Text Classification for Private Procurement: a Survey and an Analysis of Future Trends

The development of techniques for text classification, categorization, clustering and segregation has a long history of applications in a variety of fields, including social network analysis, document archiving, business document processing. The field of private procurement, in which many application domains are included, is a small but very challenging area for the aforementioned concepts. After commerce globalization (in the nineties), e-commerce B2C explosion (in the years two-thousands) and the emergence of B2B international processes for e-procurement (in the years two-thousands-ten) we are now in a post-COVID era in which the internationalisation process has reached momentum. We are in a position of considering a front made up of multi-lingual, development differential and transparent market, for which comparison processes are ubiquitously required. In this survey we found major trends in the future of text classification employed in a multilingual, multicultural and non-standardized procurement processes. The usage of Large Language Models, and in particular the development of a specific field of post-processing of answers from LLM that is the dual component of prompt engineering, an emerging field in LLM, shall settle a new environment for procurement. We envision an application domain made of the dual usage of prompt engineering and post-processing algorithms to improve the performances of classification technologies for e-procurement. Moreover, the development of translation abilities of LLM as well as other approaches of machine translation will bring novel quality levels for these applications.