Abstract: A speech interface device may be configured to act as a remote speaker peripheral for multiple audio sources such as media players and phones. Upon receiving a request from a user to connect to an audio device, the speech interface device determines which of the multiple audio devices are currently available, selects one of the audio devices based on information about or received from the user, establishes an audio connection with the selected one of the audio devices, and begins acting as a remote speaker or speakerphone for the selected audio device.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for presenting additional information for text depicted by an image. In one aspect, a method includes receiving an image. Text depicted in the image is identified. The identified text can be in one or more text blocks. A prominence presentation context is selected for the image based on the relative prominence of the one or more text blocks. Each prominence presentation context corresponds to a relative prominence of each text block in which text is presented within images. Each prominence presentation context has a corresponding user interface for presenting additional information related to the identified text depicted in the image. A user interface is identified that corresponds to the selected prominence presentation context. Additional information is presented for at least a portion of the text depicted in the image using the identified user interface.

Abstract: A sampler module may divide an audio signal into a series of sequential samples. A signal quality detector module may identify a consistent brick wall frequency of the audio signal spanning a plurality of the sequential samples at an outset of the audio signal and determine a signal treatment indication proportional to the brick wall frequency. A signal enhancer module may sequentially receive and analyze one or more sample components of the audio signal to identify lost parts of the audio signal in the one or more sample components of respective sequential samples, and generate, in accordance with the signal quality indication, a corresponding signal treatment for each of the one or more sample components of respective sequential samples having a corresponding identified lost part.

Abstract: A method of decision definition using a rules vocabulary includes: receiving free form input; identifying terms contained within the free form input; searching the rules vocabulary objects for terms; responsive to the term being found, obtaining input from a user as to whether to use the found term; responsive to the term not being found; searching the rules vocabulary attributes for terms having attributes corresponding to the term; responsive to the term being found, obtaining input from a user as to whether to use the found term; and refactoring the free form input with the found term accepted by the user. The method also includes updating the rules vocabulary with the term identified in the free form input as a synonym for the term found in said rules vocabulary. One embodiment further provides a method of determining semantic equivalence between a plurality of rules using a rules database having preferred terms.

Abstract: A voice-controlled computer system comprising a headset with a heads up display (HUD). When the system requires a user to verbally enter a character of an alphanumeric security code the system presents to the user on the HUD a plurality of different pairs of alphanumeric characters each containing a first character and an associated (pseudo) randomly generated second character. The first character of one pair is the code character to be entered, but for security reasons the user is required to verbally enter the second character associated with the code character, not the code character itself. The system deciphers verbal entry of the associated second character as equivalent to entry of the first character.

Abstract: Systems and methods for speech-based monitoring and/or control of automation devices are described. A speech-based method for monitoring and/or control of automation devices may include steps of determining a type of automation device to which first speech relates based, at least in part, on a location associated with the first speech; selecting a topic-specific speech recognition model adapted to recognize speech related to the determined type of automation device; using the topic-specific speech recognition model to recognize second speech provided at the location, wherein recognizing the second speech comprises identifying a query or command relating to the type of automation device and represented by the second speech; and issuing the query or command represented by the second speech to an automation device of the determined type.

Abstract: Described are techniques for automatically improving the accessibility of webpages and other content using machine learning and artificial intelligence systems. Webpage data may include visual data used to render visible elements and audio data used to render audible elements, such as digitized speech representative of at least a portion of the visible elements. In some cases, text data may be generated based on the audio data. The audio data may be modified based on target text strings, patterns, and characteristics determined in the text data, or the audio data may be analyzed directly. Additionally, user interactions with particular visible elements and corresponding audible elements may be compared. If the user interactions for a visible element exceed the user interactions for a corresponding audible element, the audio data associated with the audible element may be modified.

Abstract: Features are disclosed for performing speech recognition on utterances using dynamic weights with speech recognition models. An automatic speech recognition system may use a general speech recognition model, such a large finite state transducer-based language model, to generate speech recognition results for various utterances. The general speech recognition model may include sub-models or other portions that are customized for particular tasks, such as speech recognition on utterances regarding particular topics. Individual weights within the general speech recognition model can be dynamically replaced based on the context in which an utterance is made or received, thereby providing a further degree of customization without requiring additional speech recognition models to generated, maintained, or loaded.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for transforming text strings. In general, one aspect of the subject matter described in this specification can be embodied in methods that include the actions of receiving input string having a plurality of terms, the input string being in a first form; transforming the input string from the first form to a second form including: applying one or more rules to the input string to identify one or more terms for translation, the one or more identified terms being fewer than the plurality of terms, translating the identified one or more terms to one or more translated terms in the second form, and transliterating the remaining terms of the plurality of terms into transliterated terms in the second form; and concatenating the translated and transliterated terms to form a hybrid output string in the second form.

Abstract: A method and system for improving location specific (L) functions is provided. The method includes generating and maintaining an online L-word dictionary. Location specific attributes, settings, and preferences associated with a current location of a mobile device and a user are retrieved. An input question is received and an NLP analysis with respect to the input question is executed to extract a required value phrase. An interim question based on an extracted required value phrase is formed and an answer to the input question is determined in natural language based on a solved interim question. The answer is narrated in natural language and includes an automated interactive response from a hardware device in real time.

Abstract: Computer-generated speech devices for site survey and maintenance, and methods of using the same are described herein. One computer-generated speech device includes a location engine to determine a location of the computer-generated speech device at a site, a solution engine to identify an action to perform associated with one of a plurality of nodes at the site using the location of the computer-generated speech device and input data associated with the plurality of nodes, and a speech engine to broadcast the identified action as computer-generated speech using a speaker component of the computer-generated speech device for a user to perform during a survey and/or maintenance of the site.

Abstract: Embodiments described herein relate generally to content analysis technologies and natural language processing (NLP). In particular, devices, systems, and methods may implement a reverse sentence reconstruct (RSR) utility, and a sentence vectorization technique (SVT) utility. A computer server may be configured to receive a feeling classification request with text data elements, and in response, generate a feeling classification response indicating feeling for the text data elements using the RSR utility and the SVT utility.

Abstract: An exemplary computer system configured to train an ASR using the output from a TTS engine.

Abstract: A method is provided for extending the frequency band of an audio signal during a decoding or improvement process. The method includes obtaining the decoded signal in a first frequency band, referred to as a low band. Tonal components and a surround signal are extracted from the signal from the low-band signal, and the tonal components and the surround signal are combined by adaptive mixing using energy-level control factors to obtain an audio signal, referred to as a combined signal. The low-band decoded signal before the extraction step or the combined signal after the combination step are extended over at least one second frequency band which is higher than the first frequency band. Also proved are a frequency-band extension device which implements the described method and a decoder including a device of this type.

Abstract: A method of providing navigation on an electronic device when the display screen is locked. The method receives a verbal request to start navigation while the display is locked. The method identifies a route from a current location to a destination based on the received verbal request. While the display screen is locked, the method provides navigational directions on the electronic device from the current location of the electronic device to the destination. Some embodiments provide a method for processing a verbal search request. The method receives a navigation-related verbal search request and prepares a sequential list of the search results based on the received request. The method then provides audible information to present a search result from the sequential list. The method presents the search results in a batch form until the user selects a search result, the user terminates the search, or the search items are exhausted.

Abstract: A computer system may receive a textual work. The computer system may generate a knowledge graph based on the textual work. The knowledge graph may include nodes representing concepts and edges between the nodes that represent links between the concepts. The computer system may then generate a concept path for a target concept. The computer system may then identify a related background narrative block that contains a related non-target concept. The background narrative block may be a narrative block that is not in the concept path for the target concept. The computer system may then summarize the related background narrative block and output the summary to an output device coupled with the computer system.

Abstract: A device for analyzing reputation for objects by data mining is provided. The device includes a big data collector assembly collecting first information from a web page, a text preprocessor assembly analyzing morpheme information from an original text of the first information, an ontology dictionary extender assembly classifying the morpheme information according to each category and extracting an associated word, an object recognizer assembly measuring similarity between the associated word from the morpheme information to recognize a plurality of entity names associated with a first object, an information sensitivity classifier assembly classifying sensitivity of the first information based on a frequency of a sensitive word associated with the first object from the morpheme information, and a risk evaluation modeler assembly deriving a numeric risk value of the first information according to a numeric sensitivity value of the sensitive word and a weight value level for evaluating event importance.

Abstract: A computer implemented method of generating a language section from tabular data in an electronic document may include identifying, in a first tabular portion of the electronic document, a set of categories used to organize tabular data. The method may include identifying a content characteristic for each category of the set of categories in the first tabular portion. And the method may include generating a first language section from at least two distinct categories of the set of categories, wherein a format of the first language section is based on the content characteristics for the at least two distinct categories.

Abstract: Software translation quality and efficiency are improved by providing user interface (UI) context for translators. Unicode symbols are used to uniquely tag user-visible strings from the source code and into resource files. Those strings include titles, product names, error messages, strings in images and any other text that may be present on the user interface. Once the ‘pseudo’ resource files are integrated into a build, automation is run to gather screenshots of the application. Image recognition is then used to link screenshots of the UI in which a resource file string appears, such that screenshots will be brought forward and displayed to the translator when working on translating the user-visible strings of the software being localized.

Abstract: A computer implemented method of generating a language section from tabular data in an electronic document may include identifying, in a first tabular portion of the electronic document, a set of categories used to organize tabular data. The method may include identifying a content characteristic for each category of the set of categories in the first tabular portion. And the method may include generating a first language section from at least two distinct categories of the set of categories, wherein a format of the first language section is based on the content characteristics for the at least two distinct categories.