In this paper, we first use the crawler technology to obtain the audience’s comment data on the documentary film “Bright Torch” in English on Chinese non-heritage culture during the period of 2022~2023 on Amazon and YouTube platforms, and then utilize the Word Frequency-Inverse Text Frequency (TF-IDF) method to conduct information retrieval and text mining on the importance of the feature words in the dataset. After that, the feature words are subjected to deeper semantic mining using LDA topic model for topic modeling of different topics of audience, and the validity of topic modeling is verified by DPMCSKM algorithm. Then the emoji clustering algorithm is established by FP growth algorithm and retrieval distance, and the emoji library is built, and the calculation of emotional tendency is completed based on the plain Bayesian algorithm. The results show that the themes of each year of the documentary are independent of each other, and there is no overlap between the themes. Taking 2023 as an example, according to its characteristics, it is named as four types of themes, namely, “Emotional Value and Cultural Identity, Circle-Breaking and Influence of the Times, Audio-Visual Aesthetics and Technological Admiration, and Inheritance of Hope and Innovative Vigor”. The emotional tendency of the dissemination of “Bright Torch” in each region is: extremely strong positive for overseas Chinese communities, strong positive for East and Southeast Asian cultural circles, moderate positive for North America and Western Europe, and neutral positive for other regions.
