Abstract: 【Purpose/significance】To ensure the integrity of the thesaurus, it is necessary to timely include new terminology that ap⁃
pears in the field but not included in the thesaurus is combined with the time of the candidate words and the word frequency character⁃
istics of the document. It is worthwhile to explore the distribution of new terms from the perspective of time series to guide the selec⁃
tion of new terms.【Method/process】This paper mainly studies the time series corresponding to the document frequency of words, and
performs time-frequency normalization and time-equalization preprocessing, and introduces k-means clustering algorithm to cluster
the words based on time series trend change. The general rules of terminology and non-terminology trends, in turn, summarize the
trend-changing characteristics that new terms should satisfy.【Result/conclusion】Through clustering research, it is concluded that new
terms are generally in a growing trend. The selection of candidate words in the growth state, after expert judgment, the method can ef⁃
fectively select new terms from the candidate vocabulary that can be added to the thesaurus, the method has higher accuracy.【Inno⁃
vation/limitation】The innovation lies in the selection of new terms in thesaurus, which takes into account both the time variation and
the word frequency of documents, which is limited to the scale of data processing. In the empirical study, only the word frequency data
of key words in the paper are counted.