Cristina Garbacea, Samuel Carton, Shiyan Yan, Qiaozhu Mei. Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP). 2019.

Collecting a sufficient amount of 3D training data for autonomous vehicles to handle rare, but critical, traffic events (e.g., collisions) may take decades of deployment. Abundant video data of such events from municipal traffic cameras and video sharing sites (e.g., YouTube) could provide a potential alternative, but generating realistic training data in the form of 3D video reconstructions is a challenging task beyond the current capabilities of computer vision. Crowdsourcing the annotation of necessary information could bridge this gap, but the level of accuracy required to obtain usable reconstructions makes this task nearly impossible for non-experts. In this paper, we propose a novel hybrid intelligence method that combines annotations from workers viewing different instances (video frames) of the same target (3D object), and uses particle filtering to aggregate responses. Our approach can leveraging temporal dependencies between video frames, enabling higher quality through more aggressive filtering. The proposed method results in a 33% reduction in the relative error of position estimation compared to a state-of-the-art baseline. Moreover, our method enables skipping (self-filtering) challenging annotations, reducing the total annotation time for hard-to-annotate frames by 16%. Our approach provides a generalizable means of aggregating more accurate crowd responses in settings where annotation is especially challenging or error-prone.

Software APIs often contain too many methods and parameters for developers to memorize or navigate effectively. Instead, developers resort to finding answers through online search engines and systems such as Stack Overflow. However, the process of finding and integrating a working solution is often very time-consuming. Though code search engines have increased in quality, there remain significant language- and workflow-gaps in meeting end-user needs. Novice and intermediate programmers often lack the language to query, and the expertise in transferring found code to their task. To address this problem, we present CodeMend, a system to support finding and integration of code. CodeMend leverages a neural embedding model to jointly model natural language and code as mined from large Web and code datasets. We also demonstrate a novel, mixed-initiative, interface to support query and integration steps. Through CodeMend, end-users describe their goal in natural language. The system makes salient the relevant API functions, the lines in the end-user's program that should be changed, as well as proposing the actual change. We demonstrate the utility and accuracy of CodeMend through lab and simulation studies.