Achieving 90% In Data-Centric Industry Deep Learning Task

In industry deep learning application, our manually labeled data has a certain number of noisy data. To solve this problem and achieve more than 90 score in dev dataset, we present a simple method to find the noisy data and re-label the noisy data by human, given the model predictions as references in human labeling. In this paper, we illustrate our idea for a broad set of deep learning tasks, includes classification, sequence tagging, object detection, sequence generation, click-through rate prediction. The experimental results and human evaluation results verify our idea.

Session interest model for CTR prediction based on self-attention mechanism

Click-through rate prediction, which aims to predict the probability of the user clicking on an item, is critical to online advertising. How to capture the user evolving interests from the user behavior sequence is an important issue in CTR prediction. However, most existing models ignore the factor that the sequence is composed of sessions, and user behavior can be divided into different sessions according to the occurring time. The user behaviors are highly correlated in each session and are not relevant across sessions. We propose an effective model for CTR prediction, named Session Interest Model via Self-Attention (SISA). First, we divide the user sequential behavior into session layer. A self-attention mechanism with bias coding is used to model each session. Since different session interest may be related to each other or follow a sequential pattern, next, we utilize gated recurrent unit (GRU) to capture the interaction and evolution of user different historical session interests in session interest extractor module. Then, we use the local activation and GRU to aggregate their target ad to form the final representation of the behavior sequence in session interest interacting module. Experimental results show that the SISA model performs better than other models.

Rank-sensitive proportional aggregations in dynamic recommendation scenarios

In this paper, we focus on the problem of rank-sensitive proportionality preservation when aggregating outputs of multiple recommender systems in dynamic recommendation scenarios. We believe that individual recommenders may provide complementary views on the user’s preferences or needs, and therefore, their proportional (i.e. unbiased) aggregation may be beneficial for the long-term user satisfaction. We propose an aggregation framework (FuzzDA) based on a modified D’Hondt’s algorithm (DA) for proportional mandates allocation. Specifically, we adjusted DA to register fuzzy membership of items and modified the selection procedure to balance both relevance and proportionality criteria. Furthermore, we propose several iterative votes assignment strategies and negative implicit feedback incorporation strategies to make FuzzDA framework applicable in dynamic recommendation scenarios. Overall, the framework should provide benefits w.r.t. long-term novelty of recommendations, diversity of recommended items as well as overall relevance. We evaluated FuzzDA framework thoroughly both in offline simulations and in online A/B testing. Framework variants outperformed baselines w.r.t. click-through rate (CTR) in most of the evaluated scenarios. Some variants of FuzzDA also provided the best or close-to-best iterative novelty (while maintaining very high CTR). While the impact of the framework variants on user-wise diversity was not so extensive, the trade-off between CTR and diversity seems reasonable.

Advertising Popularity Feature Collaborative Recommendation Algorithm Based on Attention-LSTM Model

To accurately predict the click-through rate (CTR) and use it for ad recommendation, we propose a deep attention AD popularity prediction model (DAFCT) based on label recommendation technology and collaborative filtering method, which integrates content features and temporal information. First, we construct an Attention-LSTM model to capture the popularity trends and exploit the temporal information based on users’ feedback; finally, we use the concatenate method to fuse temporal information and content features and design a Deep Attention Popularity Prediction (DAVPP) algorithm to solve DAFCT. We experimentally adjust the weighted composite similarity metric parameters of Query pages and verify the scalability of the algorithm. Experimental results on the KDDCUP2012 dataset show that this model collaborative filtering and recommendation algorithm has better scalability and better recommendation quality. Compared with the Attention-LSTM model and the NFM model, the F1 score of DAFCT is improved by 9.80 and 3.07 percentage points, respectively.

Achieving 90% In Data-Centric Industry Deep Learning Task

In industry deep learning application, our manually labeled data has a certain number of noisy data. To solve this problem and achieve more than 90 score in dev dataset, we present a simple method to find the noisy data and re-label the noisy data by human, given the model predictions as references in human labeling. In this paper, we illustrate our idea for a broad set of deep learning tasks, includes classification, sequence tagging, object detection, sequence generation, click-through rate prediction. The experimental results and human evaluation results verify our idea.

Achieving 90% In Data-Centric Industry Deep Learning Task

In industry deep learning application, our manually labeled data has a certain number of noisy data. To solve this problem and achieve more than 90 score in dev dataset, we present a simple method to find the noisy data and re-label the noisy data by human, given the model predictions as references in human labeling. In this paper, we illustrate our idea for a broad set of deep learning tasks, includes classification, sequence tagging, object detection, sequence generation, click-through rate prediction. The experimental results and human evaluation results verify our idea.

Incentivizing Commuters to Carpool: A Large Field Experiment with Waze

Problem definition: Traffic congestion is a serious global issue. A potential solution, which requires zero investment in infrastructure, is to convince solo car users to carpool. Academic/practical relevance: In this paper, we leverage the Waze Carpool service and run the largest ever digital field experiment to nudge commuters to carpool. Methodology: Our field experiment involves more than half a million users across four U.S. states between June 10 and July 3, 2019. We identify users who can save a significant commute time by carpooling through the use of a high-occupancy vehicle (HOV) lane, users who can still use an HOV lane but have a low time saving, and users who do not have access to an HOV lane on their commute. We send them in-app notifications with different framings: mentioning the HOV lane, highlighting the time saving, emphasizing the monetary welcome bonus (for users who do not have access to an HOV lane), and a generic carpool invitation. Results: We find a strong relationship between the affinity to carpool and the potential time saving through an HOV lane. Managerial implications: Specifically, we estimate that mentioning the HOV lane increases the click-through rate (i.e., proportion of users who clicked on the button inviting them to try the carpool service) and the onboarding rate (i.e., proportion of users who signed up and created an account with the carpool service) by 133%–185% and 64%–141%, respectively, relative to a generic invitation. We conclude by discussing the implications of our findings for carpool platforms and public policy.