Upgrade to Pro — share decks privately, control downloads, hide ads and more …

AI Infused Search

AI Infused Search

This is the presentation I used for the Global AI Tour 2021.

Daron Yondem

June 15, 2021

Transcript

15 slides
  1. Levels of Artificial Intelligence Easy Medium Hard Use an existing

    model Customize a pre-trained model Train a new model from scratch A pre existing Service like a box around someone's face It can recognize face But we need specific faces Lot’s of time and effort needed
  2. What is Azure Cognitive Search? • A search engine for

    full text search • Rich indexing, with text analysis and optional AI enrichment for content extraction and transformation • Simple syntax, full Lucene syntax, and typeahead search • REST APIs and client libraries in Azure SDKs for .NET, Python, Java, and JavaScript • State-of-art ranking algorithms through semantic search (preview)
  3. Indexers • Crawler that extracts searchable text and metadata from

    an external Azure data source • Azure SQL, Azure Cosmos DB, Azure Table Storage and Blob Storage • In Preview: Amazon Redshift, Salesforce Objects, Snowflake
  4. AI Enrichment • An extension of indexers • Built-in skills

    from Microsoft or custom skills • Natural language processing • Entity recognition, language detection, key phrase extraction, text manipulation, sentiment detection, and PII detection. • Image processing skills • Optical Character Recognition (OCR) and identification of visual features, such as facial detection, image interpretation, image recognition
  5. Semantic Search • Language Understanding • Semantic Captions and Highlights

    • Semantic Answers • Semantic ranking instead of similarity scoring algorithm. • Spell Check
  6. Why not a database search? • Linguistic-aware text processing (stemming,

    lemmatization, word forms) in 56 languages • Autocorrection of misspelled words, synonyms, suggestions, scoring controls, facets, and custom tokenization. • Offload search
  7. Why a database search? • Paging in Azure Search relies

    on the $skip parameter, which is currently capped at a maximum of 100000. (use of partitions to read all data) • No backup / restore • Static schema • No joins if needed. • Optimized heavily for query performance, slower write performance

More Decks by Daron Yondem

Other Decks in Technology