Improving Pinterest Search Relevance Using Large Language Models
To improve relevance scoring on Pinterest Search, we integrate Large Language Models (LLMs) into our search relevance model, leveraging carefully designed text representations to predict the relevance of Pins effectively. Our approach uses search queries alongside content representations that includ...
Gespeichert in:
Hauptverfasser: | , , , , , , , |
---|---|
Format: | Artikel |
Sprache: | eng |
Schlagworte: | |
Online-Zugang: | Volltext bestellen |
Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
Zusammenfassung: | To improve relevance scoring on Pinterest Search, we integrate Large Language
Models (LLMs) into our search relevance model, leveraging carefully designed
text representations to predict the relevance of Pins effectively. Our approach
uses search queries alongside content representations that include captions
extracted from a generative visual language model. These are further enriched
with link-based text data, historically high-quality engaged queries,
user-curated boards, Pin titles and Pin descriptions, creating robust models
for predicting search relevance. We use a semi-supervised learning approach to
efficiently scale up the amount of training data, expanding beyond the
expensive human labeled data available. By utilizing multilingual LLMs, our
system extends training data to include unseen languages and domains, despite
initial data and annotator expertise being confined to English. Furthermore, we
distill from the LLM-based model into real-time servable model architectures
and features. We provide comprehensive offline experimental validation for our
proposed techniques and demonstrate the gains achieved through the final
deployed system at scale. |
---|---|
DOI: | 10.48550/arxiv.2410.17152 |