Abstract: 【Purpose/significance】Measuring paper novelty based on the semantic similarity algorithm of paper content, revealing the multidimensional measurement features presented by novelty papers.【Method/process】Measuring the novelty of papers, classifying the papers into high, medium and low novelty intervals based on the standardized novelty scores of the papers, comparing and analyz⁃ ing the characteristics of high novelty papers regarding text and related content.【Result/conclusion】In terms of textual content, the proportion of papers with short titles, short abstracts, and a small number of keywords, as well as those with question marks in the title, is higher in the high novelty interval than in the medium and low novelty intervals; The topic diversity of all papers in the field of Human-Computer Interaction shows a gradually increasing trend from 2002 to 2022, and papers ranked in the top 50 exhibits the strongest topic variety within the high novelty papers. In terms of associated content characteristics, the proportion of papers authored solely by a single individual is higher in the high novelty interval than in the medium and low novelty intervals. Additionally, within the high novelty interval, papers authored independently by a single author, institution, or country tended to have higher novelty scores than co-authored papers; The proportion of references cited by high novelty papers that published 10 years or earlier is higher than that by low novelty papers.【Innovation/limitation】Classifying papers into different novelty levels, systematically revealing the multidi⁃ mensional measurement characteristics of novelty papers in terms of textual and related content.