%0 Conference Paper %1 Qin_2022 %A Qin, Yuqi %A Wang, Pengfei %A Ma, Biyu %A Zhang, Zhe %B Proceedings of the 31st ACM International Conference on Information &amp$\mathsemicolon$ Knowledge Management %D 2022 %I ACM %K CIKM2022 multi-interest recommender %P 1655–1665 %R 10.1145/3511808.3557221 %T A Multi-Interest Evolution Story %U https://doi.org/10.1145%2F3511808.3557221 %X The query-based recommendation now is becoming a basic research topic in the e-commerce scenario. Generally, given a query that a user typed, it aims to provide a set of items that the user may be interested in. In this task, the customer intention (i.e., browsing or purchase) is an important factor to configure the corresponding recommendation strategy for better shopping experiences (i.e., providing diverse items when the user prefers to browse or recommending specific items when detecting the user is willing to purchase). Though necessary, this is usually overlooked in previous works. In addition, the diversity and evolution of user interests also bring challenges to inferring user intentions correctly. In this paper, we propose a predecessor task to infer two important customer intentions, which are purchasing and browsing respectively, and we introduce a novel Psychological Intention Prediction Model (PIPM for short) to address this issue. Inspired by cognitive psychology, we first devise a multi-interest extraction module to adaptively extract interests from the user-item interaction sequence. After this, we design an interest evolution layer to model the evolution of the mined multiple interests. Finally, we aggregate all evolved multiple interests to infer users' intentions in his/her next visit. Extensive experiments are conducted on a large-scale Taobao industrial dataset. The results demonstrate that PIPM gains a significant improvement on AUC and GAUC than state-of-the-art baselines. Notably, PIPM has been deployed on the Taobao e-commerce platform and obtained over 10% improvement on PCTR.

@inproceedings{Qin_2022, abstract = {The query-based recommendation now is becoming a basic research topic in the e-commerce scenario. Generally, given a query that a user typed, it aims to provide a set of items that the user may be interested in. In this task, the customer intention (i.e., browsing or purchase) is an important factor to configure the corresponding recommendation strategy for better shopping experiences (i.e., providing diverse items when the user prefers to browse or recommending specific items when detecting the user is willing to purchase). Though necessary, this is usually overlooked in previous works. In addition, the diversity and evolution of user interests also bring challenges to inferring user intentions correctly. In this paper, we propose a predecessor task to infer two important customer intentions, which are purchasing and browsing respectively, and we introduce a novel Psychological Intention Prediction Model (PIPM for short) to address this issue. Inspired by cognitive psychology, we first devise a multi-interest extraction module to adaptively extract interests from the user-item interaction sequence. After this, we design an interest evolution layer to model the evolution of the mined multiple interests. Finally, we aggregate all evolved multiple interests to infer users' intentions in his/her next visit. Extensive experiments are conducted on a large-scale Taobao industrial dataset. The results demonstrate that PIPM gains a significant improvement on AUC and GAUC than state-of-the-art baselines. Notably, PIPM has been deployed on the Taobao e-commerce platform and obtained over 10% improvement on PCTR.}, added-at = {2022-10-17T18:24:20.000+0200}, author = {Qin, Yuqi and Wang, Pengfei and Ma, Biyu and Zhang, Zhe}, biburl = {https://www.bibsonomy.org/bibtex/298af274bac492b3534dc838fe016ff3d/brusilovsky}, booktitle = {Proceedings of the 31st {ACM} International Conference on Information {\&}amp$\mathsemicolon$ Knowledge Management}, description = {A Multi-Interest Evolution Story | Proceedings of the 31st ACM International Conference on Information & Knowledge Management}, doi = {10.1145/3511808.3557221}, interhash = {6b05a8cde74179ef109c15f1195477f6}, intrahash = {98af274bac492b3534dc838fe016ff3d}, keywords = {CIKM2022 multi-interest recommender}, month = oct, pages = {1655–1665}, publisher = {{ACM}}, timestamp = {2022-10-17T18:24:20.000+0200}, title = {A Multi-Interest Evolution Story}, url = {https://doi.org/10.1145%2F3511808.3557221}, year = 2022 }