Unblocking the Search Pipeline via Greater Natural Language Understanding For Search Engines

As search engines seek to balance efficiency versus effectiveness they are striving to open up a better full-pipeline of search, which has the potential to provide a tidal wave of ‘relevance candidates’. Other approaches to achieve this are being explored, which go well beyond BERT and friends. Passages and the research enabled through passage indexing and ranking are contributing greatly to this, enabled by machine learning. Google MUM is just on the horizon and a plethora of other rich experience, and recommender system approaches. Recent developments in AI and ML signal the start of a truly intuitive search experience and I will be exploring these in my talk at Ungagged London.

Key Takeaways

• Understand the potential opportunities just around the corner
• Be more aware of the mult-stage ranking approaches of search engines
• Understand the notions around diversity in search results and probability ranking principle trade-offs
• Be armed for the future of search