSITION? That would explode quickly with too many combinations, no?
There was a problem hiding this comment.
Yes, we would try that and see (we could limit the number of components through the mapping). Then, if it does not work, we can get one for the first line, another for the last line, and then for all things in the middle, we could use weights.
Train on Taskcluster: component