Human Value Detection from Bilingual Sensory Product Reviews

Proceedings of Recent Advances in Natural Language Processing, pages 792–802 Varna, Sep 4–6, 2023 https://doi.org/10.26615/978-954-452-092-2_086 792 Abstract We applied text classification methods on a corpus of product reviews we created with the help of a questionnaire. We found that for certain values, “traditional” deep neural networks like CNN can give promising results compared to the baseline. We propose some ideas to improve the results in the future. The bilingual corpus we created which contains more than 16 000 consumer reviews associated to the human value profile of the authors can be used for different marketing purposes. Introduction In this paper, we investigate the possibility of detecting human values from consumer reviews about sensory products (perfume and other scented products such as shampoo and detergent). We carried out a series of experiments to detect human values as defined in the Schwartz’s theory (1992, 1996, 2003, 2006) in a corpus of consumer reviews about scented products that we created. These experiments are part of a research project on consumer segmentation based on psychological traits. This is a method widely used in marketing research that allows manufacturers to create products which better meet the expectations of their end users. This is particularly interesting for the fragrance 1 The difference between the two is that personality traits describe an individual, while human values describe what is important for an individual. industry, as smells have special links to emotions (Warrenburg 2002) and psychological states and profiles. There are previous works about the detection of personality traits from texts (Pennebaker et al., 2001; Mairesse et al., 2007; Majumder et al., 2017; Kazameini et al., 2020; Leonardi et al., 2020; Vásquez and Ochoa-Luna, 2021). In these works, a corpus containing texts and the personality traits auto-evaluated by the author of the texts is used – the authors were asked to answer a personality questionnaire, and the result of this questionnaire is considered as the ground truth in this task. The researchers of the previous works applied different methods to this corpus and observed the performance. As there are few existing works on the detection of human values from texts, and personality traits and human values are both psychological traits that describe the psychological profile of an individual1 , we place our work in the field of psycholinguistics and the related works from which we got inspirations are about personality detection from texts. To the best of our knowledge, this is the first work of applying NLP methods to the detection of human values in the fragrance industry. In this article, we first present the linguistic resources we used and the formalization of the task. We then describe the methods used in the experiments. After that, we present the experiments and the results we obtained. In the Human Value Detection from Bilingual Sensory Product Reviews Boyu Niu1,2, Céline Manetta2, Frédérique Segond1,3 1 Inalco, Paris, France 2 IFF, Neuilly-sur-Seine, France 3 Inria, Montpellier, France boyu.niu@{inalco.fr, iff.com}, celine.manetta@iff.com, frederique.segond@{inalco.fr, inria.fr} 793 end, we propose some ideas that may improve the …

Content truncated to 500 words