Abstract: 【 Purpose/significance】 At present, the normalization of journal impact is mainly based on discipline, overlooking the differ⁃ ences among various topics within disciplines and the diversity of journal literature types. In light of this, this study aims to construct a topic-level journal impact normalized indicator JNI (Journal Normalized Indicator), providing a reference for journal evaluation. 【Method/process】 In this research, the topic model, normalized method and citation analysis were used to construct the indicator. Then we selected 163 journals in the field of "Information Science & Library Science" of JCR list and downloaded their documents from 2019 to the 2021 for the calculation of JNI. Subsequently, a comparative analysis is conducted between JNI and multiple journal evaluation indicators, exploring their correlations and disparities.【 Result/conclusion】 The research findings indicate that JNI demon⁃ strates the normalization effect at the topic level of journals, with a minimal influence from self-citation rates. JNI exhibits a strong positive correlation with the indexes of comparison, which means JNI has validity. But JNI has a better discrimination degree than other indicators because of subject analysis, full literature calculation and standardization, JNI highlights the difference of periodicals in the division and ranking.【 Innovation/limitation】 In this paper, a new impact indicator of journal normalization, JNI, is constructed by using topic model to refine the sources of impact and standardization to eliminate the internal differences of journals, this index rep⁃ resents an effective, comprehensive, and precise journal impact indicator with practical applicability.