Haystack Search Relevance Training - April 27th, 28th, and May 1st

Training to be held before and after the Haystack conference

New to relevance? Want to solidify your foundation? Or take things to an advanced level? OpenSource Connections will be offering training before and after the conference:

New to relevance? Want to solidify your foundation? Or take things to an advanced level? OpenSource Connections will be offering their "Think Like a Relevance Engineer" (TLRE) trainings for Solr or Elasticsearch directly before the Haystack Conference. These courses give you a foundation for solving applied Information Retrieval problems with Solr or Elasticsearch.

Training Details

From the team behind Relevant Search - Learn to improve search quality from the best experts in the field. Bust through the mystique around topics like 'cognitive search' and 'Learning to Rank'. In two days, we teach you to measure, tune, and improve search quality. You'll appreciate relevance tuning from TF*IDF, to boosts, to Learning to Rank. You'll begin your journey to search & discovery applications that seem to 'get users' and deliver business results.

'Think Like a Relevance Engineer' has helped me think differently about how I solve Solr & Elasticsearch relevance problems
Matt Corkum, Disruptive Technology Director, Elsevier
What a positive experience! We have so many new ideas to implement after attending 'Think Like a Relevance Engineer' training.
Andrew Lee, Director of Engineering Search, DHI

Timings

Trainings will start at 9am sharp and run until 5pm.

What you'll get out of it

  • How to measure search quality
  • Removing the fear from relevance experimentation
  • Becoming 'hypothesis driven' in relevance work
  • Elasticsearch relevance tuning techniques
  • Building semantic search
  • Implementing machine learning to improve relevance (ie Learning to Rank)
  • Access to the best experts in the world on Elasticsearch & Solr relevance for brainstorming your problems

Day One - Managing, Measuring, and Testing Search Relevance

This day helps the class understand how working on relevance requires different thinking than other engineering problems. We teach you to measure search quality, take a hypothesis-driven approach to search projects, and safely 'fail fast' towards ever improving business KPIs

  • What is search?
  • Holding search accountable to the business
  • Search quality feedback
  • Hypothesis-driven relevance tuning
  • User studies for search quality
  • Using analytics & clicks to understand search quality

Day Two - Engineering Relevance with Elasticsearch|Solr

This day demonstrates relevance tuning techniques that actually work. Relevance can't be achieved by just tweaking field weights: Boosting strategies, synonyms, and semantic search are discussed. The day is closed introducing machine learning for search (aka "Learning to Rank").

  • Getting a feel for Elasticsearch|Solr
  • Signal Modeling (data modeling for relevance)
  • Dealing with multiple, competing objectives in search relevance
  • Synonym strategies that actually work
  • Taxonomy-based Semantic Search
  • Introduction to Learning to Rank

You'll also receive a copy of Relevant Search written by OpenSource Connections CTO Doug Turnbull and OpenSource Connections Alum John Berryman.

Learning to Rank Training

In this two day training course offered April 27th and 28th, we go hands on with machine learning tools to improve relevance. Using an open source search engine, we see our models come to life, see the pitfalls of letting machines control relevance, and work to mitigate those pitfalls. From Hello World to Learning to Rank Lessons you can avoid learning the hard way. See how you can apply open source tooling to optimize your search results with machine learning. Ideal for production focused data scientists, relevance engineers, machine learning engineers, or search developers. Some familiarity with a search engine is expected.

Timings

Trainings will start at 9am sharp and run until 5pm.

What you'll get out of it

How to:

  • Interact with the Solr & Elasticsearch Learning to Rank problems
  • Use machine learning to optimize relevance
  • Avoid common pitfalls on Learning to Rank projects
  • Avoid 'garbage-in, garbage out' - generating great features and training data
  • Hands-on work with Learning to Rank models
  • Integrate click models and conversions to generate meaningful training data

Day One: Hands on Basics

We get hands on with movie data, and a simple judgment list. We see where things can go wrong!

  1. Search Relevance as an Machine Learning Problem
  2. Cutting your Teeth With Your First Model
  3. What's Wrong with My Judgments?
  4. Iterating on Features
  5. Choosing the Best Learning to Rank Model

Day Two: Real-World Learning to Rank

At scale, in a real search applications, here are the concerns that will rear themselves.

  1. Dealing with Presentation Bias
  2. Including Non-User Relevance Concerns (Business Rules and Marketplace concerns)
  3. Model Verification, Checks, and Balances
  4. Personalization and Recommendations with Learning to Rank
  5. Including Embeddings and Other Exotic Features
  6. The Next Frontiers of ML and Search

Who Should Come to training?

This training is appropriate for members of the search team that have an interest in optimizing search with advanced programming techniques and machine learning

  • Search Engineers
  • Data Scientists
  • Data Engineers that use the search engine
  • Machine learning engineers
  • Relevance engineers
  • Product team wanting exposure to machine learning methods

Natural Language Search - Training

Perhaps your team has done what it can with normal relevance methods? And now wants to move towards using Natural Language Processing to improve understanding of your content and queries?

'Natural Language Search’ is a 2 day training course to give your team the ability to use open source NLP tools to improve search relevance. This class helps you understand how to work with different NLP tools and libraries to recognize concepts and entities, understand customer query intent, differentiate models for your specific language use cases, avoid common pitfalls, and build a language-aware search platform.

Timings

Trainings will start at 9am sharp and run until 5pm.

What you'll get out of it

You’ll learn about using language to improve search experience for your users. This includes practical how-to’s and hands-on labs for:

  • Leverage NLP tools to pre-process text and understand the basics
  • Recognize Dates, Quantities, Locations, and other Entities and use them in Search
  • Measure how accurately a model processes your text
  • Train NLP models to fit your language use cases
  • Classify query text for customer intent
  • Learn state of the art deep learning NLP transformer technology
  • Work Hands-on with NLP Libraries to build a language-aware search platform

Day One - Search, Language, and Tools

Goal for the day: Making Sense of Search with Language Enrichment. You’ll learn the core concepts of language structure and how NLP tools and models work, and how to use them for named entity recognition and query intent!

  1. Introduction to NLP
  2. Back to English Class! + LAB
  3. Named Entities
  4. Measurement + LAB
  5. Values Search + LAB
  6. Entity Location Search + LAB
  7. Query Intent + LAB

Day Two - Natural Language Search

Goal for the day: Make a Neural Search Engine! We go deep into practical exercises of building a working natural language search engine, through the use of state of the art transformer NLP technology.

  1. Vectors in Search + LAB
  2. Introduction to Transformers and BERT + LAB
  3. Fine-tuning Neural Search + LAB
  4. Hosting Considerations & Model Performance
  5. Putting it all together - the Natural Language Search Engine + LAB

Who Should Come to training?

This training is appropriate for members of the search team that have an interest in optimizing search with advanced programming techniques and machine learning

  • Search Engineers
  • Data Scientists
  • Data Engineers that use the search engine
  • Machine learning engineers
  • Relevance engineers
  • Product team wanting exposure to machine learning methods

Search Relevance for Product Managers

Agenda

This full day class for Product Owners and Product Managers responsible for search teams helps you understand how working on search requires different thinking than other engineering problems. We teach you to measure search quality, take a hypothesis-driven approach to search projects, and safely 'fail fast' towards ever improving business KPIs.

  • What is search? How is enterprise search different from other domains like e-commerce search?
  • Holding search accountable to the business.
  • How to solicit and search quality feedback, and how to prioritize it!
  • Understand hypothesis-driven relevance tuning to drive better search
  • When to under take user studies for search quality
  • How to use analytics & clicks to understand search quality and your users needs

What You’ll Get Out Of It

  • A common vocabulary to use with your search engineers.
  • How to communicate the impact of search to the business.
  • How to manage the flood of “Search doesn’t work” input in a way that lets you prioritize the work.
  • An understanding of when “bandaid” fixes are appropriate, and when you need to invest in major changes to your algorithm.

Your Trainer: Experienced Relevance Expert!

Your class will be taught by Eric Pugh.

Eric is co-founder of OpenSource Connections, whose mission is to empower the world’s search teams. He is active in the open source search communities, both Solr and Elasticsearch. Today he helps OSC’s clients build their own search teams and improve their search maturity, both by leading projects and by acting as a trusted advisor, helping transform their practices into experimental, hypothesis-driven mindsets around search relevance.

Style of Training: Small Group Workshop

Our trainers are not ‘stock tech trainer’ from central casting mindlessly reading slides. Our trainers expect to problem solve in real-time, and we want to hear your tough problems. As OpenSource Connection’s mission is to ‘empower search teams’, we see training as the central component to our mission. Our training is ‘workshop style’ where much of the value is the interactions and knowledge sharing between the small class and the trainer.

Who This Training is For

Some basic exposure to Search is recommended, but not required. We will focus on helping you talk to both the technical folks and the business. Roles that would get value out of this training:

  • Business Analyst
  • Product Managers
  • Search product owners

If you have questions about if this class is a good fit for you, please get in touch with Eric at epugh@opensourceconnections.com