Information technology / Business software and AI
Amazon: Search autocomplete that shows book covers
- Company
- Amazon
- Country
- United States
- Adoption stage
- In operation
- Source published
- Date basis
- The date the source was published. It can differ from the date adoption started.
- How the source was checked
- Read the full source text
The work problem
Customers searching for books on Amazon and Audible had to choose the desired title and format from search results even after selecting a text autocomplete suggestion. The system needed to infer intent from partial keystrokes while answering many requests with low latency, and distinguish broad genre searches from specific title searches.
Technology and data
Amazon maps partial inputs to books using historical searches and uses confidence filters to distinguish general queries from specific titles. Time-decay weighting of historical interactions reflects recent interests. Audible deployed DeepPLTR, a ranking model with separate processing for recent search context, keyword and item engagement history, and customer tastes and book descriptions. Amazon uses a probabilistic retrieval model to select a title, followed by another model personalizing audiobook, e-book or physical-book format. An intent model selects book, author or series widgets; customers view the covers and navigate directly to detail pages.
Results
Amazon Science said visual autocomplete on both services shows relevant content faster than the earlier approach and reduces discovery steps by bypassing the search-results page. The post says it handles millions of queries at low latency. The implementation shortens the path from typing to the product detail screen while leaving customers to choose their desired books.
Limits and open questions
Conversion gains, revenue growth and exact response times were not disclosed. The adoption month was not confirmed, so ordering uses the source publication date.
Sources
- Amazon | 책 표지를 보여주는 검색 자동완성amazon.science, Accessed
- Simplifying book discovery with ML-powered visual autocomplete suggestionsAmazon Science / Audible, Accessed
Compiled from public sources. These are not results from ATF Works customers.