Introduction to AI in Azure Cognitive Search
AI enrichment is a capability of Azure Cognitive Search indexing used to extract text from images, blobs, and other unstructured data sources - enriching the content to make it more searchable in an index or knowledge store. Extraction and enrichment are implemented through cognitive skills attached to an indexing pipeline. Cognitive skills built into the service fall into these categories:
Natural language processing skills include entity recognition, language detection, key phrase extraction, text manipulation, and sentiment detection. With these skills, unstructured text can assume new forms, mapped as searchable and filterable fields in an index.
Image processing skills include Optical Character Recognition (OCR) and identification of visual features, such as facial detection, image interpretation, image recognition (famous people and landmarks) or attributes like colors or image orientation. You can create text-representations of image content, searchable using all the query capabilities of Azure Cognitive Search.
Natural language and image processing is applied during the data ingestion phase, with results becoming part of a document's composition in a searchable index in Azure Cognitive Search. Data is sourced as an Azure data set and then pushed through an indexing pipeline using whichever built-in skills you need. The architecture is extensible so if the built-in skills are not sufficient, you can create and attach custom skills to integrate custom processing. Examples might be a custom entity module or document classifier targeting a specific domain such as finance, scientific publications, or medicine.
As you expand scope by increasing the frequency of processing, adding more documents, or adding more AI algorithms, you will need to attach a billable Cognitive Services resource. Charges accrue when calling APIs in Cognitive Services, and for image extraction as part of the document-cracking stage in Azure Cognitive Search. There are no charges for text extraction from documents.
When to use cognitive skills
You should consider using built-in cognitive skills if your raw content is unstructured text, image content, or content that needs language detection and translation. Applying AI through the built-in cognitive skills can unlock this content, increasing its value and utility in your search and data science apps.
Additionally, you might consider adding a custom skill if you have open-source, third-party, or first-party code that you'd like to integrate into the pipeline. Classification models that identify salient characteristics of various document types fall into this category, but any package that adds value to your content could also be used.
More about built-in skills
A skillset that's assembled using built-in skills is well suited for the following application scenarios:
Scanned documents (JPEG) that you want to make full-text searchable. You can attach an optical character recognition (OCR) skill to identify, extract, and ingest text from JPEG files.
PDFs with combined image and text. Text in PDFs can be extracted during indexing without the use of enrichment steps, but the addition of image and natural language processing can often produce a better outcome than a standard indexing provides.
Multi-lingual content against which you want to apply language detection and possibly text translation.
Unstructured or semi-structured documents containing content that has inherent meaning or context that is hidden in the larger document.
Blobs in particular often contain a large body of content that is packed into a singled "field". By attaching image and natural language processing skills to an indexer, you can create new information that is extant in the raw content, but not otherwise surfaced as distinct fields. Some ready-to-use built-in cognitive skills that can help: key phrase extraction, sentiment analysis, and entity recognition (people, organizations, and locations).
Additionally, built-in skills can also be used restructure content through text split, merge, and shape operations.
More about custom skills
Custom skills can support more complex scenarios, such as recognizing forms, or custom entity detection using a model that you provide and wrap in the custom skill web interface. Several examples of custom skills include Forms Recognizer, integration of the Bing Entity Search API, and custom entity recognition.
Components of an enrichment pipeline
An enrichment pipeline is based on indexers that crawl data sources and provide end-to-end index processing. Skills are now attached to indexers, intercepting and enriching documents according to the skillset you define. Once indexed, you can access content via search requests through all query types supported by Azure Cognitive Search. If you are new to indexers, this section walks you through the steps.
Step 1: Connection and document cracking phase
At the start of the pipeline, you have unstructured text or non-text content (such as image and scanned document JPEG files). Data must exist in an Azure data storage service that can be accessed by an indexer. Indexers can "crack" source documents to extract text from source data.
Supported sources include Azure blob storage, Azure table storage, Azure SQL Database, and Azure Cosmos DB. Text-based content can be extracted from the following file types: PDFs, Word, PowerPoint, CSV files. For the full list, see Supported formats.
Step 2: Cognitive skills and enrichment phase
Enrichment is through cognitive skills performing atomic operations. For example, once you have text content from a PDF, you can apply entity recognition language detection, or key phrase extraction to produce new fields in your index that are not available natively in the source. Altogether, the collection of skills used in your pipeline is called a skillset.
A skillset is based on built-in cognitive skills or custom skills you provide and connect to the skillset. A skillset can be minimal or highly complex, and determines not only the type of processing, but also the order of operations. A skillset plus the field mappings defined as part of an indexer fully specifies the enrichment pipeline. For more information about pulling all of these pieces together, see Define a skillset.
Internally, the pipeline generates a collection of enriched documents. You can decide which parts of the enriched documents should be mapped to indexable fields in your search index. For example, if you applied the key phrases extraction and the entity recognition skills, then those new fields would become part of the enriched document, and they can be mapped to fields on your index. See Annotations to learn more about input/output formations.
Add a knowledgeStore element to save enrichments
Search REST api-version=2019-05-06-Preview extends skillsets with a
knowledgeStore definition that provides an Azure storage connection and projections that describe how the enrichments are stored.
Adding a knowledge store to a skillset gives you the ability to project a representation of your enrichments for scenarios other than full text search. For more information, see Knowledge store (preview).
Step 3: Search index and query-based access
When processing is finished, you have a search index consisting of enriched documents, fully text-searchable in Azure Cognitive Search. Querying the index is how developers and users access the enriched content generated by the pipeline.
The index is like any other you might create for Azure Cognitive Search: you can supplement with custom analyzers, invoke fuzzy search queries, add filtered search, or experiment with scoring profiles to reshape the search results.
Indexes are generated from an index schema that defines the fields, attributes, and other constructs attached to a specific index, such as scoring profiles and synonym maps. Once an index is defined and populated, you can index incrementally to pick up new and updated source documents. Certain modifications require a full rebuild. You should use a small data set until the schema design is stable. For more information, see How to rebuild an index.
Key features and concepts
|Skillset||A top-level named resource containing a collection of skills. A skillset is the enrichment pipeline. It is invoked during indexing by an indexer.||See Define a skillset|
|Cognitive skill||An atomic transformation in an enrichment pipeline. Often, it is a component that extracts or infers structure, and therefore augments your understanding of the input data. Almost always, the output is text-based and the processing is natural language processing or image processing that extracts or generates text from image inputs. Output from a skill can be mapped to a field in an index, or used as an input for a downstream enrichment. A skill is either predefined and provided by Microsoft, or custom: created and deployed by you.||Built-in cognitive skills|
|Data extraction||Covers a broad range of processing, but pertaining to AI enrichment, the entity recognition skill is most typically used to extract data (an entity) from a source that doesn't provide that information natively.||See Entity Recognition Skill and Document Extraction Skill (preview)|
|Image processing||Infers text from an image, such as the ability to recognize a landmark, or extracts text from an image. Common examples include OCR for lifting characters from a scanned document (JPEG) file, or recognizing a street name in a photograph containing a street sign.||See Image Analysis Skill or OCR Skill|
|Natural language processing||Text processing for insights and information about text inputs. Language detection, sentiment analysis, and key phrase extraction are skills that fall under natural language processing.||See Key Phrase Extraction Skill, Language Detection Skill, Text Translation Skill (preview), Sentiment Analysis Skill|
|Document cracking||The process of extracting or creating text content from non-text sources during indexing. Optical character recognition (OCR) is an example, but generally it refers to core indexer functionality as the indexer extracts content from application files. The data source providing source file location, and the indexer definition providing field mappings, are both key factors in document cracking.||See Indexers overview|
|Shaping||Consolidate text fragments into a larger structure, or conversely break down larger text chunks into a manageable size for further downstream processing.||See Shaper Skill, Text Merger Skill, Text Split Skill|
|Enriched documents||A transitory internal structure, generated during processing, with final output reflected in a search index. A skillset determines which enrichments are performed. Field mappings determine which data elements are added to the index. Optionally, you can create a knowledge store to persist and explore enriched documents using tools like Storage Explorer, Power BI, or any other tool that connects to Azure Blob storage.||See Knowledge store (preview)|
|Indexer||A crawler that extracts searchable data and metadata from an external data source and populates an index based on field-to-field mappings between the index and your data source for document cracking. For AI enrichments, the indexer invokes a skillset, and contains the field mappings associating enrichment output to target fields in the index. The indexer definition contains all of the instructions and references for pipeline operations, and the pipeline is invoked when you run the indexer. With additional configuration, you can re-use existing processed content and execute only those steps and skills that are changed.||See Indexers and Incremental enrichment (preview).|
|Data Source||An object used by an indexer to connect to an external data source of supported types on Azure.||See Indexers overview|
|Index||A persisted search index in Azure Cognitive Search, built from an index schema that defines field structure and usage.||See Create a basic index|
|Knowledge store||A storage account where the enriched documents can be shaped and projected in addition to the search index||See Introduction to knowledge store|
|Cache||A storage account that contains cached output created by an enrichment pipeline. Enabling the cache preserves existing output that is unaffected by changes to a skillset or other components of the enrichment pipeline.||See Incremental enrichment|
Where do I start?
Step 2: Try some quickstarts and examples for hands-on experience
- Quickstart (portal)
- Tutorial (HTTP requests)
- Example: Creating a custom skill for AI enrichment (C#)
We recommend the Free service for learning purposes, however the number of free transactions is limited to 20 documents per day. To run both the quickstart and tutorial in one day, use a smaller file set (10 documents) so that you can fit in both exercises, or delete the indexer you used in the quickstart or tutorial to rest the counter to zero.
Step 3: Review the API
You can use REST
api-version=2019-05-06 on requests or the .NET SDK. If you are exploring knowledge store, use the preview REST API instead (
This step uses the REST APIs to build an AI enrichment solution. Only two APIs are added or extended for AI enrichment. Other APIs have the same syntax as the generally available versions.
|Create Data Source||A resource identifying an external data source providing source data used to create enriched documents.|
|Create Skillset (api-version=2019-05-06)||This API is specific to AI enrichment. It is a resource coordinating the use of built-in skills and custom cognitive skills used in an enrichment pipeline during indexing.|
|Create Index||A schema expressing an Azure Cognitive Search index. Fields in the index map to fields in source data or to fields manufactured during the enrichment phase (for example, a field for organization names created by entity recognition).|
|Create Indexer (api-version=2019-05-06)||A resource defining components used during indexing: including a data source, a skillset, field associations from source and intermediary data structures to target index, and the index itself. Running the indexer is the trigger for data ingestion and enrichment. The output is a search index based on the index schema, populated with source data, enriched through skillsets. This existing API is extended for cognitive search scenarios with the inclusion of a skillset property.|
Checklist: A typical workflow
Subset your Azure source data into a representative sample. Indexing takes time so start with a small, representative data set and then build it up incrementally as your solution matures.
Create a data source object in Azure Cognitive Search to provide a connection string for data retrieval.
Create a skillset with enrichment steps.
Define the index schema. The Fields collection includes fields from source data. You should also stub out additional fields to hold generated values for content created during enrichment.
Define the indexer referencing the data source, skillset, and index.
Within the indexer, add outputFieldMappings. This section maps output from the skillset (in step 3) to the inputs fields in the index schema (in step 4).
Send Create Indexer request you just created (a POST request with an indexer definition in the request body) to express the indexer in Azure Cognitive Search. This step is how you run the indexer, invoking the pipeline.
Run queries to evaluate results and modify code to update skillsets, schema, or indexer configuration.
Reset the indexer before rebuilding the pipeline.
For more information about specific questions or problems, see Troubleshooting tips.