LinkedIn Post Embeddings: Industrial Scale Embedding Generation and Usage across LinkedIn

TL;DR

The paper tackles scalable generation of semantic post embeddings for LinkedIn's retrieval and ranking systems. It introduces a multi-task fine-tuning framework on a transformer-based LLM to produce $50$-dimensional post embeddings that capture cross-task semantics across languages, enabling nearline deployment. The approach demonstrates positive transfer across multiple labeling tasks, strong zero-shot generalization, and competitive performance against OpenAI embeddings with substantially smaller dimensionality, illustrating practical advantages for production systems. The embeddings are deployed in nearline infrastructure and shown to improve Feed ranking, retrieval, and video recommendations in online A/B tests, underscoring significant real-world impact. The work also describes deriving member representations via Ward clustering and provides a comprehensive offline evaluation methodology to validate embedding quality before use in downstream surfaces.

Abstract

A post embedding (representation of text in embedding space that effectively captures semantic meaning) is a foundational component of LinkedIn that is consumed by product surfaces in retrieval and ranking (e.g., ranking posts in the feed or video tab). This paper presents the post embeddings used at LinkedIn, where a pre-trained transformer-based large language model (LLM) is taken as input and fine-tuned using multi-task learning across a diverse set of semantic labeling tasks. We observe positive transfer, leading to improved performance across all tasks, compared to training them independently. The generated post embeddings outperform baseline models in zero-shot learning, demonstrating its potential for broader applicability. Furthermore, the generated post embeddings' performance surpasses that of OpenAI's ADA-001 and ADA-002 embeddings on LinkedIn specific datasets and tasks. We also describe the offline evaluation methodology and the deployment to our near-line infrastructure, which makes the post embedding available for use within minutes of post creation for any downstream application. We present how the embeddings were applied in the Feed product surface, in both ranking and retrieval stages, and showcase the real world online impact to demonstrate the superior performance of these embeddings. Finally, we also share the results of applying the embeddings to the retrieval system of our video ranking product surface in LinkedIn. These embeddings have been battle-tested in production at LinkedIn for over two years, consistently powering multiple products.

Figures (7)

• Figure 1: Architecture used for a single task
• Figure 2: Multitask architecture
• Figure 3: Task heterogeneous sampling with 3 datasets and 4 workers
• Figure 4: Online system for post embeddings
• Figure 5: Feed ranking model architecture with LinkedIn post embeddings