BEIJING, Dec. 7, 2021 /PRNewswire/ -- JD.com announced that two papers on automatic and intelligent content generation for e-commerce product recommendation submitted by JD Retail were accepted by the 36th Association for the Advancement of Artificial Intelligence 2022 (AAAI 2022) under the track of "Highly Innovative Applications of AI." To be held in February in Vancouver, Canada, AAAI is one of the world's most prestigious conferences in the field of AI. JD is the only company to enter two papers among the total of eight under this track for AAAI 2022.
Jointly researched by JD Retail and Singapore's Nanyang Technological University, the two papers elaborate on innovative projects applied at JD to advance the automated recommendation and selling process, through Natural Language Processing (NLP), largely improving operations efficiency and saving labor costs.
JD's paper "Intelligent Online Selling Point Extraction for E-Commerce Recommendation" introduces an intelligent online selling point extraction system (IOSPE), which uses NLP and in-depth generation technology to automatically generate selling points for recommended products. As of September 2021, IOSPE has become the core service used by JD.com for more than 60 key categories of products, covering more than one million products. So far, the system has generated hundreds of millions of selling points for products, helping merchants greatly improve the efficiency of promotion and reduce costs. IOSPE effectively increases the click-through rate (CTR) of products and the average stay time of customers on the product page, which is of great benefit to increasing the conversion rate of e-commerce.
The paper "Automatic Product Copywriting for E-Commerce" explains how JD's automated product introduction copywriting generation system (APCG) works based on natural language generation (NLG) and model pretraining technologies. Since the application of the system on JD in February 2021, APCG gradually covered all product categories selling on JD, and also for merchants' new products. By September 2021, it has generated 2.53 million pieces of product descriptions, and improved the overall average click-through rate (CTR), the conversion rate (CVR) and the accumulated gross merchandise volume(GMV) by 4.22%, 3.61% and 213.42% respectively, on a year-on-year basis. The system generates up to 209,000 pieces of descriptions of products weekly, equivalent to the workload of about 1,000 professional copywriters.
Thanks to more and more application of innovative technologies such as NLP and image processing in e-commerce search and recommendation scenarios, customers worldwide are enjoying a more and more convenient shopping experience online.
Founded in 1979, the AAAI is an international nonprofit scientific society devoted to promoting research in, and the responsible use of artificial intelligence. Last year, AAAI 2021 saw over 9,000 paper submissions, of which 1,692 were accepted.