Source Code Level Word Embeddings in Aiding Semantic Test-to-Code Traceability

Proper recovery of test-to-code traceability links from source code could considerably aid software maintenance. Scientific research has already shown that this can be achieved to an extent with a range of techniques relying on various information sources. This includes information retrieval which considers the natural language aspects of the source code. Latent Semantic Indexing (LSI) is widely looked upon as the mainstream technique of this approach. Techniques utilizing word embedding information however also use similar data and nowadays enjoy immense popularity in several fields of study. In this work, we present our evaluation of both LSI and word embeddings in aiding class level test-to-code traceability of 4 open source software systems, the assessment relying on naming convention information.