Researchers use datasets of Question-Solution pairs to train machine learning models, such as source code generation models. A Question-Solution pair contains two parts: a programming question and its corresponding Solution Snippet. A Solution Snippet is a source code that solves a programming question. These datasets of Question-Solution pairs can be extracted from a number of different platforms. In this research, I study how Question-Solution pairs are extracted from Stack Overflow (SO). There are two limitations of datasets of Question-Solution pairs extracted from SO: (1) according to the authors of these datasets, some Question-Solution pairs contain Solution Snippets that do not solve the question correctly, and (2) these datasets do not contain the information on how Solution Snippets need to be reused, and such information would enhance the reusability of Solution Snippets. These limitations of datasets of pairs could adversely affect the quality of the code being generated by machine learning models. In this research, I conducted a qualitative study to categorize various presentations of Solution Snippets in SO’s answers as well as how Solution Snippets can be adapted for reuse. By doing so, I identified eight categories of how Solution Snippets are presented in SO’s answers and five categories of how Solution Snippets could be adapted. Based on these results, I concluded several potential reasons why it is not easy to create datasets of Question-Solution pairs. The first categorization informs that finding the correct location of the Solution Snippet is challenging when there are several code blocks within the answer to the question. Subsequently, the researcher must identify which code within that code block is the Solution Snippet. The second categorization informs that most Solution Snippets appear challenging to be adapted for reuse, and how Solution Snippets are potentially adapted is not explicitly stated in them. These insights shed light on creating better quality datasets from questions and answers posted on Stack Overflow. / Graduate

Finite state machines are often used for modeling the decision logic for simulated systems. MathWorks’ Stateflow has a graphical user interface that allow users to model finite state machines. A Stateflow model can be added as a block to a Matlab/Simulink model and be executed seamlessly together. Stateflow blocks are developed as “charts” but they are natively stored as XML documents. This research explores the possibility of extracting the behavior of the finite state machines as defined in a Stateflow chart. This is done by parsing the corresponding XML document and reproducing this behavior in a C++ implementation that can be instantiated within a large, C++ based simulation system. Furthermore, the goal of this research is to develop a tool that will automatically generate an equivalent C++ representation, given an arbitrary Stateflow XML model. This research is performed in the context of developing highidelity powertrain simulations to be executed in High-Performance Computing environments.

Identification of source code authorship can be a useful tool in the areas of security and forensic investigation by helping to create corroborating evidence that may send a suspected cyber terrorist, hacker, or malicious code writer to jail. When applied to academia, it can also prove a useful tool for professors who suspect students of academic dishonesty, plagiarism, or modification of source code related to programming assignments. The purpose of this dissertation is to determine whether or not cross-entropy approaches to source code authorship analysis will succeed in predicting the correct author of a given piece of source code. If so, this work will try to identify factors that affect the accuracy of the algorithm, how programmer experience determines accuracy, and whether a cross-entropy approach performs better than some known source code authorship approaches. The approach taken in the research effort will manufacture a corpus of source code writings from various authors based on the same system descriptions and varying system descriptions, from which benchmarks of different approaches can be measured.

No description available.

No description available.

This sociolinguistic study of the bilingual speech of Spanish immigrants in Toulouse, France focuses on the phenomenon of code-switching (CS). The analysis of the data showed that most CS was situational, rather than metaphorical. Three types of CS were found: insertion, alternation, and congruent lexicalization. Their examination revealed that the insertion of French words was more common than of Spanish items, the alternation was most frequent in repetitions, and the congruent lexicalization was present at the grammatical and structural level. The speech of the individual participants was also analyzed, and it was found that the sociological aspects greatly affected the use of CS. Finally, the analysis of the frequency effects was conducted revealing that the topic of the nouns influenced the language in which the nouns were used. The results proved that it is impossible to conduct a reliable grammatical analysis without including the sociolinguistic aspects.

No description available.

No description available.

No description available.

Fuzzing, the method of generating inputs to run on a target program while monitoring its execution, is a widely adopted and pragmatic methodology for bug hunting as a means of software hardening. Technical improvements in throughput have shown to be critical to increasing the rate at which new bugs can be discovered time and time again. Persistent fuzzing, which keeps the fuzz target alive via looping, provides increased throughput at the cost for manual development of harnesses to account for invalid states and coverage of the programs code base, while relying on forking to reset the state accrued by looping over the same piece of code multiple times. Stale state can lead to wasted fuzzing efforts as certain areas of code may be conditionally ignored due to a stale global. I propose Closure, a toolset which enables programs to run at persistent speeds while avoiding the downsides of stale state and other bottlenecks associated with persistent fuzzing. / Master of Science / The process of program testing to find bugs is becoming increasingly automated. A current method called "Fuzzing", is a widely adopted means for finding bugs and is required in the life cycle of program development by major companies and the US Government. I look at current improvements in fuzzing, and expand the use case of the cutting edge method called persistent fuzzing to a wider array of applications with my tool Closure. With Closure, fuzzing practitioners can experience faster fuzzing performance with less manual effort.