| Dependency and valency are closely related,but they also differ in some aspects.Valency,which indicates the potentiality of a word to combine with other words,is a static description of a word.When the word enters a particular context,the potentiality is activated,forming a dependency relation.That is to say,the dependency relation is a realization of the valency relation.It has been nearly fifty years for relevant studies on valency,widely investigated in various languages,such as German,Chinese,Czech,English,and even interlanguages.But there seem to be few studies on the quantitative comparisons of valency regularities among different genres in a specific language.The present study adopts the definition of “full valency” proposed by (?)ech et al.(2010:291)and K?hler(2012:107),which means that “all arguments,without distinguishing complements(obligatory arguments governed by the verb)and adjuncts(optional arguments directly dependent on the predicate verb),are taken into account.In the following quantitative analysis,the term “valency pattern” is used,when it refers to the linear combination of dependency relations;and the term “valence” is used,when it refers to the number of various dependency relations in a quantitative manner.The current study employs texts in 5 typical genres(with 2000 words per text and a total of 250,000 words)in LCMC(the Lancaster Corpus of Mandarin Chinese)as our research corpora,namely,News,Essays,Official,Academic,and Fictions.Following the research paradigm of quantitative linguistics,the root-related valency patterns and valences of each text in 5 genres are quantified as research objects for the following calculation of the rankfrequency distributions,as well as the different percentages of dependency relations in varying valence zones,namely,low valence zone(with the valences of 1-3),medium-low valence zone(with the valences of 4-6),medium-high valence zone(with the valences of 7-9)and high valence zone(with the valences of 10-12).The aim is to discover the homogeneity and heterogeneity underlying different genes and testify to the feasibility of distinguishing genres by the parameters of the rank-frequency distribution calculated from the root-related valences.All research sentences are first parsed by Stanford Parser and then manually checked repeatedly,in line with the Chinese dependency annotation scheme of this study,which integrates the annotation practices in Liu Haitao and the Prague Dependency Treebank.Research data are processed in three steps: Firstly,analyze as a whole the distribution of root-related valency patterns and valences in five genres;Secondly,analyze as individual texts(25 texts in each genre and a total of 125 texts)the distribution of root-related valency patterns and valences in five genres;Thirdly,analyze the percentages of dependency relations in four valence zones and explore the features of valency patterns in five genres.This study attempts to answer the following three questions:(1)Do the rank-frequency distributions of root-related valency patterns and valences in texts of five genres,observed from the perspective of five genres as a whole,share some homogeneities?(2)Do the rankfrequency distributions of root-related valency patterns and valences in texts of five genres,observed from the perspective of individual texts,exert some heterogeneity?(3)Are there any peculiarities in the percentages of different dependency relations in varying valence zones in five genres?The following results are yielded:(1)When treating five genres as a whole,the distribution of root-related valency patterns and valences in five genres abides by the same regular distribution,that is,Zipf-Mandelbrot Law with different parameters,and with excellent fitting results.The distribution of root-related valences has also been corroborated by the Friedman sum rank test results.(2)When treating five genres as individual texts,the distribution of root-related valency patterns and valences in five genres also abides by the same regular distribution,Zipf-Mandelbrot Law with different parameters but with somewhat less satisfactory fitting results.The parameter b in the Zipf-Mandelbrot Law of root-related valences proves to be genre-sensitive;the genre of Official texts can be distinguished from the rest four genres.(3)The percentages of different dependency relations in varying valence zones in five genres seem to manifest the peculiarities to some extent.To be specific,in expository-prone genres of Official and Academic,the percentages of the dependency relation of “nummod”(numbers)in Official in the high valence zone are higher than the other three genres.In comparison,the percentages of“nummod” in Academic in the low valence zone are much lower than other three genres.This may characterize the objectivity and accuracy of these two genres.In narrative-prone genres,in Fiction,the dependency relation “neg”(negation)and “mmod”(modal verb)occupy more proportions,especially in the high valence zone,and remain constant in the low and medium-low valence zone,which may have something to do with the description and depiction of characters’ personalities and emotional attitudes in texts;in News,the dependency relation designating time(“tmod”),people(“nsubj”)and the things done(“dobj”)possess relatively high percentages in the high and medium-high valence zones,which is consistent with the primary purpose of News texts to disseminate information;in Essays,the percentages of the dependency relation “advmod”(adverb)is higher in the high valence zone than in other three genres,which accords with essential characteristics of its narrative style.The research on the homogeneity and heterogeneity of rank-frequency distributions of root-related valency patterns and valences in texts of five genres may be exploited to make for the objective,accurate,and scientific description of inquiry into different genres.It is hoped that this study will shed some light on future quantitative and contrastive research on genre analysis in different languages. |
PDF Full Text Request