Abstract: Embodiments of this disclosure disclose a speech processing method and a related device, which may be applied to a scenario in which a user records a short video, a teacher records a teaching speech, or the like. The method includes: obtaining an original speech and a second text, where an original text corresponding to the original speech and a target text to which the second text belongs both include a first text; generating a second speech feature of a corrected text (that is, the second text) in the target text by referring to a first speech feature of a correct text (that is, the first text) in the original text; and then generating, based on the second speech feature, a target edited speech corresponding to the corrected text.

Abstract: The present disclosure relates to text data processing methods and apparatuses. One example method includes obtaining target text, where a phoneme of the target text includes a first phoneme and a second phoneme that are adjacent to each other. Feature extraction is performed on the first phoneme and the second phoneme to obtain a first audio feature of the first phoneme and a second audio feature of the second phoneme. By using a target recurrent neural network (RNN) and based on the first audio feature, first speech data corresponding to the first phoneme is obtained.By using the target RNN and based on the second audio feature, second speech data corresponding to the second phoneme is obtained.By using a vocoder and based on the first speech data and the second speech data, audio corresponding to the first phoneme and audio corresponding to the second phoneme are obtained.

Abstract: Disclosed is an improved approach for efficiently implementing a three-dimensional integrated circuit (3D-IC) design with heterogeneous and/or homogeneous dies. A first die design and a second die design in a three-dimensional (3D) electronic design maybe identified, and a wrapper design may be generated for at least a block of circuit component designs in the second die design for concurrent implementation of both the first and the second die designs. Both the first and the second dies of the 3D electronic design are concurrently implemented based at least upon a floorplan that is generated with at least the wrapper design for the 3D electronic design. A first wrapper and a second wrapper may be respectively generated for the first die design and the second die design based at least in part upon a result of the concurrent implementation.

Abstract: Disclosed is an improved approach for implementing a three-dimensional integrated circuit design with mixed macro and standard cell placement. This approach concurrently places both the macros and standard cells of the 3D-IC design onto two or more stacked floorplan and optimize the instance locations by timing, density, wire length and floorplan constraint.

Abstract: In a speech synthesis method, an emotion intensity feature vector is set for a target synthesis text, an acoustic feature vector corresponding to an emotion intensity is generated based on the emotion intensity feature vector by using an acoustic model, and a speech corresponding to the emotion intensity is synthesized based on the acoustic feature vector. The emotion intensity feature vector is continuously adjustable, and emotion speeches of different intensities can be generated based on values of different emotion intensity feature vectors, so that emotion types of a synthesized speech are more diversified. This application may be applied to a human-computer interaction process in the artificial intelligence (AI) field, to perform intelligent emotion speech synthesis.

Abstract: A speech synthesis method and a speech synthesis apparatus to synthesize speeches of different emotional intensities in the field of artificial intelligence, where the method includes obtaining a target emotional type and a target emotional intensity parameter that correspond to an input text, determining a corresponding target emotional acoustic model based on the target emotional type and the target emotional intensity parameter, inputting a text feature of the input text into the target emotional acoustic model to obtain an acoustic feature of the input text, and synthesizing a target emotional speech based on the acoustic feature of the input text.

Abstract: Disclosed is an approach to implement multi-die concurrent placement, routing, and/or optimization across multiple dies. This permits the multiple dies to be modeled as a single 3D space. Instead of being limited to a 2D plane, a cell can be placed to the area of any of the dies without splitting the netlist beforehand.

Abstract: The present invention discloses a missing semantics complementing method in the field of natural language processing in the artificial intelligence field, including: obtaining a question statement and a historical dialog statement; resolving a to-be-resolved item in the question statement based on the historical dialog statement and location information of the to-be-resolved item, to obtain a resolved question statement; determining whether a component in the question statement is ellipted, and if a component in the question statement is ellipted, complementing the ellipted component based on the historical dialog statement, to obtain a question statement after ellipsis resolution; merging the resolved question statement and the question statement after ellipsis resolution, to obtain a merged question statement; and determining a target complemented question statement from the resolved question statement, the question statement after ellipsis resolution, and the merged question statement.

Abstract: In a speech synthesis method, an emotion intensity feature vector is set for a target synthesis text, an acoustic feature vector corresponding to an emotion intensity is generated based on the emotion intensity feature vector by using an acoustic model, and a speech corresponding to the emotion intensity is synthesized based on the acoustic feature vector. The emotion intensity feature vector is continuously adjustable, and emotion speeches of different intensities can be generated based on values of different emotion intensity feature vectors, so that emotion types of a synthesized speech are more diversified. This application may be applied to a human-computer interaction process in the artificial intelligence (AI) field, to perform intelligent emotion speech synthesis.

Abstract: The present disclosure relates to computing resource allocation methods, devices, and systems. One example system includes a management node and a target computing node. The management node is configured to obtain M computing tasks and establish a resource assessment model, and send one or more computing tasks of the M computing tasks and information about the resource assessment model to the target computing node. The target computing node is configured to receive the one or more computing tasks and the information about the resource assessment model, substitute input data of a particular computing stage of a target task into the resource assessment model to compute a resource size required for the particular computing stage, and compute the input data by using a computing resource that is of the resource size and that is in a preset resource pool.

Abstract: The present disclosure relates to a resource allocation method for gene analysis. In one example method, a parameter value that is of the target chromosome region and that is used for resource allocation is obtained according to a sequenced read in a target chromosome region. A computing resource is allocated, according to the parameter value that is of the target chromosome region and that is used for resource allocation, to an operation in a cleansing and variant calling task that is in the gene analysis and that is performed on the sequenced read in the target chromosome region.

Abstract: A speech synthesis method and a speech synthesis apparatus to synthesize speeches of different emotional intensities in the field of artificial intelligence, where the method includes obtaining a target emotional type and a target emotional intensity parameter that correspond to an input text, determining a corresponding target emotional acoustic model based on the target emotional type and the target emotional intensity parameter, inputting a text feature of the input text into the target emotional acoustic model to obtain an acoustic feature of the input text, and synthesizing a target emotional speech based on the acoustic feature of the input text.

Abstract: A method for scheduling a data flow task and an apparatus. The method includes: preprocessing a data flow task to obtain at least one subtask; classifying the subtask into a central processing unit (CPU) task group, a graphics processing unit (GPU) task group, or a to-be-determined task group; allocating the subtask to a working node; when the subtask belongs to the CPU task group, determining that a CPU executes the subtask; when the subtask belongs to the GPU task group, determining that a GPU executes the subtask; or when the subtask belongs to the to-be-determined task group, determining, according to costs of executing the subtask by a CPU and a GPU, a running platform (e.g., the CPU or the GPU) executes the subtask, where the cost includes duration of executing the subtask.

Abstract: Embodiments of the present invention related to in Artificial Intelligence field and disclose a computing resource allocation method, a related device, and a MapReduce-based distributed system. The system includes a management node and a target computing node. The management node is configured to: obtain M computing tasks, and establish a resource assessment model; and send one or more computing tasks of the M computing tasks and information about the resource assessment model to the target computing node. The target computing node is configured to: receive the one or more computing tasks and the information about the resource assessment model, and, substitute input data of the computing stage into the resource assessment model to compute a resource size of a computing resource required for the computing stage; and compute the input data by using a computing resource that is of the resource size and that is in a preset resource pool.

Abstract: A data processing method includes traversing all sample fragments in a first sample set and collecting statistics about a first statistic of each basic element in a reference sample and included in the sample fragments, determining that a position of a basic element in the reference sample whose first statistic is less than a first threshold is a spacing position, dividing the reference sample into at least two reference sub-samples, traversing all the sample fragments in the first sample set and collecting statistics about a second statistic of each reference sub-sample of the reference sample and including the sample fragments, and combining adjacent reference sub-samples when a sum of second statistics of the adjacent reference sub-samples is less than a second threshold.

Abstract: A data processing method includes distributing, by a computing node, a pasting back result sequence corresponding to a to-be-pasted-back deoxyribonucleic acid (DNA) read string to a pasting back result sequence set corresponding to a target chromosome region, when the quantity of the pasting back result sequences included in the pasting back result sequence set is greater than or equal to the pre-determined quantity threshold, dividing the pasting back result sequence set into k pasting back result sequence subsets according to a preset division rule, and dividing the target chromosome region into k chromosome subregions in a one-to-one correspondence to the k pasting back result sequence subsets, and further dividing a gene analysis task of the pasting back result sequence set into k gene analysis subtasks, and executing in parallel the k gene analysis subtasks.

Abstract: A DNA sequence processing method and device are used to resolve a prior-art problem of low-efficiency mutation detection on a DNA sample. The method includes: performing alignment computation on each read in the read group according to a reference sequence of a chromosome to obtain an alignment result record of the read relative to the reference sequence; determining a chromosome region in which each read is located; and merging alignment result records of reads located in a same chromosome region into one intermediate result file; determining a target sequence file of each chromosome region according to the N intermediate result files corresponding to the chromosome region; and determining mutation site information of each chromosome region according to the target sequence file of the chromosome region.

Abstract: The present disclosure relates to a resource allocation method for gene analysis. In one example method, a parameter value that is of the target chromosome region and that is used for resource allocation is obtained according to a sequenced read in a target chromosome region. A computing resource is allocated, according to the parameter value that is of the target chromosome region and that is used for resource allocation, to an operation in a cleansing and variant calling task that is in the gene analysis and that is performed on the sequenced read in the target chromosome region.

Abstract: A method for scheduling a data flow task and an apparatus. The method includes: preprocessing a data flow task to obtain at least one subtask; classifying the subtask into a central processing unit (CPU) task group, a graphics processing unit (GPU) task group, or a to-be-determined task group; allocating the subtask to a working node; when the subtask belongs to the CPU task group, determining that a CPU executes the subtask; when the subtask belongs to the GPU task group, determining that a GPU executes the subtask; or when the subtask belongs to the to-be-determined task group, determining, according to costs of executing the subtask by a CPU and a GPU, a running platform (e.g., the CPU or the GPU) executes the subtask, where the cost includes duration of executing the subtask.

Abstract: A metadata updating method based on columnar storage in a distributed file system includes acquiring to-be-updated metadata in a data table, splitting data records of the data table into multiple row groups on a row basis, converting the data table into global file metadata and multiple row group files, where the row group file includes an actual data block, a data index block, a local metadata block, a metadata index block, and a file footer, determining whether the to-be-updated metadata belongs to the global file metadata, updating local metadata when the to-be-updated metadata does not belong to the global file metadata, and adding an updated local metadata block, an updated metadata index block, and an updated file footer to the multiple row group files according to updated local metadata. Dynamic updating of metadata saves time of executing an updating operation of this type and needed computing resources.