METHOD, SYSTEM, AND NON-TRANSITORY COMPUTER-READABLE RECORDING MEDIUM FOR RECOGNIZING ATTRIBUTE OF PRODUCT BY USING MULTI TASK LEARNING

• Main Authors: JANG, Yun Hun, PARK, Jun Cheol, JUN, Jae Young, CHOI, Hyung Won
• Format: Patent
• Language: English; French; Korean
• Published: 03.02.2022
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; DATA PROCESSING SYSTEMS OR METHODS, SPECIALLY ADAPTED FORADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL, SUPERVISORYOR FORECASTING PURPOSES; HANDLING RECORD CARRIERS; PHYSICS; PRESENTATION OF DATA; RECOGNITION OF DATA; RECORD CARRIERS; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE,COMMERCIAL, FINANCIAL, MANAGERIAL, SUPERVISORY OR FORECASTINGPURPOSES, NOT OTHERWISE PROVIDED FOR

According to one aspect of the present invention, provided is a method for recognizing an attribute of a product by using Multi Task Learning (MTL), the method comprising the steps of: determining, with reference to correlations between tasks, two or more tasks as a task group of tasks that a product attribute recognition model is to learn together; setting parameters or preprocessing to be commonly applied to the teaching of the task group; and teaching the task group to the product attribute recognition model by applying the set parameters or preprocessing. Selon un aspect de la présente invention, l'invention concerne un procédé de reconnaissance d'un attribut d'un produit à l'aide d'un apprentissage multitâche (MTL), le procédé comprenant les étapes consistant : à déterminer, en se référant à des corrélations entre des tâches, au moins deux tâches en tant que groupe de tâches de tâches qu'un modèle de reconnaissance d'attribut de produit doit apprendre ensemble ; à définir des paramètres ou un prétraitement devant être couramment appliqué(s) à l'enseignement du groupe de tâches ; et à enseigner le groupe de tâches au modèle de reconnaissance d'attribut de produit par l'application des paramètres définis ou du prétraitement. 본 발명의 일 태양에 따르면, 멀티 태스크 러닝(Multi Task Learning; MTL)을 이용하여 상품의 속성을 인식하는 방법으로서, 태스크 사이의 상관 관계를 참조하여 둘 이상의 태스크를 상품 속성 인식 모델이 함께 학습할 태스크 그룹(task group)으로서 결정하는 단계, 상기 태스크 그룹에 대한 학습에 공통적으로 적용될 파라미터(parameter) 또는 전처리 프로세스(preprocessing)를 설정하는 단계, 및 상기 설정되는 파라미터 또는 전처리 프로세스를 적용하여 상기 태스크 그룹에 대하여 상기 상품 속성 인식 모델을 학습시키는 단계를 포함하는 방법이 제공된다.