Microsoft Technology Licensing, LLC

Chin-Yew Lin

Yunbo Cao

Xinying Song

Zhiyuan Chen

Ming Zhou

Jianfeng Gao

Ying Li

Li Erran Li

Royi Ronen

Zhaohui Tang

Oron Nir

Brian Burdick

Shuming Shi

Avner Levi

Guihong Cao

Ohad Jassin

Cheng Niu

Jian-Yun Nie

Eylon Ami

Pei Yue

Daniel Nurieli

Shasha Li

Youngin Song

Young-In Song

Liu Jing

Tetsuya Sakai

Systems and methods are disclosed for inferring topics from a file containing both audio and video, for example a multimodal or multimedia file, in order to facilitate video indexing. A set of entities is extracted from the file and linked to produce a graph, and reference information is also obtained for the set of entities. Entities may be drawn, for example, from Wikipedia categories, or other large ontological data sources. Analysis of the graph, using unsupervised learning, permits determining clusters in the graph. Extracting features from the clusters, possibly using supervised learning, provides for selection of topic identifiers. The topic identifiers are then used for indexing the file.

Inferring topics with entity linking and ontological data

Systems, methods, and computer-readable media for dynamically generating text associated with an advertisement are provided. Core text associated with an advertisement is received from an advertiser, as is at least one attribute relevant to the advertiser and/or a user. Based upon the received attribute(s), it is determined whether customization of the core text is desired. If customization is desired, the core text is modified and presented in association with the advertisement. If customization is not desired, the core text is presented in association with the advertisement. In one embodiment, target advertisement placement information may also be utilized to determine whether customization of the core text is desired.

Dynamic generation of advertisement text

Described herein are techniques for extracting data records containing user-generated content from documents. The documents may be processed into document trees in which sub-trees represent the data records of the document. Domain constraints may be used to locate structured portions of the document tree. For example, anchor trees may be located as being sets of sibling sub-trees with similar tag paths that contain the domain constraints. The anchor trees may then be used to determine a record boundary (e.g., the start offset and length) of the data records. Finally, the data records may be extracted based on the anchor trees and the record boundaries.

Domain constraint path based data record extraction

This disclosure describes, in part, techniques for operating a search query user interface to allow seamless creating, editing and/or refining of a search query using various interactive functions. The techniques described herein may display a search query divided into segments. A selection of a segment of the search query may then be received. One or more alternatives to the selected segment may then be presented. Next, a selection of one of the presented alternative may be received. As a result, the search query may be altered using the selected alternative. Furthermore, the techniques described herein allow a user to operate on a search query using query substitution, expansion, association and/or history functions.

Search query user interface

Embodiments for a Mining Data Records based on Anchor Trees (MiBAT) process are disclosed. In accordance with at least one embodiment, the MiBAT process extracts data records containing user-generated content from web documents. The web document is processed into a Document Object Model (DOM) tree in which sub-trees of the DOM tree represent the data records of the web document. Domain constraints are used to locate structured portions of the DOM tree. Anchor trees are then located as being sets of sibling sub-trees which contain the domain constraints. The anchor trees are then used to determine a record boundary (i.e. the start offset and length) of the data records. Finally, the data records are extracted based on the anchor trees and the record boundaries.

Domain constraint based data record extraction

Some implementations disclosed herein provide techniques and arrangements to train a blocking scheme using both labeled data and unlabeled data. For example, training the blocking scheme may include iteratively: learning a conjunction, identifying first matches in the labeled data and the unlabeled data that are uncovered by the conjunction, and identifying second matches in the labeled data and the unlabeled data that are covered by the conjunction. The conjunction learned in each iteration may be combined using a disjunction. A search engine may use the search engine when searching for records that match an entity.

Record linkage based on a trained blocking scheme

One or more techniques and systems are disclosed for generating comparative patterns for use in identifying comparators. A set of comparator pairs is extracted from a first comparative pattern in a pattern database that comprises one or more comparative patterns. Questions are retrieved from a question collection using respective comparator pairs to generate comparative questions. Potential comparative patterns are generated from a combination of the comparator pairs and comparative questions, and the potential comparative patterns are evaluated by determining their reliability, in order to generate second comparative patterns for the pattern database.

Comparative entity mining

Techniques for unsupervised management of a question and answer (QA) forum include labeling of answers for quality purposes, and identification of experts. In a QA thread, a ranking of answers may include an initial labeling of the longest answer in each thread as the best answer. Such a labeling provides an initial point of reference. Then, in an iterative manner answerers are ranked using the labeling. The ranking of answerers allows selection of experts and poor or inexpert answerers. A label update is performed using the experts (and perhaps inexpert answerers) as input. The label update may be used to train a model, which may describe quality of answers in one or more QA threads and an indication of expert and inexpert answerers. The iterative process may be ended upon convergence or upon a maximum number of iterations.

Question and answer forum techniques

A question search system provides a collection of questions having words for use in evaluating the utility of the questions based on a language model. The question search system calculates n-gram probabilities for words within the questions of the collection. The n-gram probability of a word for a sequence of n&#x2212;1 words indicates the probability of that word being next after that sequence in the collection of questions. The n-gram probabilities for the words of the collection represent the language model of the collection. The question search system calculates a language model utility score for each question within a collection that indicates the likelihood that a question is repeatedly asked by users. The question search system derives the language model utility score for a question from the n-gram probabilities of the words within that question.

Determining utility of a question

A method and system for determining the relevance of questions to a queried question based on topics and focuses of the questions is provided. A question search system provides a collection of questions with topics and focuses. Upon receiving a queried question, the question search system identifies a queried topic and queried focus of the queried question. The question search system generates a score indicating the relevance of a question of the collection to the queried question based on a language model of the topic of the question and a language model of the focus of the question.

Searching questions based on topic and focus

A method and system for presenting questions that are relevant to a queried question based on clusters of topics and clusters of focuses of the questions is provided. A question search system provides a collection of questions. Each question of the collection has an associated topic and focus. Upon receiving a queried question, the question search system identifies questions of the collection that may be relevant to the queried question and generates a score or ranking indicating relevance of the identified questions. The question search system clusters the identified questions into topic clusters of questions with similar topics. The question search system may also cluster the questions within each topic cluster into focus clusters of questions with similar focuses.

Clustering question search results based on topic and focus

In a question answering system, the system identifies a type of question input by a user. The system then generates answer summaries that summarize answers to the input question in a format that is determined based on the type of question asked by the user. The answer summaries are output, in the corresponding format, in answer to the input question.

Question type-sensitive answer summarization

A probability distribution for a reference summary of a document is determined. The probability distribution for the reference summary is then used to generate a score for a machine-generated summary of the document.

Growing community of inventors

Beijing, China

Chin-Yew Lin