Fig 3 - uploaded by Alexander LeClair
Content may be subject to copyright.
F1 Score for code-only and code-description standard datasets. We observe a significant drop in performance from codedescription to code-only in all cases, though this drop was lower for nn+cd than nn+w. Best overall performer was nn+cd.
Source publication
Software Categorization is the task of organizing software into groups that broadly describe the behavior of the software, such as "editors" or "science." Categorization plays an important role in several maintenance tasks, such as repository navigation and feature elicitation. Current approaches attempt to cast the problem as text classification,...
Contexts in source publication
Context 1
... observe that both baselines suffer a significant drop in performance, when faced with code-only data versus codedescription data, as clearly visible in Figure 3. Consider first lines 1 and 5 in Table I. ...
Context 2
... observe that both baselines suffer a significant drop in performance, when faced with code-only data versus code- description data, as clearly visible in Figure 3. Consider first lines 1 and 5 in Table I. These lines correspond to the linear regression for code-only (line 1) and code-description (line 5) data. In line 5, the F1 score on the standard dataset is 65, and precision/recall is 68/64. These results are broadly similar to the 67% F1 score reported by Wang et. al [5] when using text description data (mined from the web) with linear regression to classify Java projects from SourceForge (though the results are not directly comparable due to different categories, datasets, and programming language). However, when text data is not available, we observe a drop in performance to 34% F1 score, 43/33 precision/recall -a nearly 50% performance penalty as measured by F1 ...
Similar publications
Software Categorization is the task of organizing software into groups that broadly describe the behavior of the software, such as “editors” or “science.” Categorization plays an important role in several maintenance tasks, such as repository navigation and feature elicitation. Current approaches attempt to cast the problem as text classification,...
