Talks and presentations
Using Infersharp’s Static Analysis to Automatically Mine .NET Bug Patch Data. Uğur Y. Yavuz. Data&AI internship final presentation at Microsoft. 19 August 2022. Slides available upon request.
Abstract: Infersharp is a static code analysis tool capable of detecting and reporting errors in C# code related to memory safety, access violation, and concurrency among other issues. In this talk, we will discuss our efforts to create a tool that leverages Infersharp to extract detailed bug patch information from C# repositories hosted on GitHub. Specifically, our tool has been able to gather bug fix data at the granularity level of program methods, which is useful in the training of patch generation models.
A Machine-Verified Proof of Linearizability for a Queue Algorithm. Uğur Y. Yavuz. Master’s thesis defense at Dartmouth College. 11 May 2022.
Abstract: Proofs of linearizability are typically intricate and lengthy, and readers may find it difficult to verify their correctness. We present a unique technique for producing proofs of linearizability that are fully verifiable by a mechanical proof system, thereby eliminating the need for any manual verification. Specifically, we reduce the burden of proving linearizable object implementations correct to the proof of a particular invariant whose correctness can be shown inductively. Noting that the latter is a task that many proof systems (such as the TLA+ Proof System we chose to work with) are well-suited to handle, this technique allows us to shift the responsibility of verification away from the reader and onto a machine, by enabling us to produce mechanically verifiable proofs of linearizability. We then demonstrate the effectiveness of this technique, which heretofore had only been applied to problems of a smaller scale, by proving the linearizability of a well-known queue algorithm whose proof of correctness is known to be challenging.
Translation of Asylum Testimonials from Low-Resource Languages. Uğur Y. Yavuz. ISI Natural Language Seminar at the University of Southern California. 27 August 2020.
Abstract: Many asylum seekers along the southern border of the United States speak low-resource languages that are not available on commercial translation services. As a result, the translation of their testimonials poses a real challenge to the legal system, as well as non-governmental organizations. We will discuss potential techniques that would facilitate this task, such as transfer learning, domain adaptation, and corpus expansion, and also explain our work in compiling bilingual corpora for Mixtec and Kanjobal languages.