The purpose of this paper is to show how archival expertise and diplomatic analysis can enrich the documentary corpuses analyzed by artificial intelligence tools. Here, the demonstration is based on a freely accessible material: the data from the great national debate launched in early 2019 by the French President Macron in response to the large-scale social movement known as the “yellow vests”.
Step 1 consisted of understanding the methods and conclusions of the providers responsible for processing the data of the great debate (1.5 million contributors). Step 2 was to analyze the formal elements of a random set of online contributions. Then, to compare the results.
This research shows that the processing of the data is based almost exclusively on texts, to the detriment of data on the source, date and arrangement of contributions, which could nevertheless be exploited as metadata.
The mass of data and the lack of online accessibility of part of the corpus did not make it possible to complete the experiment.
This research lays the foundation for other projects for collaboration between archival science and artificial intelligence tools.
There is a social challenge involving researchers in information sciences in public debate and governmental consultations. There is also an issue for a records manager to become more involved in the production of public records by promoting their specific skills in information management.
The originality of this paper is to show how archival science can help to improve the quality of the documentary corpuses used by artificial intelligence tools, and therefore, to improve the performance of these tools.
