Abstract: The present invention involves using one or more statistical classifiers in order to perform task classification on natural language inputs. In another embodiment, the statistical classifiers can be used in conjunction with a rule-based classifier to perform task classification.

Abstract: A training module is described for training a conditional random field (CRF) tagging model. The training module trains the tagging model based on an explicitly-labeled training set and an implicitly-labeled training set. The explicitly-labeled training set includes explicit labels that are manually selected via human annotation, while the implicitly-labeled training set includes implicit labels that are generated in an unsupervised manner. In one approach, the training module can train the tagging model by treating the implicit labels as non-binding evidence that has a bearing on values of hidden state sequence variables. In another approach, the training module can treat the implicit labels as binding or hard evidence. A labeling system is also described for providing the implicit labels.

Abstract: A method of generating a confidence measure generator is provided for use in a voice search system, the voice search system including voice search components comprising a speech recognition system, a dialog manager and a search system. The method includes selecting voice search features, from a plurality of the voice search components, to be considered by the confidence measure generator in generating a voice search confidence measure. The method includes training a model, using a computer processor, to generate the voice search confidence measure based on selected voice search features.

Abstract: One or more techniques and/or systems are disclosed for creating an expanded or improved lexicon for use in search-based semantic tagging. A set of first documents can be identified using a set of first lexicon elements as queries, and one or more first document patterns can be extracted from the set of first documents. The document patterns can be used to find one or more second documents in a query log that comprise the document patterns, which are associated with query terms used to return the second documents. The query terms for the second documents can be extracted and used to expand the lexicon. Elements within the lexicon may be weighted based upon relevance to different query domains, for example.

Abstract: A voice search system has a speech recognizer, a search component, and a dialog manager. A confidence measure generator receives speech recognition features from the speech recognizer, search features from the search component, and dialog features from the dialog manager, and calculates an overall confidence measure for voice search results based upon the features received. The invention can be extended to include the generation of additional features, based on those received from the individual components of the voice search system.

Abstract: Architecture that provides the capability to identify which parts (terms and phrases) of a voice query have been covered by predefined phrase templates, and then to concatenate matching phrase templates into a new paraphrased query. A match-drop-continue algorithm is disclosed that progressively masks out the portions (phrases, terms) of the query matched to the phrase templates. Ultimately, the matched phrase templates are accumulated and organized together dynamically into a rephrased version of the original voice query. A user interface is provided that allows the user to confirm/summarize the multiple concepts in a progressive manner.

Abstract: An answering machine detection module is used to determine whether a call recipient is an actual person or an answering machine. The answering machine detection module includes a speech recognizer and a call analysis module. The speech recognizer receives an audible response of the call recipient to a call. The speech recognizer processes the audible response and provides an output indicative of recognized speech. The call analysis module processes the output of the speech recognizer to generate an output indicative of whether the call recipient is a person or an answering machine.

Abstract: This patent application pertains to dynamic search interaction. One example includes an organizational component configured to obtain a search query from a user. The organizational component can also be configured to obtain related search queries. The organizational component can further be configured to organize the related search queries by topic and to estimate a relative likelihood that an intent of the user matches an individual topic. This example also includes an image generation component configured to cause the organized related search queries to be presented on a graphical user interface (GUI) in a manner that reflects the relative likelihood.

Abstract: Project-related data may be aggregated from various data sources, given context, and may be stored in a data repository or organizational knowledge base that may be available to and accessed by others. Documents, emails, contact information, calendar data, social networking data, and any other content that is related to a project may be brought together within a single user interface, irrespective of its data type. A user may organize and understand content, discover relevant information, and act on it without regard to where the information resides or how it was created.

Abstract: An analysis module, when triggered by a synchronization framework when a new data item is added to a project data store, runs a series of analysis feature extractors on the new content. An analysis may be conducted, and features of interest may be extracted from the data item. The analysis utilizes natural language processing, as well as other technologies, to provide an automatic or semi-automatic extraction of information. The extracted features of interest are saved as metadata within the project data store, and are associated with the data item from which it was extracted. The analysis module may be utilized to discover additional information that may be gleaned from content that is already in the project data store.

Abstract: The present invention uses a natural language understanding system that is currently being trained to assist in annotating training data for training that natural language understanding system. Unannotated training data is provided to the system and the system proposes annotations to the training data. The user is offered an opportunity to confirm or correct the proposed annotations, and the system is trained with the corrected or verified annotations.

Abstract: A directory assistance system includes a directory database and a search engine. The search engine is configured to search the directory database for a first set of residential listings based on at least one first search term. A second search term is received that is related to a cohabitant of the listing to be found. At least one search result is selected that satisfies the second search term.

Abstract: Described is a technology by which a maximum entropy model used for classification is trained with a significantly lesser amount of training data than is normally used in training other maximum entropy models, yet provides similar accuracy to the others. The maximum entropy model is initially parameterized with parameter values determined from weights obtained by training a vector space model or an n-gram model. The weights may be scaled into the initial parameter values by determining a scaling factor. Gaussian mean values may also be determined, and used for regularization in training the maximum entropy model. Scaling may also be applied to the Gaussian mean values. After initial parameterization, training comprises using training data to iteratively adjust the initial parameters into adjusted parameters until convergence is determined.

Abstract: A method of forming a shareable filler model (shareable model for garbage words) from a word n-gram model is provided. The word n-gram model is converted into a probabilistic context free grammar (PCFG). The PCFG is modified into a substantially application-independent PCFG, which constitutes the shareable filler model.

Abstract: A novel system integrates speech recognition and semantic classification, so that acoustic scores in a speech recognizer that accepts spoken utterances may be taken into account when training both language models and semantic classification models. For example, a joint association score may be defined that is indicative of a correspondence of a semantic class and a word sequence for an acoustic signal. The joint association score may incorporate parameters such as weighting parameters for signal-to-class modeling of the acoustic signal, language model parameters and scores, and acoustic model parameters and scores. The parameters may be revised to raise the joint association score of a target word sequence with a target semantic class relative to the joint association score of a competitor word sequence with the target semantic class. The parameters may be designed so that the semantic classification errors in the training data are minimized.

Abstract: A user's search experience may be enhanced by providing additional content based upon an understanding of the user's intent. Query tagging, the assigning of semantic labels to terms within a query, is one technique that may be utilized to determine the context of a user's search query. Accordingly, as provided herein, a query tagging model may be updated using one or more stratified lexicons. A list data structure (e.g., lists of phrases obtained from web pages) and seed distribution data (e.g., pre-labeled probability data) may be used by a graph learning technique to obtain an expanded set of phrases and their respective probabilities of corresponding with particular lexicons (e.g., semantic class lexicons). The expanded set of phrases may be used to group phrases into stratified lexicons. The stratified lexicons may be used as features for updating and/or executing the query tagging model.

Abstract: A training system for a speech recognition application is disclosed. In embodiments described, the training system is used to train a classification model or language model. The classification model is trained using an adaptive language model generated by an iterative training process. In embodiments described, the training data is recognized by the speech recognition component and the recognized text is used to create the adaptive language model which is used for speech recognition in a following training iteration.

Abstract: A training module is described for training a conditional random field (CRF) tagging model. The training module trains the tagging model based on an explicitly-labeled training set and an implicitly-labeled training set. The explicitly-labeled training set includes explicit labels that are manually selected via human annotation, while the implicitly-labeled training set includes implicit labels that are generated in an unsupervised manner. In one approach, the training module can train the tagging model by treating the implicit labels as non-binding evidence that has a bearing on values of hidden state sequence variables. In another approach, the training module can treat the implicit labels as binding or hard evidence. A labeling system is also described for providing the implicit labels.

Abstract: Described is a technology in which product or service reviews are automatically processed to form a summary for each single product or service. Snippets from the reviews are extracted and classified into sentiment classes (e.g., as positive or negative) based on their wording. Attributes are assigned to the reviews, e.g., based on term frequency concepts, as nouns, which may be paired with adjectives and/or verbs. The summary of the reviews belonging to a single product or service is generated based on the automatically computed attributes and the classification of review snippets into attribute and sentiment classes. For example, the summary may indicate how many reviews were positive (the sentiment class), along with text corresponding to the most similar snippet based on its similarity to the attributes (the attribute class).

Abstract: An automated “Voice Search Message Service” provides a voice-based user interface for generating text messages from an arbitrary speech input. Specifically, the Voice Search Message Service provides a voice-search information retrieval process that evaluates user speech inputs to select one or more probabilistic matches from a database of pre-defined or user-defined text messages. These probabilistic matches are also optionally sorted in terms of relevancy. A single text message from the probabilistic matches is then selected and automatically transmitted to one or more intended recipients. Optionally, one or more of the probabilistic matches are presented to the user for confirmation or selection prior to transmission. Correction or recovery of speech recognition errors avoided since the probabilistic matches are intended to paraphrase the user speech input rather than exactly reproduce that speech, though exact matches are possible.