Abstract: Answers to search queries are identified and highlighted in search result snippets for documents containing the answers. When a user submits a search query to a search engine, the search engine may query a search index to identify a set of documents as relevant search results for the search query. Additionally, an answer data store may be queried to determine if there is a short answer to the search query. If an answer is found, the answer may be used when generating search result snippets. As such, the search result snippets for documents containing the answer may include the answer. The answer may also be highlighted in the search result snippets.

Abstract: Techniques are provided for determining relevant information from a document based on document structure. A document is selected and structural elements within the document having a dominance relationship are determined. A first location within the document is selected. The structural element surrounding the first location is determined and the surrounding and non-surrounding structural elements are characterized. Additional documents are associated with the first location in the surrounding structural element based on the surrounding structural element characterization and the non-surrounding structural element characterization. Techniques for dynamically determining annotations for images based on document structure are also provided.

Abstract: A structural natural language index is created by segmenting documents within a repository into text portions and extracting named entity, co-reference, lexical entries, structural-semantic relationships, speaker attribution and meronymic derived features. A constituent structure is determined that contains the constituent elements and ordering information sufficient to reconstruct the text portion. A functional structure of the text portions is determined. A set of characterizing predicative triples are formed from the functional structure by applying linearization transfer rules. The constituent structure, the characterizing predicative triples and the derived features are combined to form a canonical form of the text portion. Each canonical form is added to the structural natural language index. A retrieved question is classified to determine question type and a corresponding canonical form for the question is generated.

Abstract: Techniques are provided for dynamically connecting and transforming user interface events between user and generic devices based on physical keys. A user and a user device are determined. The user device is authenticated based on a login sequence, smartcard or the like. A physical key associated with the user is physically associated with a generic device by insertion, wireless link, proximity, RFID detection and the like. The dynamic association is communicated over a first communications channel. A user device confirms the dynamic association over a second communications channel. The dynamically associated generic device communicates user interface events with the dynamic user interface proxy. The user device also communicates user interface events with the dynamic user interface proxy. The dynamic user interface proxy receives, optionally transforms and routes the user interface events associated with the user device and the generic device.

Abstract: A physical key is used to propose an association between a guest device and user information and services. Contact-based or contact-less connectors are used to establish the proposed association between the physical key and the guest device. The proposed association then communicated to the dynamic authentication system over a first communication channel. The dynamic authentication system determines a user confirmation over a second communications channel based on a user device and previously determined associations between users, user devices and the physical key. The guest device is then authenticated for access to information and/or services associated with the user. The information retrieved from and/or transmitted to the user's personal information repository is optionally protected using various transformations.

Abstract: Techniques are provided to determine service data features from an archive of web service transactions. Data features for functionally identical classes of service are determined. Differentiating data feature patterns uniquely identifying each service within the class are learned using machine learning, clustering, statistical analysis and the like. A service map associating services with the differentiating patterns is determined. The service map contains data feature patterns that differentiate among otherwise functionally identical services. The data features are optionally associated with past usage, objective and subjective service quality measurements and the like. The data features of the received service requests are compared to differentiating patterns in the service map. The service associated with the differentiating patterns matching the data features of the service request is selected.

Abstract: Techniques are provided for synthesizing speech using discourse function level prosodic features. An output text is determined. The discourse functions within the text are determined based on a theory of discourse analysis such as the Unified Linguistic Discourse Model. The salient prosodic features associated with the discourse functions are identified using a predictive model of discourse functions or some other model of salient prosodic features. The discourse functions are transformed into synthesized speech. Discourse function level prosodic feature adjustments are determined and applied to the synthesized speech is output.

Abstract: Techniques are provided for determining and using interaction models. Discourse functions, prosodic features and turn information are determined from the speech information in a training corpus. Statistics, decision trees, rules and/or various other methods are used to determine a predictive interaction model based on the discourse functions, the prosodic features and the turn information. Predictive interaction models are optionally determined for individual users, genres, languages and/or other characteristics of the speech information. The predictive interaction model is useable to predict turns in a dialogue based on the discourse functions and prosodic features identified in the speech information. Speech information is presented and/or received based on the predictive interaction model.

Abstract: Techniques are provided for determining predictive models of discourse functions based on prosodic features of natural language speech. Inter and intra sentential discourse functions in a training corpus of natural language speech utterances are determined. The discourse functions are clustered. The exemplary prosodic features associated with each type of discourse function are determined. Machine learning, observation and the like are used to determine a subset of prosodic features associated with each type of discourse function useful in predicting the likelihood of each type of discourse function.

Abstract: Techniques are provided for resolving ambiguity in natural language speech. Speech is recognized using automatic speech recognition. A theory of discourse analysis is determined and at least one set of candidate discourse functions is determined based on the theory of discourse analysis. Prosodic features in the speech and a correlation between the prosodic features and the discourse functions is determined. The sets of candidate discourse functions are ranked based on the prosodic features in the speech information and a correlation to the prosodic features expected for the determined discourse functions. Ambiguity is resolved between sets of candidate discourse functions based on the rank information.

Abstract: Techniques are provided for determining collaborative notes and automatically recognizing speech, handwriting and other type of information. Domain and optional actor/speaker information associated with the support information is determined. An initial automatic speech recognition model is determined based on the domain and/or actor information. The domain and/or actor/speaker language model is used to recognize text in the speech information associated with the support information. Presentation support information such as slides, speaker notes and the like are determined. The semantic overlap between the support information and the salient non-function words in the recognized text and collaborative user feedback information are used to determine relevancy scores for the recognized text. Grammaticality, well formedness, self referential integrity and other features are used to determine correctness scores.

Abstract: Techniques are provided for segmenting text into categorized discourse constituents and attaching discourse constituents into a structural representation of discourse. Techniques for determining hybrid structural and non-structural summaries of a text are also provided. A text is segmented based on a theory of discourse analysis into at least a main discourse constituent containing spatio-temporal information about a single event in a possible world view. The discourse constituents are then inserted into a structural representation of discourse. Non-structural techniques are used to determine relevance scores and important discourse constituents are determined. Relevance scores are percolated through the structural representation of discourse to determine supporting preceding discourse constituents that preserve grammaticality. A hybrid text summary is then determined based on the structural representation of the discourse and relevance scores.