Ranks relevant files in top10 with 70% accuracy.
CodeSearchFinder-AI ranks, on average, with 70% of accuracy, one or more relevant code files in top-10.
Given a service request and a source file, our approach computes two kinds of scores for the file using our unique AI logic: lexical similarity score and probabilistic score, given by Vector Space Model.
We obtain each score with 4 search types, using a different set of terms indexed from the service request and the code file.
For each of the 8 combinations of scoring, we rank all files in descending order. After this, for each file we take the best of 8 ranks.
We treat each service request and code file individually (using the summary, stack trace, stemming, comments and file names). In other words, when available and relevant, i.e. when they improve the ranking.