Microsoft Technology Licensing, LLC

Junaid Ahmed

Arnold Overwijk

Walter Sun

Li Xiong

Sasi Parthasarathy

Ziliu Li

Chenyan Xiong

Xiao Lin Zhang

Yipeng Li

Chuan Hu

Serge Smirnov

Peter Bailey

Abhishek Kumar Singh

Yan Ke

Nikhil Dandekar

Xiaodong Fan

Ting Cai

Qiaolin Mao

Siarhei Alonichau

Jingfeng Li

Xin Chen

Daniel Fernando Campos

Yatharth Saraf

Charumathi Lakshmanan

Alexander Owen Coll

Waleed Malik

Jue Wang

Lin Guo

Kumar Hemachandra Chellapilla

Yinzhe Yu

Jay Goyal

Vinay Deshpande

Fabrice Canel

Thomas Francis McElroy

Aliaksei Bondarionok

Paul Alexander Dow

Yajie Siamwalla

Jia Zhu

Arindam Mitra

Simon Julian Powers

Vishnu Challam

Thomas Arthur Ledbetter

Xiao Liu

Brahm Kiran Singh

Kwok Fung Tang

Jiaming Guo

Jiguang Shen

Paolo Codato

Ye Li

Yi Zou

Aspects of the present disclosure relate to systems and methods for performing targeted searching based on a user profile. In examples, a user profile including a user embedding may be retrieved based on the receipt of a user indication. The user embedding may be created based on one or more user interest. A plurality of document embeddings may be identified based on the user embedding, where each document embedding of the plurality of document embeddings is determined to be within a first distance of the user embedding. In examples, a ranking for each document embedding of the plurality of document embeddings may be generated, where the ranking for each document embedding of the plurality of document embeddings is based on the user embedding. At least one document may be recommend based on a ranking associated with a document embedding.

Performing targeted searching based on a user profile

Described herein are technologies related to tokenizing alphanumeric text through use of a tokenization algorithm that is at least partially implemented as a finite state machine. The tokenization algorithm is configured to output numeric identifiers that represent tokens or sub-tokens in the alphanumeric text.

Tokenizing alphanumeric text through use of finite state machines

A system for extracting key phrase candidates from a corpus of documents, including a processor, a memory, and a program executing on the processor. The system is configured to run a key phrase model to extract one or more key phrase candidates from each document in the corpus and convert each extracted key phrase candidate into a feature vector. The key phrase model also filters the feature vectors to remove duplicates using a classifier that was trained on a set of key phrase pairs with manual labels indicating whether two key phrases are duplicates of each other, to produce remaining key phrase candidates. The system uses the remaining key phrase candidates in a computer-implemented application.

Extracting key phrase candidates from documents and producing topical authority ranking

Document embedding vectors for each document of a corpus may be generated by combining embedding vectors for document subparts, thereby yielding a final embedding vector for the document. A machine learning model is trained using a query corpus and the document corpus, where the model generates a ranking score for a given (query, document) pair. During training, rankings scores are generated using the model, such that the training dataset is further refined using the generated ranking scores. For example, top documents and a negative document may be determined for a given query and subsequently used as training data. Multiple negative documents may therefore be determined for a given query. A negative document for a given query may be determined from the negative documents using noise-contrastive estimation. Such determined negative documents may be evaluated using a loss function during model training, thereby yielding a more robust model for search processing.

Document body vectorization and noise-contrastive training

To provide automated categorization of structured textual content individual nodes of textual content, from a document object model encapsulation of the structured textual content, have a multidimensional vector associated with them, where the values of the various dimensions of the multidimensional vector are based on the textual content in the corresponding node, the visual features applied or associated with the textual content of the corresponding node, and positional information of the textual content of the corresponding node. The multidimensional vectors are input to a neighbor-imbuing neural network. The enhanced multidimensional vectors output by the neighbor-imbuing neural network are then be provided to a categorization neural network. The resulting output can be in the form of multidimensional vectors whose dimensionality is proportional to categories into which the structured textual content is to be categorized. A weighted merge takes into account multiple nodes that are grouped together.

Automated structured textual content categorization accuracy with neural networks

Described herein are technologies related to inferring information about a webpage based upon semantics of a uniform resource location (URL) of the webpage. The URL is tokenized to create a sequence of tokens. An embedding for the URL is generated based upon the sequence of tokens, wherein the embedding is representative of semantics of the URL. Based upon the embedding for the URL, information about the webpage pointed to by the URL is inferred, the webpage is retrieved, and information is extracted from the webpage based upon the information inferred about the webpage.

Inferring information about a webpage based upon a uniform resource locator of the webpage

A system for extracting a key phrase from a document includes a neural key phrase extraction model ('BLING-KPE') having a first layer to extract a word sequence from the document, a second layer to represent each word in the word sequence by ELMo embedding, position embedding, and visual features, and a third layer to concatenate the ELMo embedding, the position embedding, and the visual features to produce hybrid word embeddings. A convolutional transformer models the hybrid word embeddings to n-gram embeddings, and a feedforward layer converts the n-gram embeddings into a probability distribution over a set of n-grams and calculates a key phrase score of each n-gram. The neural key phrase extraction model is trained on annotated data based on a labeled loss function to compute cross entropy loss of the key phrase score of each n-gram as compared with a label from the annotated dataset.

Keyphase extraction beyond language modeling

Technologies pertaining to electronic document understanding are described herein. A document is received, wherein the document includes a section of a type. An image of the document is generated, and a candidate region is identified in the image of the document, wherein the candidate region encompasses the section. A label is assigned to the candidate region based upon text of the section, wherein the label identifies the type of the section. An electronic document understanding task is performed based upon the label assigned to the candidate region.

Constructing a computer-implemented semantic document

A technique produces a graph data structure based on at least partially unstructured information dispersed over web documents. The technique involves applying a machine-trained model to a set of documents (or, more generally 'document units') to identify topics in the documents. The technique then generates count information by counting the occurrences of the single topics and co-occurrences of parings of topics in the documents. The technique generates conditional probability information based on the count information. An instance of conditional probability information describes a probability that a first topic will occur, given an appearance of a second topic, and a probability that the second topic will occur, given an appearance of the first topic. The technique then formulates the conditional probability information in a graph data structure. The technique also provides an application system that utilizes the graph data structure to provide any kind of computer-implemented service to a user.

Generating a graph data structure that identifies relationships among topics expressed in web documents

Technologies pertaining to ranking webpages in response to receipt of a query are described. A search engine receives a query and identifies webpages that are germane to the query. The search engine ranks the identified webpages to form a ranked list, wherein a first webpage is positioned in the ranked list based upon a static score assigned to the first webpage. The static score is based upon a weight assigned to a hyperlink in a second webpage, wherein the hyperlink points to the first webpage, and further wherein the weight is based upon a value of a feature of the hyperlink, such as a location of the hyperlink on the second webpage when the second webpage is rendered. Further, the second webpage includes several hyperlinks that point to different webpages, wherein each of the several hyperlinks has a different weight assigned thereto.

Ranking computer-implemented search results based upon static scores assigned to webpages

Keyphrase extraction beyond language modeling

A technique is described herein for allocating tasks within a distributed system. The distributed system includes plural computing devices. Each computing device, in turn, provides a bucket manager and at least one application instance. The bucket managers allocate work to the computing devices in units of buckets, in a manner governed by a set of application-specific initialization parameters. The bucket managers also perform their assignment function in cooperation with a separate storage and notification (S&N) system. The S&N system stores data nodes in a hierarchical data structure that reflect the assignment-related state of the distributed system. The S&N system also notifies the bucket managers upon changes in the data nodes. According to one provision, the bucket managers include logic that reduces expensive context changes when computing devices join and leave the distributed system.

Distributed system that promotes task-machine affinity

Context-based multi-granularity intent discovery, in relation to determining the intent of a user-problem associated with one of a plurality of products, includes identifying each of the plurality of products in a base intent level of a hierarchical intent data structure, creating a first granularity level below the base intent level, and creating a subsequent granularity level. The first granularity level includes groupings of the plurality of user-problems created according to a first common intent feature; the common intent feature is related to a respective one of the plurality of products. The subsequent granularity level includes sub-groupings of the user-problems of a respective grouping with each sub-grouping created according to a second common intent feature. The second common intent feature of each sub-grouping is determined without consideration of the first common intent feature.

Context-based multi-granularity intent discovery

A technique is described herein that receives a set of markup-language documents (e.g., HTML documents). The technique produces a set of object-model representations (VDOMs) based on the set of markup-language documents. For a given markup-language document, the producing operation involves accessing and processing information that is linked to the markup-language document. Each object-model representation also includes visual layout information that specifies a visual appearance of a page that would be produced upon display of the markup-language document by a browser application. The technique then generates transformation logic based on the set of markup-language documents and their corresponding VDOMs. The transformation logic, when applied to a new markup-language document, converts the new markup-language document into a corresponding VDOM without retrieving at least some information identified by links (if any) in the new markup-language document.

Resource-efficient generation of visual layout information associated with network-accessible documents

Growing community of inventors

Bellevue, WA, United States of America

Junaid Ahmed