US 12,443,659 B2
Session recommendation method, device and electronic equipment
Tianjian He, Beijing (CN); Yi Liu, Beijing (CN); Daxiang Dong, Beijing (CN); Yanjun Ma, Beijing (CN); and Dianhai Yu, Beijing (CN)
Assigned to Beijing Baidu Netcom Science Technology Co., Ltd., Beijing (CN)
Appl. No. 17/279,377
Filed by Beijing Baidu Netcom Science and Technology Co., Ltd., Beijing (CN)
PCT Filed Jun. 9, 2020, PCT No. PCT/CN2020/095120
§ 371(c)(1), (2) Date Mar. 24, 2021,
PCT Pub. No. WO2021/114590, PCT Pub. Date Jun. 17, 2021.
Claims priority of application No. 201911252600.6 (CN), filed on Dec. 9, 2019.
Prior Publication US 2022/0114218 A1, Apr. 14, 2022
Int. Cl. G06F 16/901 (2019.01); G06F 9/30 (2018.01); G06N 3/08 (2023.01)
CPC G06F 16/9024 (2019.01) [G06F 9/30036 (2013.01); G06F 16/9017 (2019.01); G06N 3/08 (2013.01)] 7 Claims
OG exemplary drawing
 
1. A session recommendation method, comprising:
acquiring, by an electronic device, a session control sequence having items, and acquiring a first embedding vector matrix based on an embedding vector of each of the items in the session control sequence; wherein the session control sequence is an item sequence abstracted from n items being clicked by a user in a first order on a website;
generating, by the electronic device, a position information sequence based on an arrangement sequence of the items in the session control sequence, and acquiring a second embedding vector matrix based on an embedding vector of each piece of position information in the position information sequence; wherein the position information sequence comprises n pieces of position information, the n pieces of position information correspond to the n items respectively, and values of the n pieces of position information gradually decrease in the first order;
determining, by the electronic device, a target embedding vector matrix based on the first embedding vector matrix and the second embedding vector matrix; and
determining, by the electronic device, a recommended item, based on the target embedding vector matrix and through a Session-based Recommendation Graph Neural Network (SR-GNN);
wherein determining by the electronic device the recommended item based on the target embedding vector matrix and through the SR-GNN comprises:
splicing, by the electronic device, the first embedding vector matrix and the second embedding vector matrix to obtain the target embedding vector matrix, wherein a dimension of the embedding vector corresponding to each item in the target embedding vector matrix is twice a dimension of the embedding vector corresponding to each item in the first embedding vector matrix;
wherein the determining by the electronic device the recommended item based on the target embedding vector matrix and through the SR-GNN comprises:
inputting, by the electronic device, the target embedding vector matrix into the SR-GNN to process the target embedding vector matrix, to obtain a triggering probability of each item in an item set related to the session control sequence, wherein the item set comprises the n items and other items than the n items on the website;
determining, by the electronic device, an item recommendation list based on a value of the triggering probability of each item in the item set.