Abstract: By studying the highly cited papers by years in library and information science field, the paper points out the shortcomings of high frequency keyword co-word analysis, and proposes a word selection program that acknowledges medium and low frequency keywords. It recommends deleting the high-frequency keywords and including medium and low frequency keywords that boast strong cooccurrence relationship into analysis, as well as improving representativeness of keywords. Through the comparison and analysis of multidimensional scale graph and clustering tree diagram, the paper verifies that the new method is better than the traditional high-frequency keyword co-word analysis, with higher degree of aggregation, clearer relationship between groups, and increased efficiency to reveal the thematic structure of the research field

Key words: Co-word analysis, High frequency keywords, Medium and low frequency keywords, The science of library and informatics, Highly cited papers