SIGIR LiveRAG Challenge Report
This year’s SIGIR conference featured the “LiveRAG Challenge”. In this competition participants received 500 questions synthesized from a given corpus (FineWeb10BT) and had 2 hours time to generate answers…
When Zero Search Results is the right answer (& how to measure this in Quepid)
Sometimes a query should produce zero search results – but how do we score this correctly in Quepid, the search relevance workbench?
New Quepid Features
New Quepid features include improved user interfaces, APIs and the ability to work with any search backend, plus new metrics including Jaccard similarity
Until Decompounding Do Us Part – Solving compound noun challenges with Querqy & LLMs
The challenges with compound nouns in search, how to successfully tackle them with decompounding using Query Rewriting and the role of LLMs.
CI/CD……CE! Continuous Experimentation for Search Improvement
Continuous experimentation as part of your CI/CD pipeline can drive continuous improvement of search quality
Learning to Measure AI-Powered Search with Vectara
How we worked with AI-based search and retrieval augmented generation platform Vectara to evaluate and measure their engine – and added it to Quepid!
Using GPT for Relevancy Judgements
Human ratings for relevance tuning are resource-intensive to generate – so could we use a large language model like GPT instead?
Site Search tracking with Google Analytics 4
As Google Analytics Universal Analytics (UA) approaches end-of-life, we show how to set up site search tracking using the new Google Analytics 4 (GA4), BigQuery and Google Looker Studio and how to create your site search tracking dashboards.
Applying synonyms – types, strategies, tools and a glimpse into the future
Applying synonyms – isn’t that a solved challenge by now? This post summarizes tools, types and strategies including a glimpse into the AI-powered future
Beyond Search Relevance: why we’re moving on to search result quality
What’s next after search relevance? We think there are more factors affecting search result quality – here’s how we’ll be exploring them.
Minimizing the risks of a search migration project
A search migration project can be risky – how can you identify & mitigate those risks with measurement, a great team and expert help?
Hybrid vigor: a winning way to use hybrid search
How we utilized a variant of BM25 for sparse vector scoring and combined it with cosine similarity between dense vectors to achieve a hybrid ranking algorithm with better search performance than either technique alone, validated at TREC 2021