A Comparative Study Of Knowledge Construction In Textbooks Of Artificial Intelligence And Education

Knowledge has long been studied by both sociology of education and linguistics.Bernstein takes knowledge as the object of study because the field of education has been overwhelmingly preoccupied with the analysis of its relation to society but has rarely studied the relations within education such as the relation between knowledge and education.Textbook,as a bridge that links the production and reproduction of knowledge in the educational setting,also attract much attention from scholars.However,previous studies on knowledge construction are mainly on the basis of theories either of education or of linguistics.Moreover,previous studies on textbooks of various subjects are mainly concerned about physics,history,English,etc.Few of them have explored knowledge construction in the textbooks in higher education institutions,especially the textbooks of artificial intelligence and education and made a comparison.Based on these backgrounds,this thesis aims to answer the following research questions:(1)How knowledge is constructed in textbooks through language?(2)What are the lexico-grammatical features realizing knowledge construction in the textbooks of artificial intelligence and education?This thesis applies both qualitative and quantitative methods.The former is carried out by a theoretical framework that integrates systemic functional linguistics and theories from the sociology of education including Bernstein's knowledge theory and Maton's legitimate code theory.Quantitative method is used with a corpus of four university textbooks for freshmen built by this study.To reduce the effect of personal writing style,two artificial intelligence textbooks and two education textbooks are selected.Then the distribution,regulation and recontextualization of knowledge are explored by examining the data in the corpus respectively.The distribution of knowledge can be analyzed by the taxonomic relations of terms in a subject.The regulation of knowledge is reflected by the ideational meaning of language which can be further realized by the participants and processes in a text.As for the recontextualization of knowledge,two concepts from the legitimation code theory are investigated.They are semantic density and semantic gravity.Semantic density is analyzed in terms of technicality that is realized by lexical density,percentages of type/token,and term frequency.Semantic gravity is analyzed in terms of the degree of relatedness to the context,and can be realized by generic/specific nouns,nominalizations,non-finite/finite processes.Both theoretical and empirical explorations of the research questions raised by this study result in the following major findings:First of all,the analysis of taxonomic relations shows that terms in artificial intelligence are organized more hierarchically while those in education tend to be more horizontally distributed.As a result,knowledge in artificial intelligence is mainly built by integration and subsumption while that in education is primarily built by accumulation.Second,both similarities and differences exist between these two subjects from the perspectives of participants and processes in the texts.Similarities show that both subjects have their most frequently-used types of participants tightly connected with the topic;both have more uncommon-sense participants than common-sense participants,which imply a high technicality in knowledge construction.Moreover,the first two main ways to build activities in both subjects of artificial intelligence and education are material and relational processes,which indicate that the knowledge in the textbooks of artificial intelligence and education are regulated mainly by doing and being.Differences lie in the fact that concrete participants,generic participants,technical terms and metaphoric process types occur differently in the textbooks of artificial intelligence and education.Artificial intelligence has more technical terms than education while the latter has more concrete,generic participants and metaphoric processes than the former.This suggests that although they both have a high technicality but artificial intelligence's is higher.The empirical analysis of processes shows that artificial intelligence has the material process as its most frequently occurring process while in education the relational process is used most frequently.It implies that artificial intelligence tends to build its knowledge more by the processes of doing and happening while education is more likely to use the processes of identifying and being to construct knowledge.Third,semantic gravity tends to be weaker in education but stronger in artificial intelligence,showing that knowledge in artificial intelligence is more related to its context.As for semantic density,artificial intelligence turns out to be stronger in terms of lexical density and technicality.Moreover,this study finds out two striking points:1)semantic density cannot be confirmed only by lexical density or technical term frequency alone;and 2)a stronger semantic gravity does not necessarily go with a weaker semantic density from a macroscopic perspective.This study is of great significance both theoretically and practically.It could enrich studies on knowledge construction and contribute to setting up a theoretical bridge which links sociology and linguistics together.From a practical perspective,this study can help both teachers and students to establish a deeper understanding of the intrinsic linguistic features of their own subjects.Consequently,they can write academic papers more appropriately to construct knowledge of their own fields and express their own academic views better in international journals.