Qing Da (Q. Da)

 / linkedin / Google scholar / DBLP


I received my B.Sc. degree and M.Sc. degree from Nanjing University, China in 2010 and 2013 respectively. Since Sep. 2013, I become a Ph.D. student in the LAMDA Group, Nanjing University, under the supervision of Prof. Zhi-Hua Zhou.

In Jan. 2015 (my 2nd Ph.D. year), I quit the Ph.D. program and joined the search algorithm team at Alibaba. Now I am a Senior Staff Algorithm Engineer of Alibaba Group.


Research Interests


My research interests include machine learning, data mining and their applications in search, recommendations and advertisement.

Career


Books


Environments


1. VirtualTaobao[AAAI’19], Github.

This project provides VirtualTaobao simulators trained from the real-data of Taobao, one of the largest online retail platforms. In Taobao, when a customer entered some query, the recommondation system returns a list of items according to the query and the customer profile. The system is expected to return a good list such that customers will have high chances of clicking the items.

Using VirtualTaobao simulator, one can access a “live” environment just like the real Taobao environment. Virtual                              customers will be generated once at a time, the virtual customer starts a query, and the recommendation system needs to                 return a list of items. The virtual customer will decide if it would like to click the items in the list, similar to a real customer.

How VirtualTaobao was trained is described in:

 

2. SESim, Github.

SESim is an E-Commerce search engine simulation platform for model examinations, which was a missing piece to connect evaluations of LTR researches and business objectives of real-world applications. SESim can examine models in the simulation E-commerce environment with dynamic responses, and its framework can be easily extended to other scenarios that items and users have differ- ent features. We hope to see the development of a dynamic dataset that facilitates industrial LTR researches in the future.

A typical process of industrial search engines contains three stages to produce a display list from a user query. A search engine first retrieves related items with intend of the user (i.e. the user query), then the ranker ranks these items by a fine-tuned deep LTR model, finally, the re-ranker rearranges the order of items to achieve some businesses goals such as diversity and advertising. Our proposed simulation platform SESim contains these three stages. We replace queries with category indices in our work, therefore SESim can retrieve items from a desensitized items database by the category index. After that, a customizable ranker and a customizable re-ranker produce the final item list. SESim allows us to study joint learning of multiple models, we left it as future work and focus on the correct evaluation for a single model.

Besides the set of real items, two important modules make SESim vividly reflect the behaviors of real users. Virtual user module aims at generating embeddings of virtual users and their query, and it follows the paradigm of Wasserstein Generative Adversarial Network with Gradient Penalty (WGAN-GP). Feedback module inputs the dis- play list and the information of the user, then outputs the feedback of users on the display list. To model the decision process of users, we train the feedback module by Generative Adversarial Imitation Learning (GAIL). For diversifying behaviors, we consider clicking and purchasing, which are two of the most important feedback of users in E-commerce.

How SESim was trained is described in:

Publications


Teaching Assistant


Awards & Honors