Microsoft Technology Licensing, LLC

Puneet Agrawal

Vijay Krishnan

Sree Hari Nagaralu

Manish Gupta

Adish Singla

Kishor Chamua

Ankur Khator

Shveta Verma

Sushil Kumar Chordia

Saksham Saxena

Biju Venugopal

Karthik Rg

Akriti

Rishab

Srimukha Paturi

Kumar Subham

Systems and methods for generating autocomplete text using a language model are disclosed. An image and text-prefix may be entered at an input field of a search application. The image is processed to generate an image description. The image description and the text-prefix signals may be used as input at a language model to generate an autocomplete text by the language model. A contextual history may also be included as input to the language model. The autocomplete text is an output by the language model based on the input at the language model. The auto-complete text may be a next-word ghosting.

Ghosting for multimodal dialogs

Systems and methods are directed to providing multi-level rich autosuggestions for multi-intent search queries. The system receives a query prefix entered in a search box and accesses a database of machine-learning derived intent clusters. Based on the query prefix, a top-ranking machine-learning derived intent cluster is identified, and the search user interface is updated to provide a plurality of first level autosuggestions in an autosuggestion box including a smart suggestion. The smart suggestion is a first level suggestion of the top-ranking machine-learning derived intent cluster. The system receives a selection of the smart suggestion and, in response, replaces the plurality of first level autosuggestions in the autosuggestion box with second level autosuggestions. The second level autosuggestions comprise one or more intent categories that are subsets of the smart suggestion, whereby each intent category comprises a plurality of subcategory suggestions.

Multi-level rich autosuggestions for multi-intent search query

Aspects of the present disclosure relate to providing, based on a partial query string, a plurality of autosuggestions that are diverse in nature such that the user is more likely to see the preferred complete query terms and therefore more likely to select one of the preferred suggestions which will increase search efficiency. As described herein, such functionality relates to generating cluster groups of candidate suggestions, each cluster including sub-topics, then performing the search based on a selected cluster or sub-topic. To generate the cluster groups, systems and methods, as described herein, analyze the similarity between candidate suggestions as well as the popularity of generated sub-topics. The cluster groups and sub-topics may be displayed visually, and in certain embodiments the cluster groups are ordered vertically from top to bottom and aligned to the left side of the display, while sub-topics are ordered horizontally from left to right following the cluster label.

Clustering suggestions for partial query auto-completion

Systems, methods, and computer-readable storage media are provided for providing personalized, emotionally connected responses from a digital assistant. The emotionally connected (personalized) responses may include, among other things, a handwritten response. The handwritten response may be provided in an image format in a handwriting script unique to the digital assistant. The emotionally connected responses may be provided in response to an event that triggers emotionally connected/personalized responses such as emotional query inputs, calendar entries, holidays, tasks, and the like.

Emotionally connected responses from a digital assistant

Systems, methods, and computer-readable storage media for identifying queries having a version-intent and presenting search results in accordance with the version-intent are provided. Subject-versions associated with retrievable documents are identified and associated therewith as subject-version tags. When a search query is received, it is determined whether the query has a version-intent indicative of a particular version associated with a subject of the query. Documents are retrieved that satisfy the input query and ranked for presentation based upon the particular subject-version indicated by the version-intent. The ranked documents are then presented. Also presented is a user-manipulatable tool, manipulation of which permits a user to indicate that information concerning a different subject-version is desired. Upon receiving such indication, the search results are changed such that documents having subject-version tags indicative of the newly-indicated version-intent are ranked higher and receive more prominent placement than documents lacking such subject-version tags.

Presenting search results based upon subject-versions

A search engine may allow a user to add and modify geographic constraints interactively through the use of a map. In one example, the user enters a search query, and the search engine determines whether the query is amenable to a geographic constraint. If the query is amenable to a geographic constraint, then the search engine may provide an interactive map through which the user may specify a geographic boundary. The search engine may then uses the geographic boundary as a constraint on the search, and may return documents that relate to the geographic area indicated by the boundary. The user may zoom in or out on the region defined by the boundary, or may move the boundary, thereby interactively focusing, de-focusing, and modifying the search based on geography.

Map-based selection of query component

Methods, computer systems, and computer-readable storage media for utilizing breadcrumb information associated with a Web page to improve a user's search experience on a search engine page and to improve ranking of search results are provided. Breadcrumb information is identified for a plurality of Web pages and is used to create an entity data structure. The entity data structure is used to determine related search items, deep links, and query suggestions on a search engine page. Further, contextual relationships between different breadcrumb elements that are part of a breadcrumb trail associated with a Web page are identified and used to adjust a ranking order of the Web page on a search engine results page.

Growing community of inventors

Hyderabad Telangana, India

Puneet Agrawal