Dependency network approach to the oral production of English- and Chinese-speaking healthy older adults

ABSTRACTLinguistic research has increasingly focused on the language of older adults, although findings remain controversial on whether language skills decline with age. Additionally, extant studies only deal with certain aspects of language, such as lexis or syntax, and few have systematically investigated this topic. This paper assumed a holistic perspective and dependency network approach to explore healthy older adults’ oral language at lexical, syntactic, and discourse levels. By constructing English and Chinese dependency treebanks, and corresponding dependency networks, of 100 healthy older adults and comparing them with those of all age groups, we found that 1) many parents appear before their dependents in the older adults’ English treebank due to the massive use of articles before nouns; in the Chinese treebank, the reverse applies because of the large number of aspect markers; 2) the mean dependency distance (MDD) of English older treebank is longer than that of the all-age group. Furthermore, there are many compound and complex sentences in older adults’ speech; the MDD of the Chinese treebank is shorter, while the syntactic structure conforms to grammatical rules; 3) the average degree ()s of both English and Chinese older adults’ dependency networks are higher and clustering coefficient (C)s slightly lower, indicating strong connectivity of words in the two networks and coherence in participants’ speech. In conclusion, there was no significant deterioration in the oral language of healthy older adults, although cross-language comparisons show very fine, detailed, and revealing differences between English and Chinese treebanks and networks.KEYWORDS: Healthy older adultsOral productionDependency treebankDependency networkCross-linguistic difference AcknowledgementThis work was supported by the National Planning Office of Philosophy and Social Science under grant number 22BYY079.Disclosure statementNo potential conflict of interest was reported by the author(s).Data available statementPart of the linguistic data supporting this study’s findings is extracted from DementiaBank at https://dementia.talkbank.org/, covering AD patients speaking English and Chinese and their healthy peers speaking English. Access to the data in DementiaBank is password-protected and restricted to members of the DementiaBank Consortium Group. According to the policy of the DementiaBank Consortium Group, we members are entitled to use the data concerned given the data source is marked clearly in publishing works.The other data are collected by the authors from the healthy peers speaking Chinese with both the participants’ and their immediate family’s verbal consent in this study.Notes1 TalkBank is a project organized by Brian MacWhinney at Carnegie Mellon University, whose goal is to foster fundamental research in the study of human communication, especially spoken communication. For more details, please see https://talkbank.org/.2 Stanford Parser 4.2.0 is an automatic program that works out the grammatical structure of sentences based on the dependency grammar. For more information, see https://nlp.stanford.edu/3 Pajek 5.16 is a tool for analysing large-scale complex networks. See more in http://mrvar.fdv.uni-lj.si/pajek/4 LTP (Language Technology Platform) provides a series of tools for Chinese natural language processing, which allows the users to divide and tag words, and analyse syntax. See more at http://ltp.ai/index.htmlAdditional informationFundingThis work was supported by National Planning Office of Philosophy and Social Sciences of China: [Grant Number22BYY079].