Microsoft Technology Licensing, LLC

Zaiqing Nie

Ji-Rong Wen

Wei-Ying Ma

Jun Zhu

Yong Cao

Xiaojiang Liu

Shuming Shi

Gang Luo

Yunxiao Ma

Bo Zhang

Ying-Qing Xu

Fei Xing

Ruochi Zhang

Mingjie Zhu

Qiang Wu

Lie Lu

Ying Li

Zhiwei Li

Hsiao-Wuen Hon

Liu Yang

Hua Chen

Bernhard S J Kohlmeier

Pradeep Chilakamarri

Lee Wang

Gregory Akselrod

Yue Liu

Eun Ju Nam

Prashant Thiruvengadachari

Chunyu Yang

Yunhua Hu

Lingzhi Zhao

Zhengdong Lu

Honghua Dai

The automated social networking graph mining and visualization technique described herein mines social connections and allows creation of a social networking graph from general (not necessarily social-application specific) Web pages. The technique uses the distances between a person's/entity's name and related people's/entities names on one or more Web pages to determine connections between people/entities and the strengths of the connections. In one embodiment, the technique lays out these connections, and then clusters them, in a 2-D layout of a social networking graph that represents the Web connection strengths among the related people's or entities' names, by using a force-directed model.

Automated networking graph mining and visualization

Captured data can be transformed and augmented for a particular presentation in a document, such as a note of a notebook application, based on an identified entity for the captured data. The particular presentation of captured data can be provided based on entity detection, extraction, and knowledge base resolution and retrieval. Methods, systems, and services are provided that identify a primary entity of an item input to a notebook application and create an entity object for the primary entity of the item at least from one or more structured representations for content associated with the item. A template for presenting the entity object can be determined according to the primary entity, where the template is selected from a set of templates corresponding to different primary entities such that an arrangement and presentation for one primary entity is different than that of another primary entity.

Augmenting and presenting captured data

Automated social networking graph mining and visualization

Methods and systems for Web-scale entity relationship extraction are usable to build large-scale entity relationship graphs from any data corpora stored on a computer-readable medium or accessible through a network. Such entity relationship graphs may be used to navigate previously undiscoverable relationships among entities within data corpora. Additionally, the entity relationship extraction may be configured to utilize discriminative models to jointly model correlated data found within the selected corpora.

Displaying search results with edges/entity relationships in regions/quadrants on a display device

The techniques described herein automatically generate high precision clusters and high recall clusters for a set of documents having an author with a same or similar name. The high precision clusters and the high recall clusters can then be used in a labeling process so that efficient and accurate author disambiguation is realized.

Author disambiguation

Described is a technology for understanding entities of a webpage, e.g., to label the entities on the webpage. An iterative and bidirectional framework processes a webpage, including a text understanding component (e.g., extended Semi-CRF model) that provides text segmentation features to a structure understanding component (e.g., extended HCRF model). The structure understanding component uses the text segmentation features and visual layout features of the webpage to identify a structure (e.g., labeled block). The text understanding component in turn uses the labeled block to further understand the text. The process continues iteratively until a similarity criterion is met, at which time the entities may be labeled. Also described is the use of multiple mentions of a set of text in the webpage to help in labeling an entity.

Webpage entity extraction through joint understanding of page structures and sentences

Techniques for displaying a relationship graph are described herein. In one example, a search term may be used to obtain a plurality of documents from a network, such as the Internet. A plurality of entities, and relationships between at least some of those entities, may be extracted from the documents. In an example user interface, representations of a plurality of entities may be displayed, such as by shapes (e.g., circles) labeled to identify people or organizations. Edges (e.g., lines) may be used to connect different representations of entities and to thereby indicate a relationship between the connected entities. In a particular example, input from movement of a cursor over an edge may result in display of a description of a relationship between the connected entities. In a further particular example, size of each entity may be related to a number of connections each has with others.

Web-scale entity relationship extraction

A 'Name Disambiguator' provides various techniques for implementing an interactive framework for resolving or disambiguating entity names (associated with objects such as publications) for entity searches where two or more same or similar names may refer to different entities. More specifically, the Name Disambiguator uses a combination of user input and automatic models to address the disambiguation problem. In various embodiments, the Name Disambiguator uses a two part process, including: 1) a global SVM trained from large sets of documents or objects in a simulated interactive mode, and 2) further personalization of local SVM models (associated with individual names or groups of names such as, for example, a group of coauthors) derived from the global SVM model. The result of this process is that large sets of documents or objects are rapidly and accurately condensed or clustered into ordered sets by that are organized by entity names.

Interactive framework for name disambiguation

Web-scale entity relationship extraction that extracts pattern(s) based on an extracted tuple

Described is a summarizing a web entity (e.g., a person, place, product or so forth) based upon the entity's appearance in web documents (e.g., on the order of hundreds of millions or billions of webpages). Webpages are separated into blocks, which are then processed according to various features to filter the number of blocks to further process, and rank the most relevant blocks with respect to the entity that remain. A redundancy removal mechanism removes redundant blocks, leaving a set of remaining blocks that are used to provide a summary of information that is relevant to the entity.

Web-scale entity summarization

A search method uses pseudo-anchor text associated with search objects to improve search performance. The pseudo-anchor text may be extracted in combination with an identifier of the search objects (such as a pseudo-URL) from a digital corpus such as a collection of documents. Pseudo-anchor texts for each object are preferably extracted from candidate anchor blocks using a machine learning based approach. The pseudo-anchor texts are made available for searching and used to help rank the objects in a search result to improve search performance. The method may be used in vertical search of objects such as published articles, products and images that lack explicit URLs and anchor text information.

Pseudo-anchor text extraction

A method and system is provided for determining relevance of an object to a term based on a language model. The relevance system provides records extracted from web pages that relate to the object. To determine the relevance of the object to a term, the relevance system first determines, for each record of the object, a probability of generating that term using a language model of the record of that object. The relevance system then calculates the relevance of the object to the term by combining the probabilities. The relevance system may also weight the probabilities based on the accuracy or reliability of the extracted information for each data source.

Web object retrieval based on a language model

Search results are ranked by applying sub-relevancies within a search result list. This permits search result lists to be further refined into more manageable relevant groupings from a user's perspective. The sub-relevancies or 'group-by' parameters are derived from search result attributes. Attribute values from the attributes are employed in a ranking scheme to group the search results based on attribute value relevancy. The grouped search results can then be displayed to users via a search result page. In one instance users can select which attribute value is used to group the search result list. Ranking processes are based on object ranking algorithms that consider each attribute value as an object type. Some instances provide for search result list group condensing based on relevancy of the attribute values as well. A top-k instance can be employed to limit the search results to control the amount of time required for processing groupings.

Group-by attribute value in search results

Features extracted from network browser pages and/or network search queries are leveraged to facilitate in detecting a user's browsing and/or searching intent. Machine learning classifiers constructed from these features automatically detect a user's online commercial intention (OCI). A user's intention can be commercial or non-commercial, with commercial intentions being informational or transactional. In one instance, an OCI ranking mechanism is employed with a search engine to facilitate in providing search results that are ranked according to a user's intention. This also provides a means to match purchasing advertisements with potential customers who are more than likely ready to make a purchase (transactional stage). Additionally, informational advertisements can be matched to users who are researching a potential purchase (informational stage).

Automatic detection of online commercial intention

A method and system for labeling object information of an information page is provided. A labeling system identifies an object record of an information page based on the labeling of object elements within an object record and labels object elements based on the identification of an object record that contains the object elements. To identify the records and label the elements, the labeling system generates a hierarchical representation of blocks of an information page. The labeling system identifies records and elements within the records by propagating probability-related information of record labels and element labels through the hierarchy of the blocks. The labeling system generates a feature vector for each block to represent the block and calculates a probability of a label for a block being correct based on a score derived from the feature vectors associated with related blocks. The labeling system searches for the labeling of records and elements that has the highest probability of being correct.

Growing community of inventors

Beijing, China

Zaiqing Nie