A recent study published in the Journal of Wine Economics reveals that the methodology used to assess technological similarity in the wine industry can dramatically influence our understanding of innovation trends globally.
Conducted by Francesca Chiaradia from the University of St Andrews, the research analyzed 9,439 wine patent abstracts from 53 countries spanning from 1970 to 2023. By employing advanced artificial intelligence tools that convert textual data into numerical vectors, the study compared five distinct language models to evaluate the evolution of wine inventions over time.
The findings indicate that while wine technology has indeed evolved, the interpretation of these changes can vary significantly depending on the model employed. This is crucial, as understanding technological similarity helps track the diffusion of ideas, identify clusters of production, and assess the emergence of new techniques.
Chiaradia’s dataset was sourced from Lens.org, a public patent database, focusing on abstracts rather than claims for a clearer depiction of inventions. The study utilized five models: GTE, S-BERT, USE, PaECTER, and Doc2Vec, revealing that different models yield varying interpretations of patent data.
For instance, GTE, S-BERT, and USE indicated a decline in average similarity from the 1970s to the late 1990s, followed by a plateau, while PaECTER and Doc2Vec suggested an increase in convergence. These discrepancies are significant enough to influence broader claims regarding whether wine innovation is trending towards standardization or diversity.
Ultimately, the study underscores that the conclusions drawn about wine innovation are contingent on the analytical model utilized, emphasizing the need for a nuanced approach in evaluating technological advancements in the wine sector.
Source: Vinetur
