This is a very crude implementation of an assisted learning process which can analyse Linux logs and identify the best possible error.
The statistics part has not been implemented, but using a queue based ranking algorithm. Have to further refine this with nltk.. To carry on this later..
-
written in emacs, so expected alignment problems..
-
Not the final piece but a sketch
-
basic idea is that the texts floser to the standard error messages in unix/linux have the highest possible option to detect an error. If one analyses the words closer to the error message, Tokenize using NLTK ,rank the error , , CONFIRM to the algorithm that this is the files/scripts which could be the problem, the algorithm will take care of analysing the error,