Abstract: 【Purpose/significance】Expecting to improve the accuracy and scientific, both papers and patent data sources are compre?
hensively used in this study to identify emerging technologies and construct evolution paths.【Method/process】Aiming at the specific
area of integrated circuit, this paper combines semantic similarity based on Word2Vec Model and string similarity to build the bag of
words. And LDA topic model is applied to extract the implicit technical themes in this field, after which the paper establishes multidi?
mensional index to compare and identify the emerging technologies topics, including novelty, strength and popularity. Then the paper
divides the time window uses dot product cosine similarity to compute the similarity between topics in adjacent windows to detect the
birth, succession, merge, split and extinction of topic evolution relationship, which are presented visually.【Result/conclusion】Accord?
ing to this study, within the field of integrated circuits, the scientific research and the patent research have formed mutual promotion.
By combining the papers and patent data sources, technology topics could be detected as popular, growing, mature and potential types and their evolution paths could be constructed in a more effective way, which reveals the interaction and transfer of knowledge between science and technology.【Innovation/limitation】In this study, the combination of Word2Vec semantic similarity and string similarity was innovatively used to construct word bags, which improved the quality of word bags and laid the foundation for the subsequent iden? tification of implicit technical topics based on LDA topic model. However, this study still has limitations in the diversity of data sources, time delay, and objectivity of model threshold setting, which need to be further strengthened.