Policy-Grounded Dynamic Facet Suggestions for Job Search

2026-05-15

AI Summary

This study addresses the challenge of ambiguous and underspecified queries in job search, which hinder accurate intent recognition and degrade job recommendation performance. To resolve this, the authors propose a dynamic, fine-grained attribute suggestion mechanism that uniquely integrates policy-guided exploration with retrieval-augmented ranking. The approach leverages an offline-constructed taxonomy, embedding-based retrieval, and a distilled small language model to score candidates, enabling real-time generation of personalized semantic attributes that disambiguate user intent. The system employs single-token sequential scoring, batch processing, and prefix caching to ensure low-latency inference. Offline evaluations demonstrate high suggestion precision, while online A/B tests show significant improvements in both attribute suggestion click-through rates and overall job search conversion rates.

Abstract

Job seekers often initiate search with short, underspecified queries. At LinkedIn, over 80% of job-related queries contain three or fewer keywords, making accurate user intent inference and relevant job retrieval particularly challenging. We present dynamic facet suggestion (DFS), an interactive query refinement mechanism that facilitates intent disambiguation by surfacing personalized semantic attributes conditioned on the joint user-query context in real time. We propose a policy-grounded, retrieval-augmented ranking framework for facet suggestion, comprising offline taxonomy curation, embedding-based retrieval of top-K candidates, and distilled small language model (SLM) based candidate scoring. The system is optimized for real-time serving via pointwise single-token scoring with batching and prefix caching. Offline evaluation demonstrates high precision for generated suggestions, and online A/B tests show significant improvements in suggestion engagement and job search outcomes.

Problem

• job search
• query refinement
• user intent inference
• facet suggestion
• short queries

Innovation

• dynamic facet suggestion
• retrieval-augmented ranking
• small language model
• real-time query refinement
• policy-grounded optimization