Azure Search Review

Incorporate Artificial Intelligence Through Cognitive Search Feature

One of the unique and powerful capabilities offered by Azure Search is that incorporates artificial intelligence into the indexing process. This can be observed in its feature called Cognitive Search. This feature enables users to automatically extract data from source documents or contents and convert them into structured and searchable data which will be a part of the search index later on. How the data is transformed depends on the set of steps and resources that users set up which are known as cognitive skills. These skills use AI algorithms and they are very useful for structuring and enriching data.

Source Data And Document Cracking

The Cognitive Search feature operates through a pipeline which is divided into different phases which include source data and document cracking, cognitive skills and enrichment, and search index and query-based access phase. The first phase of the pipeline involves the process of extracting text or image-based content from source documents. Through document cracking, users will be able to extract or generate text-based content from image files or non-text sources. In this phase, the content is not yet structured and searchable.

Convert Unstructured Content To Searchable Content

The second phase of the pipeline is very critical, as this is the phase where the cognitive skills play a major role, helping users gain insights into the extracted or generated content, make inferences on it, and convert it to a structured and searchable content. Azure Search is comprised of two general types of cognitive skills: Predefined cognitive skills and custom cognitive skills. A set of cognitive skills is called skillset. One of the predefined skills that users can apply is the Named Entity Recognition cognitive skill. With this skill, named entities or data are instantly extracted from the content such as names of persons, locations, or organizations. Another predefined skill is the Key Phrase Extraction cognitive skill. This skill permits users to generate a list of phrases from an unstructured content.

Custom Cognitive Skills And Enriched Documents

When it comes to custom cognitive skills, these are skills that can be set up by users to perform special data evaluation processes and indexing operations. For instance, they can extract data based on a specific area of specialization. After the contents are processed, they are then organized into a collection of enriched documents. Whatever data or content found in those documents can be mapped to the index fields that comprised the search index. However, users have the option to decide which data or content they want to include in the search index. Once the content is mapped to the index field in the search index, it will now become one of those contents that the search solution will return as it generates search results for queries.

Apply Full-Text Search And Text Analysis To Heterogeneous Contents

Azure Search like other search solutions and services is using a very common technique called full-text search and text analysis. In this technique, a search engine or solution checks all the words contained within a stored document or a full-text database and returns search results based on the query made. However, although others are applying such technique, it sometimes only work on contents that are managed by the same database management system. On the contrary, Azure Search’s full-text search capability can be implemented even for contents that belong to different database management systems, allowing users to store and search contents from heterogeneous sources.

Leverage The Lucene Query Syntax

In performing full-text search and text analysis, Azure Search supports the implementation of the industry-leading search query syntax called Lucene query syntax. This is a query language developed by Apache and users can apply and enable such syntax as they build search queries for their search solution. The Lucene query syntax is recommended for various types of query operations such as what is popularly known as fuzzy search. A fuzzy search is used for searching for words or terms that are different in spelling but may share one or more similar letters (e.g. blue, blues, and glue).

Another type of query operation where the syntax can be applied is when performing a proximity search. Here, words that are close to each other are the ones being searched. A tilde “~” symbol is added at the end of the two words and followed by a number. This number is the proximity boundary and it defines the distance which is the maximum number of intermediary words. For example, when the expression “hotel airport”~5 is entered, “hotel” and “airport” will be searched within 5 words of each other in the document.

Tailor Your Search Results To Your Business Using Rank Scoring

Azure Search is also equipped with a relevance modeling capability. With this capability, users can customize how they score the ranking of contents or items appearing in their search results. Basically, a content that has a high score also has a high relevance, placing it at the top of the search results. By creating their own scoring profiles, users will be able to define the relevance and ranking of contents in the search results. In other words, they can configure how contents are showing in the search results, tailoring them to their business requirements and goals. For example, if they are promoting new products, they can set up scoring profiles for these products so that they will rank higher in the search results.