{"id":1098888,"date":"2024-10-31T12:23:03","date_gmt":"2024-10-31T19:23:03","guid":{"rendered":"https:\/\/www.microsoft.com\/en-us\/research\/?p=1098888"},"modified":"2024-10-31T16:23:30","modified_gmt":"2024-10-31T23:23:30","slug":"introducing-drift-search-combining-global-and-local-search-methods-to-improve-quality-and-efficiency","status":"publish","type":"post","link":"https:\/\/www.microsoft.com\/en-us\/research\/blog\/introducing-drift-search-combining-global-and-local-search-methods-to-improve-quality-and-efficiency\/","title":{"rendered":"Introducing DRIFT Search: Combining global and local search methods to improve quality and efficiency"},"content":{"rendered":"\n<figure class=\"wp-block-image size-full\"><img loading=\"lazy\" decoding=\"async\" width=\"1400\" height=\"788\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/10\/DRIFT-Search-GraphRAG-BlogHeroFeature-1400x788-1.jpg\" alt=\"Three icons that represent local and global search and GraphRAG. These icons sit on a blue to pink gradient.\" class=\"wp-image-1098951\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/10\/DRIFT-Search-GraphRAG-BlogHeroFeature-1400x788-1.jpg 1400w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/10\/DRIFT-Search-GraphRAG-BlogHeroFeature-1400x788-1-300x169.jpg 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/10\/DRIFT-Search-GraphRAG-BlogHeroFeature-1400x788-1-1024x576.jpg 1024w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/10\/DRIFT-Search-GraphRAG-BlogHeroFeature-1400x788-1-768x432.jpg 768w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/10\/DRIFT-Search-GraphRAG-BlogHeroFeature-1400x788-1-1066x600.jpg 1066w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/10\/DRIFT-Search-GraphRAG-BlogHeroFeature-1400x788-1-655x368.jpg 655w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/10\/DRIFT-Search-GraphRAG-BlogHeroFeature-1400x788-1-240x135.jpg 240w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/10\/DRIFT-Search-GraphRAG-BlogHeroFeature-1400x788-1-640x360.jpg 640w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/10\/DRIFT-Search-GraphRAG-BlogHeroFeature-1400x788-1-960x540.jpg 960w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/10\/DRIFT-Search-GraphRAG-BlogHeroFeature-1400x788-1-1280x720.jpg 1280w\" sizes=\"auto, (max-width: 1400px) 100vw, 1400px\" \/><\/figure>\n\n\n\n<p>GraphRAG is a technique that uses large language models (LLMs) to create knowledge graphs and summaries from unstructured text documents and leverages them to improve retrieval-augmented generation (RAG) operations on private datasets. It offers comprehensive global overviews of large, private troves of unstructured text documents while also enabling exploration of detailed, localized information. By using LLMs to create comprehensive knowledge graphs that connect and describe entities and relationships contained in those documents, GraphRAG leverages semantic structuring of the data to generate responses to a wide variety of complex user queries. <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"http:\/\/uncharted.software\/\" target=\"_blank\" rel=\"noopener noreferrer\">Uncharted<span class=\"sr-only\"> (opens in new tab)<\/span><\/a>, one of Microsoft\u2019s research collaborators, has recently been expanding the frontiers of this technology by developing a new approach to processing local queries: DRIFT search (Dynamic Reasoning and Inference with Flexible Traversal). This approach builds upon Microsoft\u2019s GraphRAG technique, combining characteristics of both global and local search to generate detailed responses in a method that balances computational costs with quality outcomes.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\" id=\"how-graphrag-works\">How GraphRAG works<\/h2>\n\n\n\n<p>GraphRAG has two primary components, an indexing engine and a query engine.<\/p>\n\n\n\n<p>The indexing engine breaks down documents into smaller chunks, converting them into a knowledge graph with entities and relationships. It then identifies communities within the graph and generates summaries\u2014or &#8220;community reports&#8221;\u2014that represent the global data structure.&nbsp;<\/p>\n\n\n\n<p>The query engine utilizes LLMs to build graph indexes over unstructured text and query them in two primary modes:&nbsp;<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Global search<\/strong> handles queries that span the entire dataset. This mode synthesizes information from diverse underlying sources to answer questions that require a broad understanding of the whole corpus. For example, in a dataset about tech company research efforts, a global query could be: &#8220;What trends in AI research have emerged over the past five years across multiple organizations?&#8221; While effective for connecting scattered information, global search can be resource intensive.&nbsp;<\/li>\n\n\n\n<li><strong>Local search<\/strong> optimizes for targeted queries, drawing from a smaller subset of documents that closely match the user&#8217;s input. This mode works best when the answer lies within a small number of text units. E.g. a query asking: \u201cWhat new features and integrations did Microsoft\u2019s Cosmos DB team release on October 4th?\u201d <\/li>\n<\/ul>\n\n\n\n<p>The creation of these summaries often involves a human in the loop (HITL), as user input shapes how information is summarized (e.g., what kinds of entities and relationships are extracted). To index documents using GraphRAG, a clear description of the intended user persona (as defined in the indexing phase) is needed, as it influences how nodes, edges, and community reports are structured.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\" id=\"introducing-drift-search\">Introducing DRIFT Search<\/h2>\n\n\n\n<p>DRIFT Search introduces a new approach to local search queries by including community information in the search process. This greatly expands the breadth of the query\u2019s starting point and leads to retrieval and usage of a far higher variety of facts in the final answer. This addition expands the GraphRAG query engine by providing a more comprehensive option for local search, which uses community insights to refine a query into detailed follow-up questions. These follow-ups allow DRIFT to handle queries that may not fully align with the original <a href=\"https:\/\/www.microsoft.com\/en-us\/research\/blog\/graphrag-auto-tuning-provides-rapid-adaptation-to-new-domains\/\" target=\"_blank\" rel=\"noreferrer noopener\">extraction templates<\/a> defined by the user at index time.<\/p>\n\n\n\n<figure class=\"wp-block-table\"><table><thead><tr><th>Answer details<\/th><th>Drift (DS_Default)<\/th><th>Local (LS)<\/th><\/tr><\/thead><tbody><tr><td>Supply Chain<\/td><td>Traced back to cinnamon in Ecuador and Sri Lanka<br>[Redacted Brand] and [Redacted Brand] Brands Impacted<br>Products sold at [Redacted Brand] and [Redacted Brand]<\/td><td>Plants in Ecuador<\/td><\/tr><tr><td>Contamination Levels<\/td><td>2000 times higher than FDA max<\/td><td>Blood lead levels ranging from 4 to 29 micrograms per deciliter<\/td><\/tr><tr><td>Actions<\/td><td>Recalls and health advisories<br>Investigating plant in Ecuador<br>Issued warnings to retailers<\/td><td>Recalls and health advisories<\/td><\/tr><\/tbody><\/table><figcaption class=\"wp-element-caption\"><strong>Table 1<\/strong>: An example of summarized responses from two search techniques (DRIFT and Local Search) on a dataset of AP News articles to the query<strong>: \u201c<\/strong>Describe what actions are being taken by the U.S. Food and Drug Administration and the Centers for Disease Control and Prevention to address the lead contamination in apple cinnamon fruit puree and applesauce pouches in the United States during November 2023\u201d. As shown in the table, DRIFT search was able to surface details not immediately available with the two other approaches.<\/figcaption><\/figure>\n\n\n\n\t<div class=\"border-bottom border-top border-gray-300 mt-5 mb-5 msr-promo text-center text-md-left alignwide\" data-bi-aN=\"promo\" data-bi-id=\"999693\">\n\t\t\n\n\t\t<p class=\"msr-promo__label text-gray-800 text-center text-uppercase\">\n\t\t<span class=\"px-4 bg-white display-inline-block font-weight-semibold small\">Spotlight: Event Series<\/span>\n\t<\/p>\n\t\n\t<div class=\"row pt-3 pb-4 align-items-center\">\n\t\t\t\t\t\t<div class=\"msr-promo__media col-12 col-md-5\">\n\t\t\t\t<a class=\"bg-gray-300 display-block\" href=\"https:\/\/www.microsoft.com\/en-us\/research\/event\/microsoft-research-forum\/past-episodes\/?OCID=msr_researchforum_MCR_Blog_Promo\" aria-label=\"Microsoft Research Forum\" data-bi-cN=\"Microsoft Research Forum\" target=\"_blank\">\n\t\t\t\t\t<img decoding=\"async\" class=\"w-100 display-block\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2025\/05\/Research-Forum-hero_1400x788.jpg\" alt=\"Research Forum | abstract background with colorful hexagons\" \/>\n\t\t\t\t<\/a>\n\t\t\t<\/div>\n\t\t\t\n\t\t\t<div class=\"msr-promo__content p-3 px-5 col-12 col-md\">\n\n\t\t\t\t\t\t\t\t\t<h2 class=\"h4\">Microsoft Research Forum<\/h2>\n\t\t\t\t\n\t\t\t\t\t\t\t\t<p id=\"microsoft-research-forum\" class=\"large\">Join us for a continuous exchange of ideas about research in the era of general AI. Watch the latest episodes on demand.<\/p>\n\t\t\t\t\n\t\t\t\t\t\t\t\t<div class=\"wp-block-buttons justify-content-center justify-content-md-start\">\n\t\t\t\t\t<div class=\"wp-block-button\">\n\t\t\t\t\t\t<a href=\"https:\/\/www.microsoft.com\/en-us\/research\/event\/microsoft-research-forum\/past-episodes\/?OCID=msr_researchforum_MCR_Blog_Promo\" aria-describedby=\"microsoft-research-forum\" class=\"btn btn-brand glyph-append glyph-append-chevron-right\" data-bi-cN=\"Microsoft Research Forum\" target=\"_blank\">\n\t\t\t\t\t\t\tWatch on-demand\t\t\t\t\t\t<\/a>\n\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t\t\t<\/div><!--\/.msr-promo__content-->\n\t<\/div><!--\/.msr-promo__inner-wrap-->\n\t<\/div><!--\/.msr-promo-->\n\t\n\n\n<h2 class=\"wp-block-heading\" id=\"drift-search-a-step-by-step-process\">DRIFT Search: A step-by-step process&nbsp;<\/h2>\n\n\n\n<ol class=\"wp-block-list\">\n<li><strong>Primer:<\/strong> When a user submits a query, DRIFT compares it to the top K most semantically relevant community reports. This generates an initial answer along with several follow-up questions, which act as a lighter version of global search. To do this, we expand the query using Hypothetical Document Embeddings (<span data-contrast=\"auto\" xml:lang=\"EN-US\" lang=\"EN-US\" class=\"TextRun EmptyTextRun SCXW48893886 BCX8\" style=\"-webkit-user-drag: none; -webkit-tap-highlight-color: transparent; margin: 0px; padding: 0px; user-select: text; font-size: 11pt; line-height: 19.7625px; font-family: Aptos, Aptos_EmbeddedFont, Aptos_MSFontService, sans-serif; font-variant-ligatures: none !important;\"><\/span>HyDE), to increase sensitivity (recall), embed the query, look up the query against all community reports, select the top K and then use the top K to try to answer the query. The aim is to leverage high-level abstractions to guide further exploration.<\/li>\n\n\n\n<li><strong>Follow-Up:<\/strong> With the primer in place, DRIFT executes each follow-up using a local search variant. This yields additional intermediate answers and follow-up questions, creating a loop of refinement that continues until the search engine meets its termination criteria, which is currently configured for two iterations (further research will investigate reward functions to guide terminations). This phase represents a globally informed query refinement. Using global data structures, DRIFT navigates toward specific, relevant information within the knowledge graph even when the initial query diverges from the indexing persona. This follow-up process enables DRIFT to adjust its approach based on emerging information.&nbsp;<\/li>\n\n\n\n<li><strong>Output Hierarchy:<\/strong> The final output is a hierarchy of questions and answers ranked on their relevance to the original query. This hierarchical structure can be customized to fit specific user needs. During benchmark testing, a naive map-reduce approach aggregated all intermediate answers, with each answer weighted equally.&nbsp;<\/li>\n<\/ol>\n\n\n\n<figure class=\"wp-block-image size-full is-resized\"><img loading=\"lazy\" decoding=\"async\" width=\"6585\" height=\"1091\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/10\/414141-original-w-labels.png\" alt=\"An image that shows a hierarchical tree with each node represented as a pie chart of weighting. \" class=\"wp-image-1099287\" style=\"width:900px;height:auto\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/10\/414141-original-w-labels.png 6585w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/10\/414141-original-w-labels-300x50.png 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/10\/414141-original-w-labels-1024x170.png 1024w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/10\/414141-original-w-labels-768x127.png 768w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/10\/414141-original-w-labels-1536x254.png 1536w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/10\/414141-original-w-labels-2048x339.png 2048w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/10\/414141-original-w-labels-240x40.png 240w\" sizes=\"auto, (max-width: 6585px) 100vw, 6585px\" \/><figcaption class=\"wp-element-caption\">Figure 1. An entire DRIFT search hierarchy highlighting the three core phases of the DRIFT search process. <strong>A (Primer):<\/strong> DRIFT compares the user&#8217;s query with the top <em>K<\/em> most semantically relevant community reports, generating a broad initial answer and follow-up questions to steer further exploration. <strong>B (Follow-Up):<\/strong> DRIFT uses local search to refine queries, producing additional intermediate answers and follow-up questions that enhance specificity, guiding the engine towards context-rich information. A glyph on each node in the diagram shows the confidence the algorithm has to continue the query expansion step.&nbsp; <strong>C (Output Hierarchy):<\/strong> The final output is a hierarchical structure of questions and answers ranked by relevance, reflecting a balanced mix of global insights and local refinements, making the results adaptable and comprehensive.<\/figcaption><\/figure>\n\n\n\n<h3 class=\"wp-block-heading\" id=\"why-drift-search-is-effective\">Why DRIFT search is effective<\/h3>\n\n\n\n<p>DRIFT search excels by dynamically combining global insights with local refinement, enabling navigation from high-level summaries down to original text chunks within the knowledge graph. This layered approach ensures that detailed, context-rich information is preserved even when the initial query diverges from the persona used during indexing. By decomposing broad questions into fine-grained follow-ups, DRIFT captures granular details and adjusts based on the emerging context, making it adaptable to diverse query types. This makes it particularly effective when handling queries that require both breadth and depth without losing specific details.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\" id=\"benchmarking-drift-search\">Benchmarking DRIFT search<\/h3>\n\n\n\n<p>As shown, we tested the effectiveness of DRIFT search by performing a comparative analysis across a variety of use cases against GraphRAG local search and a highly tuned variant of semantic search methods. The analysis evaluated each method&#8217;s performance based on key metrics such as:&nbsp;&nbsp;<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Comprehensiveness<\/strong>: Does the response answer all aspects of the question?<\/li>\n\n\n\n<li><strong>Diversity of responses<\/strong>: Does the response provide different perspectives and insights on the question?<\/li>\n<\/ul>\n\n\n\n<p>In our results, DRIFT search provided significantly better results on both comprehensiveness and diversity in the metrics.\u00a0We set up an experiment where we ingested 5K+ news articles from the Associated Press and ingested those articles using GraphRAG.\u00a0After ingestion, we generated 50 \u201clocal\u201d questions on this dataset and used both DRIFT and Local Search to generate answers for each of these questions.\u00a0These \u201clocal\u201d questions were questions that target specific details in the dataset that could be attributed to a small number of text units containing the answer.\u00a0These answers were then used with an LLM judge to score for comprehensiveness and diversity.<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>On comprehensiveness, DRIFT search outperformed Local Search 78% of the time.<\/li>\n\n\n\n<li>On diversity, DRIFT search outperformed Local Search 81% of the time.<\/li>\n<\/ul>\n\n\n\n<h2 class=\"wp-block-heading\" id=\"availability\">Availability<\/h2>\n\n\n\n<p>DRIFT search is available now on the <a class=\"msr-external-link glyph-append glyph-append-open-in-new-tab glyph-append-xsmall\" href=\"https:\/\/github.com\/microsoft\/graphrag\" target=\"_blank\" rel=\"noopener noreferrer\">GraphRAG GitHub<span class=\"sr-only\"> (opens in new tab)<\/span><\/a>.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\" id=\"future-research-directions\">Future research directions<\/h2>\n\n\n\n<p>A future version of DRIFT will incorporate an improved version of Global Search that will allow it to more directly address questions currently serviced best by global search.&nbsp;The hope is to then move towards a single query interface that can service questions of both local and global varieties.&nbsp;This work will further evolve DRIFT&#8217;s termination logic, potentially through a reward model that balances novel information with redundancy.&nbsp;Additionally, executing follow-up queries using either global or local search modes could improve efficiency. Some queries require broader data access, which can be achieved by leveraging a query router and a lite-<strong>global search<\/strong> variant that uses fewer community reports, tokens, and overall resources.<\/p>\n\n\n\n<p>DRIFT search is the first of several major optimizations to GraphRAG that are being explored.&nbsp; It shows how a global index can even benefit local queries.&nbsp;In our future work, we plan to explore more approaches to bring greater efficiency to the system by leveraging the knowledge graph that GraphRAG creates.<\/p>\n","protected":false},"excerpt":{"rendered":"<p>GraphRAG leverages semantic structuring of data to generate responses to complex user queries. A collaboration with Uncharted expands the frontiers of this technology, developing a new approach to processing local queries: DRIFT search.<\/p>\n","protected":false},"author":42735,"featured_media":1098951,"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":null,"msr_hide_image_in_river":null,"footnotes":""},"categories":[1],"tags":[],"research-area":[13556],"msr-region":[],"msr-event-type":[],"msr-locale":[268875],"msr-post-option":[269148,243984,269142],"msr-impact-theme":[],"msr-promo-type":[],"msr-podcast-series":[],"class_list":["post-1098888","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-research-blog","msr-research-area-artificial-intelligence","msr-locale-en_us","msr-post-option-approved-for-river","msr-post-option-blog-homepage-featured","msr-post-option-include-in-river"],"msr_event_details":{"start":"","end":"","location":""},"podcast_url":"","podcast_episode":"","msr_research_lab":[],"msr_impact_theme":[],"related-publications":[],"related-downloads":[],"related-videos":[],"related-academic-programs":[],"related-groups":[],"related-projects":[1027041],"related-events":[],"related-researchers":[{"type":"guest","value":"julian-whiting","user_id":"1098894","display_name":"Julian Whiting","author_link":"Julian Whiting","is_active":true,"last_first":"Whiting, Julian","people_section":0,"alias":"julian-whiting"},{"type":"guest","value":"zachary-hills","user_id":"1099242","display_name":"Zachary Hills","author_link":"Zachary Hills","is_active":true,"last_first":"Hills, Zachary","people_section":0,"alias":"zachary-hills"},{"type":"user_nicename","value":"Alonso Guevara Fern&aacute;ndez","user_id":42522,"display_name":"Alonso Guevara Fern&aacute;ndez","author_link":"<a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/alonsog\/\" aria-label=\"Visit the profile page for Alonso Guevara Fern&aacute;ndez\">Alonso Guevara Fern&aacute;ndez<\/a>","is_active":false,"last_first":"Guevara Fern\u00e1ndez, Alonso","people_section":0,"alias":"alonsog"},{"type":"user_nicename","value":"Ha Trinh","user_id":43245,"display_name":"Ha Trinh","author_link":"<a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/trinhha\/\" aria-label=\"Visit the profile page for Ha Trinh\">Ha Trinh<\/a>","is_active":false,"last_first":"Trinh, Ha","people_section":0,"alias":"trinhha"},{"type":"guest","value":"adam-bradley","user_id":"1099251","display_name":"Adam Bradley","author_link":"Adam Bradley","is_active":true,"last_first":"Bradley, Adam","people_section":0,"alias":"adam-bradley"},{"type":"user_nicename","value":"Jonathan Larson","user_id":32385,"display_name":"Jonathan Larson","author_link":"<a href=\"https:\/\/www.microsoft.com\/en-us\/research\/people\/jolarso\/\" aria-label=\"Visit the profile page for Jonathan Larson\">Jonathan Larson<\/a>","is_active":false,"last_first":"Larson, Jonathan","people_section":0,"alias":"jolarso"}],"msr_type":"Post","featured_image_thumbnail":"<img width=\"960\" height=\"540\" src=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/10\/DRIFT-Search-GraphRAG-BlogHeroFeature-1400x788-1-960x540.jpg\" class=\"img-object-cover\" alt=\"Three icons that represent (left to right) search, GraphRAG, and the globe.\" decoding=\"async\" loading=\"lazy\" srcset=\"https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/10\/DRIFT-Search-GraphRAG-BlogHeroFeature-1400x788-1-960x540.jpg 960w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/10\/DRIFT-Search-GraphRAG-BlogHeroFeature-1400x788-1-300x169.jpg 300w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/10\/DRIFT-Search-GraphRAG-BlogHeroFeature-1400x788-1-1024x576.jpg 1024w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/10\/DRIFT-Search-GraphRAG-BlogHeroFeature-1400x788-1-768x432.jpg 768w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/10\/DRIFT-Search-GraphRAG-BlogHeroFeature-1400x788-1-1066x600.jpg 1066w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/10\/DRIFT-Search-GraphRAG-BlogHeroFeature-1400x788-1-655x368.jpg 655w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/10\/DRIFT-Search-GraphRAG-BlogHeroFeature-1400x788-1-240x135.jpg 240w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/10\/DRIFT-Search-GraphRAG-BlogHeroFeature-1400x788-1-640x360.jpg 640w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/10\/DRIFT-Search-GraphRAG-BlogHeroFeature-1400x788-1-1280x720.jpg 1280w, https:\/\/www.microsoft.com\/en-us\/research\/wp-content\/uploads\/2024\/10\/DRIFT-Search-GraphRAG-BlogHeroFeature-1400x788-1.jpg 1400w\" sizes=\"auto, (max-width: 960px) 100vw, 960px\" \/>","byline":"","formattedDate":"October 31, 2024","formattedExcerpt":"GraphRAG leverages semantic structuring of data to generate responses to complex user queries. A collaboration with Uncharted expands the frontiers of this technology, developing a new approach to processing local queries: DRIFT search.","locale":{"slug":"en_us","name":"English","native":"","english":"English"},"_links":{"self":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/posts\/1098888","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\/42735"}],"replies":[{"embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/comments?post=1098888"}],"version-history":[{"count":18,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/posts\/1098888\/revisions"}],"predecessor-version":[{"id":1099455,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/posts\/1098888\/revisions\/1099455"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/media\/1098951"}],"wp:attachment":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/media?parent=1098888"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/categories?post=1098888"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/tags?post=1098888"},{"taxonomy":"msr-research-area","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/research-area?post=1098888"},{"taxonomy":"msr-region","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-region?post=1098888"},{"taxonomy":"msr-event-type","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-event-type?post=1098888"},{"taxonomy":"msr-locale","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-locale?post=1098888"},{"taxonomy":"msr-post-option","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-post-option?post=1098888"},{"taxonomy":"msr-impact-theme","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-impact-theme?post=1098888"},{"taxonomy":"msr-promo-type","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-promo-type?post=1098888"},{"taxonomy":"msr-podcast-series","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-podcast-series?post=1098888"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}