SONDE: a database for exploring the semantics of nouns derived from verbs in French

Interest in the semantics of word formation has grown significantly in recent years, addressing research questions that require detailed semantic information and extensive data analysis. In this paper, we present the SONDE database, which was developed to investigate the semantics of nouns derived from verbs in French. The database relies on the manual analysis of 5272 deverbal nouns, annotated for their semantic type (taking into account the lexical ambiguity of both bases and derivatives), the lexical aspect of verbs and nouns, and the semantic roles assigned to their arguments. We first describe the construction of the database, including sample selection, the annotation scheme, and an evaluation of annotation reliability. We then analyze the collected data, focusing on (i) the similarity and polyfunctionality of deverbal processes, (ii) the preservation of argument structure and lexical aspect through nominalization, and (iii) the semantic organization of morphological families formed by verbs and their derived nouns. Results show that a network of semantic functions underlies the diversity of deverbal processes, partly shaped by semantically motivated relationships. It appears that the mechanisms of semantic extension that drive polyfunctionality also contribute to the emergence of similarities between processes. Another finding is that the aspectual and argumental properties of base verbs are not necessarily preserved in nominalizations, and that this non-preservation is conditioned by derivational patterns. Finally, the analysis of morphological families reveals distinct structural configurations depending on the semantic alignment of nouns across families. These configurations exhibit varying degrees of defectiveness and are largely determined by the semantic properties of the base verbs.