After decades of massive digitization, a sub-stantial amount of documents exists in digital form. The accessibility of these documents is strongly impacted by the quality of document indexing. Most of these documents are indexed in noisy versions that include numerous errors. The noise can be due to manual input mistakes or optical character recognition process and results in errors like spelling mistakes, missing characters, and others. This paper presents a study of the impact of noise on document ranking, an essential task in natural language processing (NLP) with wide-ranging practical applications. We provide a deep and quantitative analysis of the impact of recognition errors on document ranking by testing two popular ranking models on several noisy versions of a subset of the MS MARCO passage ranking dataset, with various levels and types of noise. Our study provides insights into the challenges of document ranking under noisy conditions and advocates for developing ranking models that are more robust to noise.

The earth observation industry provides satellite imagery with high spatial resolution and short revisit time. To allow efficient operational employment of these images, automating certain tasks has become necessary. In the defense domain, aircraft detection on satellite imagery is a valuable tool for analysts. Obtaining high performance detectors on such a task can only be achieved by leveraging deep learning and thus using a large amount of labeled data. To obtain labels of a high enough quality, the knowledge of military experts is needed. We propose a hybrid clustering active learning method to select the most relevant data to label, thus limiting the amount of data required and further improving the performances. It combines diversity- and uncertainty-based active learning selection methods. For aircraft detection by segmentation, we show that this method can provide better or competitive results compared to other active learning methods.