Reproduce “An End-to-End Solution for Named Entity Recognition in eCommerce Search” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Xiang Cheng, Mitchell Bowden, Bhushan Ramesh Bhange, Priyanka Goyal, Thomas Packer, Faizan Javed
A new end-to-end solution for Named Entity Recognition (NER) in eCommerce search, featuring a novel "TripleLearn" framework for efficient iterative model training, achieved an F1 score of 93.3 and demonstrated a $60 million annual incremental revenue by improving search query understanding at The Home Depot.
Key insights
- The "TripleLearn" iterative training framework effectively manages heterogeneous data sources, leading to cumulative performance improvements and expanded entity coverage without relying solely on large, expensive human-annotated datasets.
- Incorporating character-level embeddings significantly boosts NER performance, particularly for handling domain-specific, uncommon, or misspelled words in noisy eCommerce queries.
- Custom-trained, domain-specific Word2vec embeddings demonstrate superior semantic relevance and vocabulary coverage in eCommerce, outperforming general-purpose embeddings and contributing to higher F1 scores.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.