Prediction Of Children's Abstract Concept Semantic Distance Using Word Co-occurrence Model Based On Children's Corpu

concepts refer to concepts that lack a single,bounded,and clear referent,and are closely related to human abstract thinking.Currently,research focused on adult populations widely accepts the multiple representation view of abstract concepts,which suggests that abstract concepts are represented by both linguistic and non-linguistic experiences,with linguistic experience playing a central role.However,language abilities of children at the primary school stage are rapidly developing,and their overall linguistic experience is still limited.Therefore,the role of linguistic experience in children’s representation of abstract concepts is not clear.With the continuous innovation of natural language processing technology,the co-occurrence patterns of words based on distributional semantic models(DSMs)have been proven to be an important way of representing abstract concepts/meaning of words in adults.Specifically,the higher the probability of co-occurrence between two words or the probability of having common co-occurring words between two words,the closer the distance between these two words in the semantic system and the more related their semantics.Therefore,this study first proposes to use DSM based on children’s language corpus to quantify the co-occurrence patterns in children’s linguistic experience(Study 1),and then explores the applicability and development of linguistic co-occurrence patterns in children’s representation of abstract words(Study2).The specific research is as follows:Study 1 uses the 2020 National Curriculum Textbook Catalogue for Compulsory Education and Guidelines for Reading Instruction for Primary and Secondary School Students(2020 Edition)released by the Ministry of Education,with a total of 10.147 million words,as the corpus for children.The high-performance word2 vec algorithm in current DSMs is used to train a distributed semantic space suitable for children’s groups(referred to as children’s DSM).At the same time,40 specific words and 40 abstract words are selected from the textbooks of grades 1-3,and a semantic-related word and two unrelated words are set for each target word.The performance of children’s DSM is detected using a synonym task.The results show that children’s DSM can select the high-correlation target options 100% correctly.Study 2 uses 160 third-grade and 180 fifth-grade students from a primary school in Tianjin as participants,with 50 abstract words appearing in grades 1-3 textbooks as materials,and 1225 word pairs are formed by pairing them up.Through the semantic relatedness rating task,the semantic relatedness of word pairs based on children’s DSM,non-linguistic experience,and subjective ratings are obtained for third and fifth-grade children,and the role of linguistic experience in the representation and development of abstract concepts is measured by using children’s subjective ratings as the standard and non-linguistic experience as a comparison.The study shows that,regardless of whether it is the third or fifth grade,the semantic distance based on the co-occurrence patterns of corpus words is significantly correlated with the semantic distance based on subjective ratings,and can significantly predict the semantic distance rating of abstract words for corresponding grade children.With the increase of grade,this predictive effect is also increasing.Compared with non-linguistic experience based on embodied features,the predictive effect of semantic distance based on co-occurrence patterns of words on the psychological semantic distance of abstract concepts in third and fifth-grade children is greater.In summary,under the conditions of this study,it was found that the representation principle of co-occurrence statistics of linguistic experience is also applicable to primary school children(7-11 years old),and its representation effect is enhanced with the increase of children’s language experience.In addition,compared with non-linguistic experience based on embodied features,the representation of co-occurrence patterns of words based on linguistic experience has a greater predictive effect on the psychological semantic distance of abstract concepts in children.