Call Me Maybe: Using NLP to Automatically Generate Unit Test Cases Respecting Temporal Constraints
Arianna Blasi, Alessandra Gorla, Michael D. Ernst, Mauro Pezzè

Towards Detecting Inconsistent Comments in Java Source Code Automatically
Nataliia Stulova, Arianna Blasi, Alessandra Gorla, Oscar Nierstrasz

Translating code comments to procedure specifications
Arianna Blasi, Alberto Goffi, Konstantin Kuznetsov, Alessandra Gorla, Michael D. Ernst, Mauro Pezzè, Sergio Delgado C.

RepliComment: identifying clones in code comments
Arianna Blasi and Alessandra Gorla

A Decade of Code Comment Quality Assessment: A Systematic Literature Review
Pooja Rani, Arianna Blasi, Nataliia Stulova, Sebastiano Panichella, Alessandra Gorla, Oscar Nierstrasz

Replicomment: identifying clones in code comments
Arianna Blasi, Nataliia Stulova, Alessandra Gorla, Oscar Nierstrasz

MeMo: Automatically identifying metamorphic relations in Javadoc comments for test automation
Arianna Blasi, Alessandra Gorla, Michael D. Ernst, Mauro Pezzè, Antonio Carzaniga

Panelist of Ask Me Anything – Untapped, Underestimated, and Underresourced Communities in CS/IT within the Diversity and Inclusion Events-track"

Automatic Generation of Test Oracles from Natural Language Artifacts
Arianna Blasi

Research on NLP for RE at Università della Svizzera italiana (USI): a Report
Arianna Blasi, Mauro Pezzè, Alessandra Gorla, Michael D. Ernst

Semantic-based Analysis of Javadoc Comments
Arianna Blasi, Alberto Goffi, Konstantin Kuznetsov, Alessandra Gorla, Michael D. Ernst, Mauro Pezzè, Sergio Delgado C.

Jdoctor
Jdoctor (originally known as Toradocu) is a tool that takes in input the Java source code of a class, extracts the Javadoc semi-structured comments at the method level, and translates @param, @return and @exception tags into executable specifications. Such specifications can act as test oracles and can be integrated with an automatic test case generator: Jdoctor is currently integrated with Randoop. Up to now, Jdoctor's precision and recall computed over 800 translations are respectively above 90% and 80%.

RepliComment
RepliComment is a prototype born in 2018 that takes in input the Java source code of a whole project and searches for comment clones at the method and field level. RepliComment relies on some heuristics to understand whether a comment clone could be legitimate (e.g. overriding) or if there is an issue such as an unintentional mistake of copy-and-paste. RepliComment was extended in 2020: Now it automatically assesses the severity of the clones and provides fix suggestions for the copy-and-paste mistakes.

MeMo
MeMo is a tool that takes in input the Java source code of a class and extracts method summaries (i.e., unstructured comments) to look for metamorphic relations. If found, a MR is translated into an executable assertion that can be automatically integrated into existing test suites. MeMo so far discovered over 250 MRs across different Java projects, which it can translate with good precision and recall. The translations, when used as assertions, prove to help both regression and developers-written assertions in bug detection.

CallMeMaybe
CallMeMaybe uses natural language processing to analyze Javadoc comments to identify temporal properties. This information can guide a test case generator towards executing sequences of method calls that respect the temporal properties. CallMeMaybe is currently integrated with Randoop. Up to now, CallMeMaybe's precision and recall are respectively above 80% and 70%, and the translated constraints proved to improve thousands of Randoop's outputs.