Abstract: Systems, apparatuses, methods, and computer program products are disclosed for automatically determining accuracy of entity recognition of text. An example method includes segmenting a service entity recognition analysis of the text and a gold entity recognition analysis of the text into common superstrings that define entity spans. The example method further includes classifying each of the entity spans based on an accuracy of entity recognition in the service analysis of the text corresponding to the entity spans using a classification system that differentiates accurately identified but improperly bounded entities into at least three subcategories to obtain an entity accuracy classification. The example method also includes obtaining a score report based on the entity accuracy classification. The example method additionally includes performing an action set based on the entity accuracy classification.

Abstract: Provided are an off-grid start method and system for a new energy power generation system. The method includes: gradually boosting the voltage of a master according to a plurality of preset voltage given values, and slaves determining, by means of measuring a voltage of the load of a system, a target voltage given value used by the master; and the master determining, by means of monitoring an output current of the master itself, that a slave is successfully connected in parallel, and then continuing to boost the output voltage until all the slaves run in parallel. Therefore, according to the solution, no upper-layer synchronous control is required during a black-start process, and no communication between a master and slaves is required.

Abstract: In general, this disclosure describes techniques for generating and evaluating automatic transcripts of audio recordings containing human speech. In some examples, a computing system is configured to: generate transcripts of a plurality of audio recordings; determine an error rate for each transcript by comparing the transcript to a reference transcript of the audio recording; receive, for each transcript, a subjective ranking selected from a plurality of subjective rank categories; determine, based on the error rates and subjective rankings, objective rank categories defined by error-rate ranges; and assign an objective ranking to a new machine-generated transcript of a new audio recording, based on the objective rank categories and an error rate of the new machine-generated transcript.

Abstract: Systems, apparatuses, methods, and computer program products are disclosed for automatically determining accuracy of entity recognition of text. An example method includes segmenting a service entity recognition analysis of the text and a gold entity recognition analysis of the text into common superstrings that define entity spans. The example method further includes classifying each of the entity spans based on an accuracy of entity recognition in the service analysis of the text corresponding to the entity spans using a classification system that differentiates accurately identified but improperly bounded entities into at least three subcategories to obtain an entity accuracy classification. The example method also includes obtaining a score report based on the entity accuracy classification. The example method additionally includes performing an action set based on the entity accuracy classification.

Abstract: Systems and methods for detecting demographic bias in automatic speech recognition (ASR) systems. Corpuses of transcriptions from different demographic groups are analyzed, where one of the groups is known to be susceptible to bias and another group is known not to be susceptible to bias. A difference between the transcription accuracy for the first group and a transcription accuracy for a second group is measured. ASR accuracy for each group is measured and compared to each other using both statistics-based and practicality-based methodologies to determine whether a given ASR system or model exhibits a meaningful level of bias. Based on the statistical significance and the practical significance, an alert including a recommendation to adjust the ASR model is generated.

Abstract: An inverter circuit control method and a device thereof are provided. In the control method, after determining a DC bus voltage and an output leakage current of a target inverter circuit, a modulation harmonic wave to be injected into a SPWM signal is adjusted according to the DC bus voltage or the output leakage current if it is determined according to the DC bus voltage that the target inverter circuit satisfies a preset modulation condition, such that total harmonic distortion of a current of the target inverter circuit is within a preset range. The DC bus voltage and the output leakage current of the target inverter circuit are used as references for adjusting the modulation harmonic wave, ensuring total harmonic distortion of the current of the target inverter circuit to be within a preset range.

Abstract: Systems and methods for detecting demographic bias in automatic speech recognition (ASR) systems. Corpuses of transcriptions from different demographic groups are analyzed, where one of the groups is known to be susceptible to bias and another group is known not to be susceptible to bias. ASR accuracy for each group is measured and compared to each other using both statistics-based and practicality-based methodologies to determine whether a given ASR system or model exhibits a meaningful level of bias.

Abstract: An inverter circuit control method and a device thereof are provided. In the control method, after determining a DC bus voltage and an output leakage current of a target inverter circuit, a modulation harmonic wave to be injected into a SPWM signal is adjusted according to the DC bus voltage or the output leakage current if it is determined according to the DC bus voltage that the target inverter circuit satisfies a preset modulation condition, such that total harmonic distortion of a current of the target inverter circuit is within a preset range. The DC bus voltage and the output leakage current of the target inverter circuit are used as references for adjusting the modulation harmonic wave, ensuring total harmonic distortion of the current of the target inverter circuit to be within a preset range.

Abstract: A method for suppressing a common mode impulse current for an inverter generated when an alternating current switch is switched on and a device for applying the method are provided. Before the alternating current switch is switched on for the inverter, a target value of a common mode voltage between the inverter and ground is calculated based on at least one of a voltage between an alternating current port and ground, a voltage between the inverter and ground and a voltage across the alternating current switch sampled in real time. Then, a compensation power supply arranged between a main circuit of the inverter and ground is controlled to change the common mode voltage to the target value, so that a common mode voltage difference between two ends of the alternating current switch is zero. Subsequently, the inverter is controlled to switch on the alternating current switch.

Abstract: A method for suppressing a common mode impulse current for an inverter generated when an alternating current switch is switched on and a device for applying the method are provided. Before the alternating current switch is switched on for the inverter, a target value of a common mode voltage between the inverter and ground is calculated based on at least one of a voltage between an alternating current port and ground, a voltage between the inverter and ground and a voltage across the alternating current switch sampled in real time. Then, a compensation power supply arranged between a main circuit of the inverter and ground is controlled to change the common mode voltage to the target value, so that a common mode voltage difference between two ends of the alternating current switch is zero. Subsequently, the inverter is controlled to switch on the alternating current switch.