Abstract: This article applies natural language sentiment analysis technology to the calculation of semantic prosody strength, with a view to exploring the feasibility and operability of a more precise measurement of semantic prosody under the theoretical framework of extended units of meaning.The existing measurement methods of semantic prosody strength are mostly based on the frequency-based polarity ratio calculation.The commonality of this type of methods is that collocation words in the same polarity category have equivalent attitudinal or emotional associations, but they cannot penetrate the inside of a polarity category and perform a more detailed and differentiated analysis of specific semantic prosody strengths under the same polarity category.We hope that the high-precision semantic prosody strength value supported by sentiment analysis technology can more finely describe the similarities and differences of semantic prosodies among different node words and for the same node word in different collocation patterns and semantic preferences, and hence help us better understand the consistent aura of meaning pervading the context around the node word, and the relationship between the elements of extended units of meaning.