本文整理了KDD21的Accepted Papers^[1]中，工業(yè)界在搜索、推薦、廣告、nlp上的文章。整理的論文列表比較偏個(gè)人口味，選取的方式是根據(jù)論文作者列表上看是否是公司主導(dǎo)的，但判斷比較偏主觀，存在漏掉的可能。整理的方式主要按照公司和方向來(lái)劃分，排名不計(jì)先后順序。

1. 按照方向分類

主要挑選了一些筆者比較感興趣的方向，并整理了對(duì)應(yīng)的文章名稱。讀者可以大致讀一下文章名，判斷是否和自己的研究方向或工作方向一致，從中選擇感興趣的文章進(jìn)行精讀。

1.1 推薦系統(tǒng)

1.1.1 樣本

涉及到采樣、負(fù)樣本等。

• Google: Bootstrapping for Batch Active Sampling

• Google: Bootstrapping Recommendations at Chrome Web Store

• Alibaba：Real Negatives Matter: Continuous Training with Real Negatives for Delayed Feedback Modeling

1.1.2 表征學(xué)習(xí)

• Google: Learning to Embed Categorical Features without Embedding Tables for Recommendation

• 華為：An Embedding Learning Framework for Numerical Features in CTR Prediction

• 騰訊：Learning Reliable User Representations from Volatile and Sparse Data to Accurately Predict Customer Lifetime Value

• 阿里：Representation Learning for Predicting Customer Orders

1.1.3 跨域推薦

• 阿里：Debiasing Learning based Cross-domain Recommendation

• 騰訊：Adversarial Feature Translation for Multi-domain Recommendation

1.1.4 糾偏

• 阿里：Contrastive Learning for Debiased Candidate Generation in Large-Scale Recommender Systems

• 阿里：Debiasing Learning based Cross-domain Recommendation

1.1.5 圖神經(jīng)網(wǎng)絡(luò)

• 華為：Dual Graph enhanced Embedding Neural Network for CTR Prediction

• 美團(tuán)：Signed Graph Neural Network with Latent Groups

• 阿里：DMBGN: Deep Multi-Behavior Graph Networks for Voucher Redemption Rate Prediction

• 百度：MugRep: A Multi-Task Hierarchical Graph Representation Learning Framework for Real Estate Appraisal

1.1.6 多任務(wù)學(xué)習(xí)

• Google：Understanding and Improving Fairness-Accuracy Trade-offs in Multi-Task Learning

• 美團(tuán)：Modeling the Sequential Dependence among Audience Multi-step Conversions with Multi-task Learning for Customer Acquisition

• 百度：MugRep: A Multi-Task Hierarchical Graph Representation Learning Framework for Real Estate Appraisal

1.1.7 多模態(tài)/短視頻推薦

• 阿里：SEMI: A Sequential Multi-Modal Information Transfer Network for E-Commerce Micro-Video Recommendations

1.1.8 知識(shí)圖譜

• Microsoft：Reinforced Anchor Knowledge Graph Generation for News Recommendation Reasoning

1.1.9 推薦系統(tǒng)架構(gòu)

• Facebook：Training Recommender Systems at Scale: Communication-Efficient Model and Data Parallelism

• Facebook：Hierarchical Training: Scaling Deep Recommendation Models on Large CPU Clusters

• 阿里，F(xiàn)leetRec: Large-Scale Recommendation Inference on Hybrid GPU-FPGA Clusters

• 騰訊，Large-Scale Network Embedding in Apache Spark

• Microsoft，On Post-Selection Inference in A/B Testing

1.2 搜索

1.2.1 向量檢索

• 阿里：Embedding-based Product Retrieval in Taobao Search

1.2.2 查詢/內(nèi)容理解

• Facebook：Que2Search: Fast and Accurate Query and Document Understanding for Search at Facebook

1.2.3 概念圖譜

• 阿里巴巴：AliCG: Fine-grained and Evolvable Conceptual Graph Construction for Semantic Search at Alibaba

• 阿里巴巴：AliCoCo2: Commonsense Knowledge Extraction, Representation and Application in E-commerce

1.2.4 預(yù)訓(xùn)練

• 百度：Pretrained Language Models for Web-scale Retrieval in Baidu Search

• 微軟：Domain-Specific Pretraining for Vertical Search: Case Study on Biomedical Literature

1.2.5 Query改寫/自動(dòng)補(bǔ)全

• 微軟：Diversity driven Query Rewriting in Search Advertising

• 百度：Meta-Learned Spatial-Temporal POI Auto-Completion for the Search Engine at Baidu Maps

1.2.6 圖神經(jīng)網(wǎng)絡(luò)

• 百度：HGAMN: Heterogeneous Graph Attention Matching Network for Multilingual POI Retrieval at Baidu Maps

1.2.7 多模態(tài)

• Google: Mondegreen: A Post-Processing Solution to Speech Recognition Error Correction for Voice Search Queries

• Facebook：VisRel: Media Search at Scale

1.2.8 邊緣計(jì)算

• 阿里：FIVES: Feature Interaction Via Edge Search for Large-Scale Tabular Data

1.2.9 搜索引擎架構(gòu)

• 百度：Norm Adjusted Proximity Graph for Fast Inner Product Retrieval

• 百度：JIZHI: A Fast and Cost-Effective Model-As-A-Service System for Web-Scale Online Inference at Baidu

1.3 廣告

這一塊文章不是很多，就不細(xì)分了。

• Google: Clustering for Private Interest-based Advertising

• 阿里：A Unified Solution to Constrained Bidding in Online Display Advertising

• 阿里：Exploration in Online Advertising Systems with Deep Uncertainty-Aware Learning

• 阿里：Neural Auction: End-to-End Learning of Auction Mechanisms for E-Commerce Advertising

• 阿里：We Know What You Want: An Advertising Strategy Recommender System for Online Advertising

1.4 NLP

1.4.1 預(yù)訓(xùn)練

• 微軟：NAS-BERT: Task-Agnostic and Adaptive-Size BERT Compression with Neural Architecture Search

• 阿里：M6: Multi-Modality-to-Multi-Modality Multitask Mega-transformer for Unified Pretraining

• 微軟：TUTA: Tree-based Transformers for Generally Structured Table Pre-training

1.4.2 命名實(shí)體識(shí)別

• 微軟：Reinforced Iterative Knowledge Distillation for Cross-Lingual Named Entity Recognition

1.4.3 少樣本學(xué)習(xí)

• 微軟：Generalized Zero-Shot Extreme Multi-label Learning

• 微軟：Zero-shot Multi-lingual Interrogative Question Generation for "People Also Ask" at Bing

1.4.4 摘要

• 微軟：Reinforcing Pretrained Models for Generating Attractive Text Advertisements

1.4.5 意圖識(shí)別

• 阿里：MeLL: Large-scale Extensible User Intent Classification for Dialogue Systems with Meta Lifelong Learning

1.4.6 多模態(tài)

• 阿里：M6: Multi-Modality-to-Multi-Modality Multitask Mega-transformer for Unified Pretraining

2.按照公司分類

2.1 Google

• Learning to Embed Categorical Features without Embedding Tables for Recommendation

• NewsEmbed: Modeling News through Pre-trained Document Representations

• Understanding and Improving Fairness-Accuracy Trade-offs in Multi-Task Learning

• Bootstrapping for Batch Active Sampling

• Bootstrapping Recommendations at Chrome Web Store

• Clustering for Private Interest-based Advertising

• Dynamic Language Models for Continuously Evolving Content

• Mondegreen: A Post-Processing Solution to Speech Recognition Error Correction for Voice Search Queries

• On Training Sample Memorization: Lessons from Benchmarking Generative Modeling with a Large-scale Competition

2.2 Facebook

• Training Recommender Systems at Scale: Communication-Efficient Model and Data Parallelism
• Preference Amplification in Recommender Systems
• Hierarchical Training: Scaling Deep Recommendation Models on Large CPU Clusters
• Network Experimentation at Scale
• Que2Search: Fast and Accurate Query and Document Understanding for Search at Facebook
• VisRel: Media Search at Scale
• Balancing Consistency and Disparity in Network Alignment

2.3 Microsoft

• Generalized Zero-Shot Extreme Multi-label Learning

• Learning Multiple Stock Trading Patterns with Temporal Routing Adaptor and Optimal Transport

• NAS-BERT: Task-Agnostic and Adaptive-Size BERT Compression with Neural Architecture Search

• Reinforced Anchor Knowledge Graph Generation for News Recommendation Reasoning

• Table2Charts: Recommending Charts by Learning Shared Table Representations

• TabularNet: A Neural Network Architecture for Understanding Semantic Structures of Tabular Data

• TUTA: Tree-based Transformers for Generally Structured Table Pre-training

• Contextual Bandit Applications in a Customer Support Bot

• Diversity driven Query Rewriting in Search Advertising

• Domain-Specific Pretraining for Vertical Search: Case Study on Biomedical Literature

• On Post-Selection Inference in A/B Testing

• Reinforced Iterative Knowledge Distillation for Cross-Lingual Named Entity Recognition

• Reinforcing Pretrained Models for Generating Attractive Text Advertisements

• Zero-shot Multi-lingual Interrogative Question Generation for "People Also Ask" at Bing

2.4 阿里

• A Unified Solution to Constrained Bidding in Online Display Advertising
• AliCG: Fine-grained and Evolvable Conceptual Graph Construction for Semantic Search at Alibaba
• AliCoCo2: Commonsense Knowledge Extraction, Representation and Application in E-commerce
• Contrastive Learning for Debiased Candidate Generation in Large-Scale Recommender Systems
• Debiasing Learning based Cross-domain Recommendation
• Device-Cloud Collaborative Learning for Recommendation
• Deep Inclusion Relation-aware Network for User Response Prediction at Fliggy
• DMBGN: Deep Multi-Behavior Graph Networks for Voucher Redemption Rate Prediction
• Dual Attentive Sequential Learning for Cross-Domain Click-Through Rate Prediction
• Embedding-based Product Retrieval in Taobao Search
• Exploration in Online Advertising Systems with Deep Uncertainty-Aware Learning
• FIVES: Feature Interaction Via Edge Search for Large-Scale Tabular Data
• FleetRec: Large-Scale Recommendation Inference on Hybrid GPU-FPGA Clusters
• Intention-aware Heterogeneous Graph Attention Networks for Fraud Transactions Detection
• Live-Streaming Fraud Detection: A Heterogeneous Graph Neural Network Approach
• M6: Multi-Modality-to-Multi-Modality Multitask Mega-transformer for Unified Pretraining
• Markdowns in E-Commerce Fresh Retail: A Counterfactual Prediction and Multi-Period Optimization Approach
• MeLL: Large-scale Extensible User Intent Classification for Dialogue Systems with Meta Lifelong Learning
• Multi-Agent Cooperative Bidding Games for Multi-Objective Optimization in e-Commercial Sponsored Search
• Neural Auction: End-to-End Learning of Auction Mechanisms for E-Commerce Advertising
• Real Negatives Matter: Continuous Training with Real Negatives for Delayed Feedback Modeling
• Representation Learning for Predicting Customer Orders
• SEMI: A Sequential Multi-Modal Information Transfer Network for E-Commerce Micro-Video Recommendations
• We Know What You Want: An Advertising Strategy Recommender System for Online Advertising

2.5 百度

• Norm Adjusted Proximity Graph for Fast Inner Product Retrieval
• Curriculum Meta-Learning for Next POI Recommendation
• Pretrained Language Models for Web-scale Retrieval in Baidu Search
• HGAMN: Heterogeneous Graph Attention Matching Network for Multilingual POI Retrieval at Baidu Maps
• JIZHI: A Fast and Cost-Effective Model-As-A-Service System for Web-Scale Online Inference at Baidu
• Meta-Learned Spatial-Temporal POI Auto-Completion for the Search Engine at Baidu Maps
• MugRep: A Multi-Task Hierarchical Graph Representation Learning Framework for Real Estate Appraisal
• SSML: Self-Supervised Meta-Learner for En Route Travel Time Estimation at Baidu Maps
• Talent Demand Forecasting with Attentive Neural Sequential Model

2.6 騰訊

• Why Attentions May Not Be Interpretable?
• Adversarial Feature Translation for Multi-domain Recommendation
• Large-Scale Network Embedding in Apache Spark
• Learn to Expand Audience via Meta Hybrid Experts and Critics
• Learning Reliable User Representations from Volatile and Sparse Data to Accurately Predict Customer Lifetime Value

2.7 美團(tuán)

• Modeling the Sequential Dependence among Audience Multi-step Conversions with Multi-task Learning for Customer Acquisition
• User Consumption Intention Prediction in Meituan
• Signed Graph Neural Network with Latent Groups
• A Deep Learning Method for Route and Time Prediction in Food Delivery Service

2.8 華為

• An Embedding Learning Framework for Numerical Features in CTR Prediction
• Dual Graph enhanced Embedding Neural Network for CTR Prediction
• Discrete-time Temporal Network Embedding via Implicit Hierarchical Learning
• Retrieval & Interaction Machine for Tabular Data Prediction
• A Multi-Graph Attributed Reinforcement Learning Based Optimization Algorithm for Large-scale Hybrid Flow Shop Scheduling Problem

結(jié)語(yǔ)

后續(xù)筆者會(huì)針對(duì)感興趣的文章進(jìn)行解讀。如果大家有感興趣的文章，也歡迎在公眾號(hào)后臺(tái)跟我留言，我會(huì)優(yōu)先挑選大家感興趣的文章進(jìn)行解讀。當(dāng)然，如果你有解讀好的筆記，也歡迎投稿或交流~~

參考

[1] KDD2021 Accepted Papers: https://kdd.org/kdd2021/accepted-papers/index

[2] KDD2021 | 推薦系統(tǒng)論文集錦