Measurement of Innovation in Service Firms Using LDA Theme Model——A Comparative Analysis with Traditional Innovation Measurement Methods
Service innovation is an important connotation of high-quality development of the service industry.The tradi-tional innovation indicators represented by R&D and patent have obvious defects in measuring the innovation of service en-terprises.Using the Latent Dirichlet Allocation(LDA)thematic model,this paper measures the innovation level of 914 Chi-nese listed companies in the service sector from 2008 to 2022 and compares the results with the traditional innovation indi-cators from the dimensions of full sample,industry classification,regional distribution,firm scale and firm ownership.The re-sults show that:The LDA method can①reflect both technical and non-technical innovation of the service enterprises,and is more suitable for measuring the innovation of non-R&D and non-patent service enterprises;②Better capture the character-istics of innovation in lifestyle service enterprises such as education,culture,sports and entertainment industries,and truly reflect their innovation capabilities;③ Upward adjust the innovation level of service enterprises in the central and western regions,downward adjust it in the eastern region;④Upward adjust the innovation level of the small and medium-sized serv-ice enterprises and downward adjust it of the large service enterprises;⑤Explain the essential reasons for the higher innova-tion level of non-state-owned service enterprises through exploring the differences in terms of innovation frequency and themes between state-owned and non-state-owned service enterprises.Thus five policy recommendations are put forward as following-expanding the new cognition of the connotation of service innovation,accelerating the innovation of living service industry enterprises,creating a good innovation ecology,deepening the penetration of digital technology,and giving full play to the advantages of differentiation in services firms.
LDA theme modelService enterprisesInnovation measurementAnalyst research reportsText analysis
NETL
NSTL
万方数据
