<p>This paper makes important substantive and methodological contributions. Substantively, it shows how Chinese political discourse has changed overtime under specific Premiers. It highlights a shift from revolutionary mobilization to technocratic management. Methodologically, it shows how Natural Language Processing (NLP) tools of automatic textual analysis, based on the Stanford CoreNLP and Stanza freeware NLP packages, can link qualitative and quantitative social science. We apply NLP tools to a corpus of 54 English translations of China’s Government Work Reports (CGWR). Results based on part-of-speech, Dependency Relations, and Named Entity Recognition tags illustrate how various word classes are used (e.g., nouns, verbs). “Development” is the most frequent noun with a wide range of co-occurring adjectives (e.g., economic, social, steady). Passive verb voices and nominalizations increase over time, revealing a tendency to remove agency (who did it? Who is responsible for some actions? ). Modality also increases, with higher overtime use of weaker forms of modality, such as “we should,” versus the stronger “we must,” which reveals a tendency toward reduced moral commitments by the government. The NER tag PERSON shows that during Xi Jinping’s presidency, Xi became almost an exclusive focus, indicative of a personality cult. Algorithms of sentence complexity, vocabulary richness, and text readability reveal an overtime simplification of the CGWR’s prose, denoting efforts to reach wider audiences. The SVO (Subject-Verb-Object) extractor shows that the pronoun <i>we</i> and missing subjects of passive clauses are the most frequent SVO subjects. The increase in passive constructs and nominalization and a softening of governmental commitment through shifting modality reveal a tendency to hide agency.These combined results ultimately validate the use of an NLP-driven “distant reading” methodological approach to texts beyond both simple word-frequency counts and qualitative “close reading.”</p>