Product Recommendation using Image and Text Processing

Production recommendation systems allow users to review other information that are relating to the product that they are interested in. The fundament of this problem in computer and technology perspective is to how extract information from the product that can be used for matching the related products. This work presents a technique that integrates information from production images and the description of the product (text format) to match a set of products collected in a databased. The matching will be used as the product recommendation system. Shape-based representation is extracted from the image. This includes HOG, Shape Context and Hu Moments. The description of the product is embedded by the information presented using LSTM technique. Integration between image and text information is performed using a simple weighting technique. The matching id carried out using cosine similarity measurement. The data is collected from online stores. The experimental results show that the proposed technique gives promising results.