52

Speeches produced by political parties are often reputed to be repetitive and uninformative. Yet, over time such speeches can yield extremely valuable information, as shown here by applying statistical methods to chronological series of texts. Analysis of a compilation of reports presented during the eight congresses of the Chinese Communist Party (CCP) held between 1982 and 2017 reveals variations of vocabulary over time in the form of textual highlights that point up key moments of lexical evolution (2). Textometric analysis sheds a critical light on the tokenization of texts into words as operated by automatic tokenizers of Chinese. More effective methods for observing lexical variations linked to ongoing political changes are then presented (3). Continuous changes in the enunciation style over the observed period are highlighted (4). The methodology described in this study is applicable to a very wide range of textual corpora.