I will work on a dataset of texts called “Deutsch-französische Kunstvermittlung”, which present texts published in German and French journals between 1870 and 1940, on Franco-German art mediation and art education. Ultimately, the goal is to provide researchers efficient tools for their search in the document database. Currently, we can only filter the documents by their title and date. I will attempt to automatically extract relevant information about each text, so that it facilitates the search in the database. An example of interesting attribute would be “which artwork is discussed in the text?”. Then we would allow to filter-search documents based on that information. That will be a task of Named Entity Recognition, and I will try to use state-of-the-art methods for it. Other information such as the type of impression (color or not) or the format could be interesting for the search. Identifying more relevant properties for the queries is also part of the job.