I am the driving force behind Import Tourism with a vision to broaden the company’s readership throughout 2016. I am an editor and reporter of “Technology” category.
Address: 3194 Flynn Street, Garfield Heights, OH 44125, USA
Phone: (+1) 440-829-5759
Latest posts by Raymond Young (see all)
Amazon can predict their customer’s queries context with the help of Artificial intelligence (AI) and machine learning. In recent conference of ACM SIGIR on Human Information and Retrieval, a preprint paper is presented about the schedule to take place this month. In that paper, Amazon researchers gives the description about the system that predicts some interesting activities such as ‘running’ from customer’s queries like “Nike men’s shoes.” Therefore, this technology could help the company to improve the quality of search results of the company’s website Amazon.com. In particular, it could enhance overall Amazon shopping experience for the customers. In recent blog post, Amazon Search customer experience applied scientist and contributing author Adrian Boteanu explains about products searching activities. According to them, most product discovery algorithms look for correlations between queries and products. By contrast, on the basis of context of use by the customer, AI of researchers’ identifies the best matches. Hence the Amazon team assembles total number of 173 context of use categories to train the AI system. These 173 categories further divided into 112 activities and 61 audiences. 112 activities include running, cleaning, reading and many more. Likewise, 61 audiences include man, woman, child, daughter, professional and others on the basis of common product queries by the customers. Expert used standard reference texts to create aliases for the terms they used in order to denote the categories on the basis of queries. Each model was trained to predict context of use on the basis of query strings, and in tests, the best-performing managed to anticipate product annotations with 97% accuracy for activity categories and 92% for audience categories. When human reviewers were presented with rank-ordered lists of categories generated by the activity models, the reviewers agreed an average of 81% of the time with the system’s per-item predictions.