Abstract: 【Purpose/significance】The rule of the development of children's sentiment has always been a concern of various parties.
However,existing researches have shortcomings in the long-term,accurate and efficient collection,processing and analysis of sentimen‐
tal data.【Method/process】The research firstly collects the data of children's free narrative text from grade 1 to grade 6 in primary
school,then uses the text sentiment analysis to identify the sentimental state of narrative text,and finally uses polynomial regression to
study the linear and nonlinear trend of sentimental development【. Result/conclusion】The results show that,with the increase of grades,positive sentiment generally presents a curve downward trend,negative sentiment presents a curve upward trend,and neutral sentiment presents a sinusoidal shape in the whole development process【. Innovation/limitation】Although there are limitations such as the inabil‐ity to determine direct causality from the text,free-narrative text sentiment analysis still provides researchers with the opportunity to use "big data +AI" technology to conveniently,accurately and efficiently explore the rules of children's long-span sentimental develop‐ment.