{"id":729205,"date":"2021-03-02T06:00:54","date_gmt":"2021-03-02T14:00:54","guid":{"rendered":"https:\/\/www.microsoft.com\/en-us\/research\/?p=729205"},"modified":"2021-03-04T12:15:43","modified_gmt":"2021-03-04T20:15:43","slug":"the-science-behind-semantic-search-how-ai-from-bing-is-powering-azure-cognitive-search","status":"publish","type":"post","link":"https:\/\/www.microsoft.com\/en-us\/research\/blog\/the-science-behind-semantic-search-how-ai-from-bing-is-powering-azure-cognitive-search\/","title":{"rendered":"The science behind semantic search: How AI from Bing is powering Azure Cognitive Search"},"content":{"rendered":"\n<figure class=\"wp-block-image alignwide size-large\"><img decoding=\"async\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2021\/03\/1400x788_Search_Semantic_nologo.gif\" alt=\"\"\/><\/figure>\n\n\n\n<p><a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/azure.microsoft.com\/en-us\/services\/search\/\" target=\"_blank\" rel=\"noopener noreferrer\">Azure Cognitive Search<span class=\"sr-only\"> (opens in new tab)<\/span><\/a>&nbsp;is a cloud search service that gives developers APIs and tools&nbsp;to&nbsp;build rich search experiences&nbsp;over private, heterogeneous content in web, mobile, and enterprise&nbsp;applications.&nbsp;It has multiple components,&nbsp;including&nbsp;an&nbsp;API for indexing and querying, seamless integration through Azure&nbsp;data ingestion,&nbsp;deep integration with&nbsp;Azure&nbsp;Cognitive Services,&nbsp;and persistent storage of user-owned indexed content.&nbsp;At the heart of Azure Cognitive Search is its full text, keyword-based&nbsp;search&nbsp;engine built&nbsp;on&nbsp;the&nbsp;BM25&nbsp;algorithm\u2014an industry standard in information retrieval.&nbsp;&nbsp;<\/p>\n\n\n\n<p>We\u2019ve found that what customers desire next is higher-quality results out of the box with less effort, enabling them to deliver more relevant and actionable results to <em>their <\/em>customers.<\/p>\n\n\n\n<p>As part of our <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/project\/ai-at-scale\/\">AI at Scale<\/a> effort, we lean heavily on recent developments in large Transformer-based language models to improve the relevance quality of Microsoft Bing. These improvements allow a search engine to go beyond keyword matching to searching using the semantic meaning behind words and content. We call this transformational ability <em>semantic search<\/em>\u2014a major showcase of what AI at Scale can deliver for customers.<\/p>\n\n\n\n<p>Semantic search has significantly advanced the quality of Bing search results, and it has been a companywide effort: top applied scientists and engineers from Bing leverage the latest technology from Microsoft Research and Microsoft Azure. Maximizing the power of AI at Scale requires a lot of sophistication. One needs to pretrain large Transformer-based models, perform multi-task fine-tuning across various tasks, and distill big models to a servable form with very minimal loss of quality. We recognize that it takes a large group of specialized talent to integrate and deploy AI at Scale products for customers, and many companies can\u2019t afford these types of teams. To empower every person and every organization on the planet, we need to significantly lower the bar for everyone to use AI at Scale technology.<\/p>\n\n\n\n<p>Today, Microsoft is excited to announce that we\u2019re bringing semantic search capabilities to all Azure customers in preview. You no longer need a team of deep learning experts to take advantage of this technology: we packaged all the best <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" rel=\"noopener noreferrer\" target=\"_blank\" href=\"http:\/\/aka.ms\/aiatscale\">AI at Scale <span class=\"sr-only\"> (opens in new tab)<\/span><\/a>technology, from models to software to hardware, into a single end-to-end AI solution. This is AI at Scale for everyone.<\/p>\n\n\n\n<p>This post goes deeper into the Bing technology that made semantic search possible. We also encourage you to read the post \u201c<a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" rel=\"noopener noreferrer\" target=\"_blank\" href=\"https:\/\/aka.ms\/AnnouncingSemanticSearch\">Introducing semantic search: Bringing more meaningful results to Azure Cognitive Search<span class=\"sr-only\"> (opens in new tab)<\/span><\/a>,\u201d which explains what new capabilities are available to you and how you can get started today.<\/p>\n\n\n\n<figure class=\"wp-block-embed is-type-video is-provider-youtube wp-block-embed-youtube wp-embed-aspect-16-9 wp-has-aspect-ratio\"><div class=\"wp-block-embed__wrapper\">\n<div class=\"yt-consent-placeholder\" role=\"region\" aria-label=\"Video playback requires cookie consent\" data-video-id=\"d_6ZNyV1MvA\" data-poster=\"https:\/\/img.youtube.com\/vi\/d_6ZNyV1MvA\/maxresdefault.jpg\"><iframe aria-hidden=\"true\" tabindex=\"-1\" title=\"Build Semantic Search into Your Apps | Azure Cognitive Search\" width=\"500\" height=\"281\" data-src=\"https:\/\/www.youtube-nocookie.com\/embed\/d_6ZNyV1MvA?feature=oembed&rel=0&enablejsapi=1\" frameborder=\"0\" allow=\"accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture; web-share\" referrerpolicy=\"strict-origin-when-cross-origin\" allowfullscreen><\/iframe><div class=\"yt-consent-placeholder__overlay\"><button class=\"yt-consent-placeholder__play\"><svg width=\"42\" height=\"42\" xmlns=\"http:\/\/www.w3.org\/2000\/svg\" aria-hidden=\"true\" focusable=\"false\"><g fill=\"none\" fill-rule=\"evenodd\"><circle fill=\"#000\" opacity=\".556\" cx=\"21\" cy=\"21\" r=\"21\"\/><path stroke=\"#FFF\" d=\"M27.5 22l-12 8.5v-17z\"\/><\/g><\/svg><span class=\"yt-consent-placeholder__label\">Video playback requires cookie consent<\/span><\/button><\/div><\/div>\n<\/div><\/figure>\n\n\n\n<h2 id=\"semantic-search-capabilities-in-azure-cognitive-search\">Semantic search capabilities in Azure Cognitive Search<\/h2>\n\n\n\n<p>Below are the features enabled by semantic search in Azure Cognitive Search. Before our public preview release, we tested many of these features on our own products, and the examples below show the dramatic improvements in results we achieved by applying AI at Scale technology.<\/p>\n\n\n\n<p><strong>Semantic ranking: Massively improved relevance<\/strong><\/p>\n\n\n\n<p>Semantic search is moving beyond keyword-based ranking to a Transformer-based semantic ranking engine that understands the meaning behind the text. We performed A\/B testing of these technologies on Microsoft Docs, a giant library of technical documentation for users, developers, and IT professionals. The Microsoft Docs team has been tuning its search index for search relevance the last couple of years and ran multiple experiments during that period. Semantic search increased the clickthrough rate of the search result page by 2.0 percent compared to its current production system and by 4.5 percent on longer queries (three or more words) as measured by an A\/B experiment. This was the largest single improvement of key performance indicators for search engagement the Microsoft Docs team has ever seen.<\/p>\n\n\n\n<div class=\"wp-block-image\"><figure class=\"aligncenter size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"804\" height=\"429\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2021\/03\/SemanticSearch_Fig1.jpg\" alt=\"\" class=\"wp-image-729331\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2021\/03\/SemanticSearch_Fig1.jpg 804w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2021\/03\/SemanticSearch_Fig1-300x160.jpg 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2021\/03\/SemanticSearch_Fig1-768x410.jpg 768w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2021\/03\/SemanticSearch_Fig1-16x9.jpg 16w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2021\/03\/SemanticSearch_Fig1-710x380.jpg 710w\" sizes=\"auto, (max-width: 804px) 100vw, 804px\" \/><figcaption>Semantic ranking goes beyond keywords to capture the meaning behind search terms to return more relevant results. Above is an example of improved relevance through semantic ranking (right) on Microsoft Docs, a library of technical documentation.<\/figcaption><\/figure><\/div>\n\n\n\n<p><strong>Semantic captions: Extractive summarization<\/strong><\/p>\n\n\n\n<p>We\u2019ve found that returning relevant results isn\u2019t the only important aspect of a search engine. The captions and snippets are also important to inform people about why a result is relevant\u2014or perhaps not relevant\u2014so they can proceed with a result, skip it, or reformulate their queries. Semantic captions uses extractive summarization to pull a snippet from the document that best summarizes why we think it\u2019s relevant for the query.<\/p>\n\n\n\n<p><strong>Semantic highlights: Machine reading comprehension<\/strong><\/p>\n\n\n\n<p>Going beyond keyword-based highlighting in captions and snippets is also valuable for improving user experience. With the right hit highlighting, people can immediately and directly get the answer they\u2019re looking for, quickly scan a page of results to find the document they want, or even get a summary of a topic. Machine reading comprehension enables semantic highlights of relevant words or phrases on answers and captions to save people time.<\/p>\n\n\n\n<div class=\"wp-block-image\"><figure class=\"aligncenter size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"616\" height=\"490\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2021\/03\/Ai-for-image_figure-1.png\" alt=\"\" class=\"wp-image-731113\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2021\/03\/Ai-for-image_figure-1.png 616w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2021\/03\/Ai-for-image_figure-1-300x239.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2021\/03\/Ai-for-image_figure-1-16x12.png 16w\" sizes=\"auto, (max-width: 616px) 100vw, 616px\" \/><figcaption>Semantic captions and highlighting can help people more easily determine whether a result is relevant or even provide a direct answer or topic summary. Above is an example from Microsoft Bing of semantic captions and highlighting (the bolded terms) powered by extractive summarization and machine reading comprehension.<\/figcaption><\/figure><\/div>\n\n\n\n<p><strong>Semantic answers: Instant answers<\/strong><\/p>\n\n\n\n<p>Queries formulated as questions are one of the more important query segments on Bing. Our customers aren\u2019t just looking for relevant documents but <em>instant answers.<\/em> We aim to provide them through machine learning\u2014reading through all the documents in the corpus, running extractive summarization, and then using machine reading comprehension\u2014and finally promoting a direct answer to an individual\u2019s question to the top of the results.<\/p>\n\n\n\n<figure class=\"wp-block-image size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"906\" height=\"359\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2021\/03\/Semantic-Search-Teams-Updated-Fig.png\" alt=\"\" class=\"wp-image-729931\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2021\/03\/Semantic-Search-Teams-Updated-Fig.png 906w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2021\/03\/Semantic-Search-Teams-Updated-Fig-300x119.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2021\/03\/Semantic-Search-Teams-Updated-Fig-768x304.png 768w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2021\/03\/Semantic-Search-Teams-Updated-Fig-16x6.png 16w\" sizes=\"auto, (max-width: 906px) 100vw, 906px\" \/><figcaption>Above is an example of an instant answer over the Microsoft Docs index. Semantic search can highlight the exact phrase from a corpus of millions of documents, easing the cumbersome nature of today\u2019s search experience.<\/figcaption><\/figure>\n\n\n\n<p><strong>Automatic spell correction<\/strong><\/p>\n\n\n\n<p>Ten to 15 percent of the queries issued to search engines are misspelled. When a query is misspelled, it\u2019s difficult for any of the downstream search components to deliver results that match intent. Semantic search enables automatic spell correction, so customers don\u2019t have to worry about having the perfect spelling.<\/p>\n\n\n\n<h2 id=\"the-best-of-microsoft-research\">The best of Microsoft Research <\/h2>\n\n\n\n<p>One of our promises for semantic search is to deliver to customers state-of-the-art technologies from research and product groups across Microsoft and the broader semantic search community, at the lowest cost. Microsoft Research has been at the forefront of some of the most important breakthroughs in natural language understanding and semantic search. Microsoft Research, together with other AI teams across the company, has developed a wide range of neural language models that substantially lift the state of the art across all major research benchmarks, including achieving human parity on public natural language understanding benchmarks such as <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" rel=\"noopener noreferrer\" target=\"_blank\" href=\"https:\/\/blogs.microsoft.com\/ai\/microsoft-creates-ai-can-read-document-answer-questions-well-person\/\">SQuAD<span class=\"sr-only\"> (opens in new tab)<\/span><\/a>, <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" rel=\"noopener noreferrer\" target=\"_blank\" href=\"https:\/\/docs.microsoft.com\/en-us\/archive\/blogs\/stevengu\/microsoft-achieves-human-performance-estimate-on-glue-benchmark\">GLUE<span class=\"sr-only\"> (opens in new tab)<\/span><\/a>, and <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/blog\/microsoft-deberta-surpasses-human-performance-on-the-superglue-benchmark\/\">SuperGLUE<\/a>. These innovations from Microsoft Research have been adapted for real-world scenarios and incorporated into Microsoft products (including Azure Cognitive Search).<\/p>\n\n\n\n<p><strong>UniLM<\/strong><\/p>\n\n\n\n<p><a href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/unilmv2-pseudo-masked-language-models-for-unified-language-model-pre-training\/\" target=\"_blank\" rel=\"noreferrer noopener\">&#8220;UniLMv2: Pseudo-Masked Language Models for Unified Language Model Pre-Training&#8221;<\/a>&nbsp;(ICML&nbsp;2020)&nbsp;&nbsp;<\/p>\n\n\n\n<p><a href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/unified-language-model-pre-training-for-natural-language-understanding-and-generation\/\" target=\"_blank\" rel=\"noreferrer noopener\">&#8220;Unified Language Model Pre-training for Natural Language Understanding and Generation&#8221;<\/a>&nbsp;(NeurIPS&nbsp;2019)&nbsp;<\/p>\n\n\n\n<p><a href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/unified-language-model-pre-training-for-natural-language-understanding-and-generation\/#:~:text=Unified%20Language%20Model%20Pre-training%20for%20Natural%20Language%20Understanding,language%20modeling%20tasks:%20unidirectional,%20bidirectional,%20and%20sequence-to-sequence%20prediction.\">UniLM<\/a> (Unified Language Model Pre-training) is a unified pretraining and fine-tuning framework that supports both natural language understanding and generation tasks, combining the strengths of autoencoding models like Google\u2019s BERT and autoregressive models like OpenAI\u2019s GPT-2. The unified modeling is achieved by employing a shared Transformer network and utilizing specific self-attention masks to control what context the prediction conditions on. We also propose a novel training procedure, referred to as a <em>pseudo-masked language model<\/em>, to effectively and efficiently pretrain different language modeling objectives in one forward pass. UniLM is the first unified pretrained model that shows strong results on both language understanding and generation benchmarks, <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" rel=\"noopener noreferrer\" target=\"_blank\" href=\"https:\/\/venturebeat.com\/2019\/10\/16\/microsofts-unilm-ai-achieves-state-of-the-art-performance-on-summarization-and-language-generation\/\">creating a new state of the art on natural language benchmarks when it was introduced<span class=\"sr-only\"> (opens in new tab)<\/span><\/a> in addition to its competitive results on natural language tasks. The second version of the model, UniLMv2, outperformed other related models such as BERT, Google\u2019s XLNet, and Facebook\u2019s RoBERTa across SQuAD and GLUE at the time. Although Bing search technology was <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" rel=\"noopener noreferrer\" target=\"_blank\" href=\"https:\/\/azure.microsoft.com\/en-us\/blog\/bing-delivers-its-largest-improvement-in-search-experience-using-azure-gpus\/\">BERT-based early in 2019<span class=\"sr-only\"> (opens in new tab)<\/span><\/a>, we\u2019ve moved it to UniLMv2 in 2020 for improved quality, and UniLMv2 has been incorporated into the Microsoft pretrained Turing language models.<\/p>\n\n\n\n<div class=\"wp-block-image\"><figure class=\"aligncenter size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"720\" height=\"516\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2021\/03\/Semantic-Search-Fig-4.png\" alt=\"\" class=\"wp-image-729337\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2021\/03\/Semantic-Search-Fig-4.png 720w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2021\/03\/Semantic-Search-Fig-4-300x215.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2021\/03\/Semantic-Search-Fig-4-16x12.png 16w\" sizes=\"auto, (max-width: 720px) 100vw, 720px\" \/><figcaption>This figure provides an overview of Unified Language Model Pre-training (UniLM). The model parameters are shared across the language model (LM) objectives\u2014that is, bidirectional LM, left-to-right LM, and sequence-to-sequence LM\u2014and different self-attention masks to control the access to context for each word token are used.<\/figcaption><\/figure><\/div>\n\n\n\n<p><strong>Graph&nbsp;<\/strong><strong>a<\/strong><strong>ttention&nbsp;<\/strong><strong>n<\/strong><strong>etworks<\/strong><strong>&nbsp;for&nbsp;<\/strong><strong>machine r<\/strong><strong>eading<\/strong><strong>&nbsp;<\/strong><strong>c<\/strong><strong>omprehension<\/strong>&nbsp;<\/p>\n\n\n\n<p><a href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/document-modeling-with-graph-attention-networks-for-multi-grained-machine-reading-comprehension\/\" target=\"_blank\" rel=\"noreferrer noopener\">&#8220;Document Modeling with&nbsp;Graph Attention Networks for Multi-grained Machine Reading Comprehension&#8221;<\/a>&nbsp;<\/p>\n\n\n\n<p>The task of machine reading comprehension (MRC) is to find short answers, such as&nbsp;phrases,&nbsp;or long answers,&nbsp;such as&nbsp;paragraphs,&nbsp;from documents\u202fwith respect to given questions.&nbsp;Because of&nbsp;the max length limitation, most existing MRC methods treat documents as separate paragraphs in the answer extraction procedure and&nbsp;don\u2019t&nbsp;consider their internal relationship. To better model the documents for MRC,&nbsp;we propose&nbsp;a multi-grained machine reading comprehension framework where we first build a graph for each document based on its hierarchical nature\u2014that is, documents, paragraphs, sentences, and tokens\u2014and then use graph attention networks to learn different levels of representations. In this way, we can directly derive scores of long answers from its paragraph-level representations and obtain scores of short answers from the start and end positions on the token-level representations. Thus,\u202fboth\u202flong and short answer extraction tasks can be trained jointly to promote each other. The model has been\u202fat the top of&nbsp;<a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/ai.google.com\/research\/NaturalQuestions\/leaderboard\" target=\"_blank\" rel=\"noopener noreferrer\">Google&#8217;s Natural Question leaderboard<span class=\"sr-only\"> (opens in new tab)<\/span><\/a>&nbsp;since\u202fits submission in February 2020 and&nbsp;is&nbsp;being used inside&nbsp;Bing&nbsp;for much of its question answering.\u202f&nbsp;<\/p>\n\n\n\n<p><strong>Multi-Task\u00a0Deep Neural Networks<\/strong>\u00a0<\/p>\n\n\n\n<p><a href=\"https:\/\/www.microsoft.com\/en-us\/research\/publication\/multi-task-deep-neural-networks-for-natural-language-understanding-2\/\" target=\"_blank\" rel=\"noreferrer noopener\">&#8220;Multi-Task Deep Neural Networks for Natural Language Understanding&#8221;<\/a>\u00a0<\/p>\n\n\n\n<p>Multi-Task Deep Neural Network (MT-DNN), developed by Microsoft Research and Microsoft Dynamics 365 AI, is <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" rel=\"noopener noreferrer\" target=\"_blank\" href=\"https:\/\/docs.microsoft.com\/en-us\/archive\/blogs\/stevengu\/microsoft-achieves-human-performance-estimate-on-glue-benchmark\">the first AI model that surpassed human performance on the GLUE leaderboard<span class=\"sr-only\"> (opens in new tab)<\/span><\/a>, and it remained the best model on the leaderboard until December 2019, outperforming all competing models, including BERT, Google\u2019s T5, and RoBERTa. MT-DNN combines multi-task learning and language model pretraining for learning language representations across multiple natural language understanding tasks. MT-DNN not only leverages large amounts of cross-task data, but it also benefits from a regularization effect that leads to more general representations to adapt to new tasks and domains. The models in semantic search are leveraging multi-task learning across various search tasks to maximize their quality.<\/p>\n\n\n\n<p><strong>MiniLM<\/strong><\/p>\n\n\n\n<p><a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/arxiv.org\/pdf\/2002.10957.pdf%22%20\/\" target=\"_blank\" rel=\"noopener noreferrer\">MiniLM:\u00a0Deep Self-Attention Distillation for\u00a0Task-Agnostic Compression of Pre-Trained Transformers&#8221;<span class=\"sr-only\"> (opens in new tab)<\/span><\/a>\u00a0(NeurIPS\u00a02020)\u00a0<\/p>\n\n\n\n<p>Although large-scale Transformer-based models are powerful, they\u2019re expensive for online serving in production settings. Microsoft Research has developed state-of-the-art task-agnostic knowledge distillation techniques to compress big pretrained Transformer-based models into small and fast ones for online serving while minimizing performance loss. MiniLM is a model compression method based on deep self-attention distillation, where a compressed (small) model is trained by deeply mimicking the self-attention module of a large model to be compressed. MiniLM achieves state-of-the-art results in small-model settings on both monolingual and multilingual natural language benchmarks. We\u2019ve used MiniLM distillation for all the large models in our semantic search solutions; these large models retain 95 percent of the accuracy at only 20 percent of the cost.<\/p>\n\n\n\n<div class=\"wp-block-image\"><figure class=\"aligncenter size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"720\" height=\"327\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2021\/03\/SemanticSearch-Figure5.png\" alt=\"\" class=\"wp-image-729352\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2021\/03\/SemanticSearch-Figure5.png 720w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2021\/03\/SemanticSearch-Figure5-300x136.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2021\/03\/SemanticSearch-Figure5-16x7.png 16w\" sizes=\"auto, (max-width: 720px) 100vw, 720px\" \/><figcaption>This figure provides an overview of MiniLM, a model compression method based on deep self-attention distillation. The student is trained by deeply mimicking the self-attention behavior of the last Transformer layer of the teacher.<\/figcaption><\/figure><\/div>\n\n\n\n<h2 id=\"conclusion\">Conclusion<\/h2>\n\n\n\n<p>The AI models behind semantic search are extremely powerful and have been validated both against external academic benchmarks and real user traffic on Bing. By bringing semantic search to Azure Cognitive Search, we\u2019re taking a major step toward democratizing advanced machine learning technologies for everyone. Semantic search will deliver better experiences for your end users while also enabling anyone to tap into the power of AI at Scale using fewer resources and without specialized expertise. We believe semantic search on Azure Cognitive Search offers the best combination of search relevance, developer experience, and cloud service capabilities available on the market. Start using semantic search in your service today by signing up for the preview at <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" rel=\"noopener noreferrer\" target=\"_blank\" href=\"http:\/\/aka.ms\/semanticpreview\">aka.ms\/semanticpreview<span class=\"sr-only\"> (opens in new tab)<\/span><\/a>.<\/p>\n\n\n\n<p>If you\u2019re interested in applying the latest deep learning techniques to innovate in search, <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" rel=\"noopener noreferrer\" target=\"_blank\" href=\"https:\/\/careers.microsoft.com\/us\/en\/search-results?keywords=%23semanticsearch%23\">our Search and AI team is hiring globally<span class=\"sr-only\"> (opens in new tab)<\/span><\/a>!<\/p>\n\n\n\n<p><em><strong>Acknowledgme<\/strong><\/em><strong><em>nts<\/em><\/strong><br><em>This work in semantic search has been the result of companywide collaboration, including teams from Microsoft Research, Microsoft Azure, Microsoft Bing, and Azure Cognitive Search. Blog post authors Rangan Majumder, Alec Berntson, and Daxin Jiang are from Bing, while authors Jianfeng Gao, Furu Wei, and Nan Duan are from Microsoft Research.<\/em><\/p>\n","protected":false},"excerpt":{"rendered":"<p>Azure Cognitive Search (opens in new tab)&nbsp;is a cloud search service that gives developers APIs and tools&nbsp;to&nbsp;build rich search experiences&nbsp;over private, heterogeneous content in web, mobile, and enterprise&nbsp;applications.&nbsp;It has multiple components,&nbsp;including&nbsp;an&nbsp;API for indexing and querying, seamless integration through Azure&nbsp;data ingestion,&nbsp;deep integration with&nbsp;Azure&nbsp;Cognitive Services,&nbsp;and persistent storage of user-owned indexed content.&nbsp;At the heart of Azure Cognitive Search [&hellip;]<\/p>\n","protected":false},"author":38838,"featured_media":729925,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"msr-url-field":"","msr-podcast-episode":"","msrModifiedDate":"","msrModifiedDateEnabled":false,"ep_exclude_from_search":false,"_classifai_error":"","msr-author-ordering":[{"type":"user_nicename","value":"Rangan Majumder","user_id":"38931"},{"type":"user_nicename","value":"Alec Berntson","user_id":"40123"},{"type":"user_nicename","value":"Daxin Jiang (\u59dc\u5927\u6615)","user_id":"31642"},{"type":"user_nicename","value":"Jianfeng Gao","user_id":"32246"},{"type":"user_nicename","value":"Furu Wei","user_id":"31830"},{"type":"user_nicename","value":"Nan Duan","user_id":"33052"}],"msr_hide_image_in_river":0,"footnotes":""},"categories":[1],"tags":[],"research-area":[13556],"msr-region":[],"msr-event-type":[],"msr-locale":[268875],"msr-post-option":[],"msr-impact-theme":[],"msr-promo-type":[],"msr-podcast-series":[],"class_list":["post-729205","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-research-blog","msr-research-area-artificial-intelligence","msr-locale-en_us"],"msr_event_details":{"start":"","end":"","location":""},"podcast_url":"","podcast_episode":"","msr_research_lab":[199560],"msr_impact_theme":[],"related-publications":[],"related-downloads":[],"related-videos":[],"related-academic-programs":[],"related-groups":[],"related-projects":[649749],"related-events":[],"related-researchers":[{"type":"user_nicename","value":"Rangan Majumder","user_id":38931,"display_name":"Rangan Majumder","author_link":"<a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/ranganm\/\" aria-label=\"Visit the profile page for Rangan Majumder\">Rangan Majumder<\/a>","is_active":false,"last_first":"Majumder, Rangan","people_section":0,"alias":"ranganm"},{"type":"user_nicename","value":"Alec Berntson","user_id":40123,"display_name":"Alec Berntson","author_link":"<a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/alecb\/\" aria-label=\"Visit the profile page for Alec Berntson\">Alec Berntson<\/a>","is_active":false,"last_first":"Berntson, Alec","people_section":0,"alias":"alecb"},{"type":"user_nicename","value":"Jianfeng Gao","user_id":32246,"display_name":"Jianfeng Gao","author_link":"<a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/jfgao\/\" aria-label=\"Visit the profile page for Jianfeng Gao\">Jianfeng Gao<\/a>","is_active":false,"last_first":"Gao, Jianfeng","people_section":0,"alias":"jfgao"},{"type":"user_nicename","value":"Furu Wei","user_id":31830,"display_name":"Furu Wei","author_link":"<a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/fuwei\/\" aria-label=\"Visit the profile page for Furu Wei\">Furu Wei<\/a>","is_active":false,"last_first":"Wei, Furu","people_section":0,"alias":"fuwei"}],"msr_type":"Post","featured_image_thumbnail":"<img width=\"960\" height=\"540\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2021\/03\/1400x788_semantic_search_still_no_logo-960x540.jpg\" class=\"img-object-cover\" alt=\"graphical user interface, text, application, email\" decoding=\"async\" loading=\"lazy\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2021\/03\/1400x788_semantic_search_still_no_logo-960x540.jpg 960w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2021\/03\/1400x788_semantic_search_still_no_logo-300x169.jpg 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2021\/03\/1400x788_semantic_search_still_no_logo-1024x576.jpg 1024w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2021\/03\/1400x788_semantic_search_still_no_logo-768x432.jpg 768w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2021\/03\/1400x788_semantic_search_still_no_logo-1536x865.jpg 1536w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2021\/03\/1400x788_semantic_search_still_no_logo-2048x1153.jpg 2048w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2021\/03\/1400x788_semantic_search_still_no_logo-16x9.jpg 16w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2021\/03\/1400x788_semantic_search_still_no_logo-1066x600.jpg 1066w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2021\/03\/1400x788_semantic_search_still_no_logo-655x368.jpg 655w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2021\/03\/1400x788_semantic_search_still_no_logo-343x193.jpg 343w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2021\/03\/1400x788_semantic_search_still_no_logo-640x360.jpg 640w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2021\/03\/1400x788_semantic_search_still_no_logo-1280x720.jpg 1280w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2021\/03\/1400x788_semantic_search_still_no_logo-1920x1080.jpg 1920w\" sizes=\"auto, (max-width: 960px) 100vw, 960px\" \/>","byline":"","formattedDate":"March 2, 2021","formattedExcerpt":"Azure Cognitive Search (opens in new tab)&nbsp;is a cloud search service that gives developers APIs and tools&nbsp;to&nbsp;build rich search experiences&nbsp;over private, heterogeneous content in web, mobile, and enterprise&nbsp;applications.&nbsp;It has multiple components,&nbsp;including&nbsp;an&nbsp;API for indexing and querying, seamless integration through Azure&nbsp;data ingestion,&nbsp;deep integration with&nbsp;Azure&nbsp;Cognitive Services,&nbsp;and persistent storage&hellip;","locale":{"slug":"en_us","name":"English","native":"","english":"English"},"_links":{"self":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/posts\/729205","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/users\/38838"}],"replies":[{"embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/comments?post=729205"}],"version-history":[{"count":43,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/posts\/729205\/revisions"}],"predecessor-version":[{"id":731119,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/posts\/729205\/revisions\/731119"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/media\/729925"}],"wp:attachment":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/media?parent=729205"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/categories?post=729205"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/tags?post=729205"},{"taxonomy":"msr-research-area","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/research-area?post=729205"},{"taxonomy":"msr-region","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-region?post=729205"},{"taxonomy":"msr-event-type","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-event-type?post=729205"},{"taxonomy":"msr-locale","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-locale?post=729205"},{"taxonomy":"msr-post-option","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-post-option?post=729205"},{"taxonomy":"msr-impact-theme","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-impact-theme?post=729205"},{"taxonomy":"msr-promo-type","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-promo-type?post=729205"},{"taxonomy":"msr-podcast-series","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-podcast-series?post=729205"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}