International Business Machines Corporation

Microsoft Technology Licensing, LLC

Canon Kabushiki Kaisha

Memblaze Technology (beijing) Co., Ltd.

Yong Ni

Peng Hu

Yong Xu

Yong Jie Pan

Wen Jie Wang

Mei-Yuh Hwang

Yongfeng Xu

Hai Tao Wang

Ruishan Guo

Jun Li

Ai Jie Niu

Pengfei Hu

Yongjie Pan

A method, an apparatus and a system for information retrieval are provided. The method for information retrieval may comprise: receiving a query in a natural language form from a user (); extracting a plurality of feature vectors from the query based on a plurality of knowledge entries in a knowledge base associated with the user and a built-in featurization source, one feature vector per knowledge entry (); obtaining, with a pre-trained language understanding model, a plurality of language understanding results based on the plurality of feature vectors (); and selecting a knowledge entry corresponding to the query among the plurality of entries based on the plurality of language understanding results (). The method may extract more features and combine information retrieval and language understanding in one shot to improve efficiency.

Combo of language understanding and information retrieval

A power supply circuit coupled to a power source for providing electronic energy to a workload includes a first switch, a control circuit, a first diode, and a backup power, in which the input of the first switch is coupled to the power source, the output of the first switch is coupled to the workload, the control circuit is coupled to the control terminal of the first switch for determining the state of the power source, the anode of the first diode is coupled to the power source, the cathode of the first diode is coupled to the first terminal of the backup power, the first terminal of the backup power is further coupled to the workload, when the control circuit determines that the power source is in a power-off state, a control signal is applied to the control terminal of the first switch to turn off the first switch.

Power supply circuit and power supply method

Adjusting the layout size of a hyperlink includes displaying at least one hyperlink in a user interface; detecting a touch operation for the at least one hyperlink, and extracting position coordinates of a touch point formed by the touch operation on the user interface. Adjusting the layout size further includes determining a target hyperlink from the at least one hyperlink, and determining the precision of the touch operation with respect to the target hyperlink based on the position coordinates of the touch point; and adjusting layout size of the target hyperlink based on the determined precision. The hyperlink layout in a web page can be adapted to the touch precision of user's finger automatically, which facilitates the recognition of hyperlinks by the user's finger.

Adjusting layout size of hyperlink

Embodiments include a method for adjusting layout size of a hyperlink. The method comprises: displaying at least one hyperlink in a user interface; detecting a touch operation for the at least one hyperlink, and extracting position coordinates of a touch point formed by the touch operation on the user interface; determining a target hyperlink from the at least one hyperlink, and determining the precision of the touch operation with respect to the target hyperlink based on the position coordinates of the touch point; and adjusting layout size of the target hyperlink based on the determined precision. A hyperlink layout in a web page can be adapted to the touch precision of user's finger automatically, which facilitates the recognition of hyperlinks by the user's finger.

Annotated training data (e.g., sentences) in a first language are used to generate annotated training data for a second language. For example, annotated sentences in English are manually collected first, and then is used to generate annotated sentences in Chinese. The annotated training data includes slot labels, slot values and carrier phrases. The carrier phrases are the portions of the training data that is outside of a slot. The carrier phrases are translated from the first language to one or more translations in the second language. The translations may include machine translations as well as human translations. Entities for the slot values are determined for the translated sentences using content sources that include locale-dependent entities. The determined entities are used to fill the slots in the translations of the second language. All or a portion of the resulting sentences may be used for training models in the second language.

Filled translation for bootstrapping language understanding of low-resourced languages

The present invention provides a method for diagnosing a problem of a software product running in a cloud environment and a corresponding apparatus, the method comprising: receiving a problem in the operation of the monitored software product from a diagnosis agent deployed on a node in the cloud environment; capturing the cloud environment including the software product, and deploying the captured cloud environment in a diagnosis cloud, and the step comprising: deploying the image of each node of the cloud environment in the diagnosis cloud; and applying corresponding configuration data for a cluster system deployed in the cloud environment to configure each node in the diagnosis cloud. The method and apparatus of the present invention can diagnose problems of a software product running in a cloud environment, and rebuild the cloud environment to facilitate the diagnosis of the problems.

Diagnosing a problem of a software product running in a cloud environment

A method, system and computer program product for saving power for a mobile terminal. Configuration parameters of a first virtual screen are obtained in response to the occurrence of a first trigger event, where the first virtual screen is at least part of the screen of the mobile terminal. The scope of the first virtual screen is determined according to the configuration parameters of the first virtual screen. Data to be displayed on the screen of the mobile terminal is displayed within the determined scope of the first virtual screen. The display of the screen of the mobile terminal is closed outside the determined scope of the first screen. By realizing a virtual screen and using only part of the screen of the mobile terminal for displaying, there is a reduction in the power consumed by the screen display so that the battery life of the mobile terminal is improved.

Reducing power consumption for a mobile terminal

The present invention disclosed a web based OCR system including an OCR service apparatus that communicates with a web browser of a client apparatus and an optical character recognition method for controlling the OCR service apparatus that communicates with the web browser of the client apparatus, the OCR service apparatus comprising: a storage unit for storing an image; a sending unit for sending the image to the web browser of said client apparatus; a receiving unit for receiving a control instruction of a character recognition process from the web browser of said client apparatus, wherein the received control instruction is generated in the web browser of said client apparatus; and a character recognition processing unit for executing the character recognition process for the image based on the control instruction. With such optical character recognition method and system based on web service application on a network, the user can get control of the recognition process flow and the efficiencies are improved.

Growing community of inventors

Beijing, China

Yong Ni