The main goal of this thesis is to design, implement and evaluate a module capable of selecting the most probable answer out of a given dataset of question and answer sentences. The module will utilize recurrent neural networks to achieve the task. The model should be optimized for use with the Czech language. The module implementation will be used as part of the AQA (Automatic Question answering) system. Detailed evaluation and error analysis will be provided alongside the implemented module.

The submitted thesis will consist of a thesis text and a practical part. The textual part will include a detailed description of neural question answering systems and related approaches, design and implementation of the answer selection module, and the details of its evaluation and error analysis. The practical part will contain all code and data for replicating the results.