Grammatical inference, or grammar induction, studies how to learn automatically implicit rules behind some sequential data. This domain has a real scientific purpose and can be useful in numerous domains like natural language processing or bioinformatics as they often manipulate sequences. The tool we use to describe these data is the formal grammar. There exists some categories of grammars that are more expressive than others and therefore, more complicated to learn. In order to infer these expressive grammars some options have been developped. One of them is to make substitutable languages assumptions. The goal of my internship is to search for methods to improve the results of expressive grammar inference using these subsitutable languages. These improvements will be especially made for practical applications, and more particularly for biological sequences. In this report, I will first describe the state-of-art algorithm that can learn an expressive class of substitutable language and how it is currently implemented. Then, I will develop how I improved the current parser to make it useful for real cases. Finally, I will talk about my contributions to improve the learning capability of the state-of-the-art algorihtm adapting it to a new class of subsitutable languages.