2018-Real-time Personalization using Embeddings for Search Ranking at Airbnb

Table of Contents

1. Problem Definition
2. Previous Methods & New Methods
3. Impacts
4. Further Improvement
5. TODO & Questions

Problem Definition

1. key metrics
2. Two application
   1. real-time personalization rec
   2. Similar listing rec

Previous Methods & New Methods

1. Key contribution
   1. real-time embedding
   2. adapt it to their marketing
      1. Congregated training
      2. more weight to booking item
   3. short & long term interest
      1. short: item embedding based on clicked listings
      2. long: user_type, item_type embedding based on booked listings

Impacts

1. similar hotel rec
   1. 21% increase in CTR
   2. 4.6% increase in bookings

Further Improvement

TODO & Questions

1. data info
   1. listing embedding:
      1. 800M click sessions
      2. 4.5M listings
2. how to eval?? -> While some listing characteristics, such as price, do not need to be learned because they can be extracted from listing meta-data, other types of listing characteristics, such as architecture, style and feel are much harder to extract in form of listing features.
   1. Clustering -> shown on map
   2. Similarity difference in difference dimensions
   3. Embedding evaluation tools
3. Video demostration of this paper
4. blogs from Wangzhe
   1. 从KDD 2018 Best Paper看Airbnb实时搜索排序中的Embedding技巧 - 王喆的文章 - 知乎
   2. Airbnb如何解决Embedding的数据稀疏问题？ - 王喆的文章 - 知乎

• Previous
  2010-Personalized Click Prediction in Sponsored Search
• Next
  2021-Hotel2vec, Learning Hotel Embeddings from User Click Sessions with Side Information