Assessing Semantic Value using Machine Learning

Abstract

The purpose of this study is to propose a method to achieve product integrity by using machine learning technology. In many cases, customer needs are subjective and ambiguous. Therefore, it is difficult for the design department to understand them and achieve product integrity for them in the product development process. To address this problem, we developed a machine learning model that predicts expression words of customer evaluations from datasets such as product specifications and images. By using predictive models for customer evaluations, design departments digest customer needs without intermediators like marketing departments, so that they can directly identify the gap between customer needs and product value, and quickly fix it. To demonstrate this concept, we select camera lens as a target product and built a machine learning model based on the technology of image caption generation. After calibrating parameters by training dataset, the model obtained an accuracy that is close to the results of human captioning of images reported in the prior study. Finally, the study provides some examples of applications such as a cognitive map of product values and also a product positioning map of competing products, to show practical use cases in product development processes.