Abstract: A method of detecting a defect in a sheet of glass includes: (i) directing a converging beam from a source of illumination onto a surface of the sheet of glass to illuminate the defect; (ii) focussing an image capture device onto a first plane to image the defect in the sheet of glass; (iii) capturing a first image of the defect; (iv) carrying out an adjustment step; and (v) capturing a second image of the defect. Each of the first and second images has a respective first portion from the illuminated defect and a respective second portion due to reflection of a portion of the beam from the glass surface. In the first image of the defect the first portion is brighter than the second portion, and in the second image of the defect the first portion is darker than the second portion. An apparatus for carrying out the method is also provided.

Abstract: The techniques described herein may provide for an evidence-based, clinical practice guideline-driven, patient communication system. A user (e.g., a patient, a client, etc.) may interact with, or use, a patient communication system for improved healthcare according to one or more aspects. The patient communication system may credibly categorize user inputs into injury patterns (e.g., correlated to healthcare industry established clinical practice guidelines). For instance, an evidence-based, clinical practice guideline-driven, patient communication system may utilize a library of extensive and credible healthcare information (e.g., reviewed by a credible body established by the healthcare industry). Such information may include physical therapy videos, recovery strategies, workplace group or client-group healthcare information, etc. The patient communication system may generate (e.g.

Abstract: The technology provides an approach to train translation models that are robust to transcription errors and punctuation errors. The approach includes introducing errors from actual automatic speech recognition and automatic punctuation systems into the source side of the machine translation training data. A method for training a machine translation model includes performing automatic speech recognition on input source audio to generate a system transcript. The method aligns a human transcript of the source audio to the system transcript, including projecting system segmentation onto the human transcript. Then the method performs segment robustness training of a machine translation model according to the aligned human and system transcripts, and performs system robustness training of the machine translation model, e.g., by injecting token errors into training data.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer-storage media, for re-translation for simultaneous, spoken-language machine translation. In some implementations, a stream of audio data comprising speech in a first language is received. A transcription for the speech in the stream of audio data is generated using an automated speech recognizer through a series of updates. A translation of the transcription into a second language is generated using a machine translation module. The translation is generated with translation iterations that translate increasing amounts of the transcription, including re-translating previously portions of the transcription. A series of translation updates are provided to a client device based on the translation iterations.

Abstract: A crack climbing unit is provided and comprises a body having an inwardly facing jamming surface. The surface is provided with a compliant liner.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer-storage media, for re-translation for simultaneous, spoken-language machine translation. In some implementations, a stream of audio data comprising speech in a first language is received. A transcription for the speech in the stream of audio data is generated using an automated speech recognizer through a series of updates. A translation of the transcription into a second language is generated using a machine translation module. The translation is generated with translation iterations that translate increasing amounts of the transcription, including re-translating previously portions of the transcription. A series of translation updates are provided to a client device based on the translation iterations.

Abstract: The techniques described herein may provide for an evidence-based, clinical practice guideline-driven, patient communication system. A user (e.g., a patient, a client, etc.) may interact with, or use, a patient communication system for improved healthcare according to one or more aspects. The patient communication system may credibly categorize user inputs into injury patterns (e.g., correlated to healthcare industry established clinical practice guidelines). For instance, an evidence-based, clinical practice guideline-driven, patient communication system may utilize a library of extensive and credible healthcare information (e.g., reviewed by a credible body established by the healthcare industry). Such information may include physical therapy videos, recovery strategies, workplace group or client-group healthcare information, etc. The patient communication system may generate (e.g.

Abstract: A computer-implemented method includes compiling one or more segments of code during run-time of a process executing at one or more processors of a computer system. The compilation produces a high-level intermediate representation of the one or more segments of the code. The high-level intermediate representation is modifiable by the process, without executing the high-level intermediate representation, to generate a modified high-level intermediate representation that is executable by the process.

Abstract: A system and method to achieve low power and/or low supply operation of a delta-sigma modulator by taking advantage of the inherent virtual ground of the delta-sigma loop to make the input to a low power integrator small and largely independent of the input signal. This results in improved linearity of the integrator and relaxed constraints on the supply for the first stage integrator. The architecture also enables direct access to the quantization error of the feedback loop and thus can be used to either/or: 1. Calibrate the modulator, 2. Achieve reduced quantization noise, 3. Stabilize the loop by compensating for excess loop delay. Low voltage common-mode-feedback is also achieved using the techniques described.

Abstract: A system and method to achieve low power and/or low supply operation of a delta-sigma modulator by taking advantage of the inherent virtual ground of the delta-sigma loop to make the input to a low power integrator small and largely independent of the input signal. This results in improved linearity of the integrator and relaxed constraints on the supply for the first stage integrator. The architecture also enables direct access to the quantization error of the feedback loop and thus can be used to either/or: 1. Calibrate the modulator, 2. Achieve reduced quantization noise, 3. Stabilize the loop by compensating for excess loop delay. Low voltage common-mode-feedback is also achieved using the techniques described.

Abstract: A method includes, during compilation of a portion of source code at a compiler, initiating token resolution with respect to a token included in the portion of source code. A static context and a dynamic context of the compiler are queried to identify one or more lookup results of the token. A particular lookup result is selected to resolve the token based on one or more semantic rules of the compiler. The particular lookup result is selected independently of whether the particular lookup result is sourced from the static context or from the dynamic context. The method includes resolving the token based on the selected particular lookup result.

Abstract: An automated published data monitoring system implements a content validation service capable of validating published data in accordance with programmable criteria. A root data location is provided and validation of such data includes crawling a hierarchical organization of additional data. Deserializers are specific to identified collections of data and deserialize data into strongly typed data structures that are programmatically validatable. Deserializers register themselves to handle collections of data identified based upon the location and domain of such data. Additionally, validators are specific to types of data structures and programmatically validate such data structures including validating their type and their correctness, the latter as compared to statically or dynamically defined limits. Validators register themselves to handle specified types of data structures originating from specific data collections.

Abstract: Described is a technology for performing unsupervised learning using global features extracted from unlabeled examples. The unsupervised learning process may be used to train a log-linear model, such as for use in morphological segmentation of words. For example, segmentations of the examples are sampled based upon the global features to produce a segmented corpus and log-linear model, which are then iteratively reprocessed to produce a final segmented corpus and a log-linear model.

Abstract: A method includes, during compilation of a portion of source code at a compiler, initiating token resolution with respect to a token included in the portion of source code. A static context and a dynamic context of the compiler are queried to identify one or more lookup results of the token. A particular lookup result is selected to resolve the token based on one or more semantic rules of the compiler. The particular lookup result is selected independently of whether the particular lookup result is sourced from the static context or from the dynamic context. The method includes resolving the token based on the selected particular lookup result.

Abstract: A computer-implemented method includes compiling one or more segments of code during run-time of a process executing at one or more processors of a computer system. The compilation produces a high-level intermediate representation of the one or more segments of the code. The high-level intermediate representation is modifiable by the process, without executing the high-level intermediate representation, to generate a modified high-level intermediate representation that is executable by the process.

Abstract: Described is a transliteration engine/substring decoder that back-transliterates an input string from a source language into an output string in a target language. The transliteration engine may be based upon discriminately weighted indicator features and/or generative models in which the decoder's discriminative parameters are learned. The training data may be based on source-target pairs, which may be transformed into derivations. Features extracted from these derivations include indicator features and hybrid generative model features.

Abstract: Described is a technology for performing unsupervised learning using global features extracted from unlabeled examples. The unsupervised learning process may be used to train a log-linear model, such as for use in morphological segmentation of words. For example, segmentations of the examples are sampled based upon the global features to produce a segmented corpus and log-linear model, which are then iteratively reprocessed to produce a final segmented corpus and a log-linear model.

Abstract: Reusable component data persistence using markup languages may be provided. A file including data associated with a reusable software component may be opened. Next, it may be determined if the file includes a persistence element comprising a dataset. The dataset may contain the data associated with the reusable software component. The data associated with the reusable software component may be described in the markup language and may reference a sub-element in a binary format. In addition, if the file includes the persistence element comprising the dataset, saving the data associated with the reusable software component to the dataset may be performed or loading the data associated with the reusable software component from the dataset may be performed.

Abstract: Disclosed is a frequency-locked loop (FLL), which attempts to bring about frequency and phase synchronization between two signals over the control bandwidth of the loop: a reference signal and a voltage-controlled oscillator (VCO) signal. For example, the FLL employs a reference signal generated by a crystal oscillator of frequency fREF and a VCO signal generated by the oscillations of an unquenched SRG resonator with tunable resonant frequency fRES. These signals are connected to the inputs of a phase/frequency detector (PFD) which produces output pulses in response to the relationship between fREF and fRES. These pulses are applied to a loop filter (LF) which creates a voltage using some kind of charge-storage element. This loop filter voltage is a so-called error voltage whose value is used to control the frequency of the resonator to bring the reference signal and VCO signal into phase synchrony.