Methods, devices and systems enhance compression and decompression of data values when they comprise a plurality of semantically meaningful data fields. Compression is sometimes not applied to each data value as a whole, but instead to at least one of the semantically meaningful data fields of each data value, and in isolation from the other ones. Data fields can be organized that share the same semantic meaning together to accelerate compression and decompression as multiple compressors and decompressors can be used in parallel. A system can be used where methods and devices are tailored to perform compression and decompression of the semantically meaningful data fields of floating-point numbers after first partitioning further at least one of said data fields into two or a plurality of sub-fields to increase the degree of value locality and improve compressibility of floating-point values.

Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for generating descriptive text for images. In one aspect, a method includes identifying a set of seed descriptors for an image in a document that is hosted on a website. For each seed descriptor, structure information is generated that specifies a structure of the document with respect to the image and the seed descriptor. One or more templates are generated for each seed descriptor using the structure information for the seed descriptor. Each template can include image location information, document structure information, image feature information, and a generative rule that generates descriptive text for other images in other documents. Descriptive text for other images is generated using the templates and the other documents. The descriptive text is associated with the images.

An approach is provided in which a knowledge manager detects an annotation error while attempting to annotate a document and, as such, segments the document into multiple document segments. The knowledge manager then detects another annotation error while attempting to annotate the multiple document segments, resulting in a set of unannotated document segments. In turn, the knowledge manager generates a notification that identifies the set of unannotated document segments.

An in-memory computing method and apparatus, adapted for a processor to perform MAC operations on a memory, are provided. In the method, a format of binary data of weights is transformed from a floating-point format into a quantized format by truncating at least a portion of fraction bits of the binary data and calculating complements of remaining bits, and programming the transformed binary data into cells of the memory. A tuning procedure is performed by iteratively inputting binary data of input signals into the memory, integrating outputs of the memory, and adjusting the weights programmed to the cells based on the integrated outputs. The binary data of the weights is reshaped based on a probability of reducing bits with a value of one in the binary data of each weight. The tuning procedure is repeated until an end condition is met.

Embodiments of the invention may be used to implement a rate converter that includes: 6 channels in forward (audio) path, each channel having a 24-bit signal path per channel, an End-to-end SNR of 110 dB, all within the 20 Hz to 20 KHz bandwidth. Embodiment may also be used to implement a rate converter having: 2 channels in a reverse path, such as for voice signals, 16-bit signal path per channel, an End-to-end SNR of 93 dB, all within 20 Hz to 20 KHz bandwidth. The rate converter may include sample rates such as 8, 11.025, 12, 16, 22.05, 24, 32 44.1, 48, and 96 KHz. Further, rate converters according to embodiments may include a gated clock in low-power mode to conserve power.

Embodiments of the invention may be used to implement a rate converter that includes: 6 channels in forward (audio) path, each channel having a 24-bit signal path per channel, an End-to-end SNR of 110 dB, all within the 20 Hz to 20 KHz bandwidth. Embodiment may also be used to implement a rate converter having: 2 channels in a reverse path, such as for voice signals, 16-bit signal path per channel, an End-to-end SNR of 93 dB, all within 20 Hz to 20 KHz bandwidth. The rate converter may include sample rates such as 8, 11.025, 12, 16, 22.05, 24, 32 44.1, 48, and 96 KHz. Further, rate converters according to embodiments may include a gated clock in low-power mode to conserve power.

A non-transitory computer readable medium storing a program causing a computer to execute a parsing process including extracting one or more figure descriptions from a first entity of document in a sequential order and calculating the number of figure descriptions from the first entity of the document having figure descriptions; extracting one or more drawing pages from a second entity of document in a sequential order and calculating the number of drawing pages from the second entity of the document having drawing pages; computing a varying list of figure descriptions; and combining the varying list of figure descriptions with one of the one or more drawing pages to produce the single output, whereby a user of the single output can readily identify a figure number of the one of the one or more drawing pages and associate the figure number with a description from the varying list of figure descriptions.

A computer program product comprising a computer-readable storage medium having program instructions embodied therewith, the program instructions executable by a processor to cause the processor to receive a document having multiple section headers, segment the document into at least first and second sections based on the section headers, segment items in the first section into fragments and identify a section type for each of the fragments, determine that the identified section type for at least one of the fragments better matches a type of the second section than it matches a type of the first section, and re-locate the at least one of the fragments to the second section.

A method comprising receiving a document having multiple sections of different types using a processor. The method also comprises obtaining a plurality of lexicons using the processor, each of the lexicons for interpreting fragments in one or more of the section types. The method further comprises interpreting fragments in a first section of the multiple sections using the processor and one or more lexicons. The method still further comprises determining, based upon the interpretation and using the processor, that a fragment in the first section is misplaced. The method still further comprises re-locating, using the processor, the misplaced fragment to a second section of the multiple sections in the document to generate a re-organized document. The method additionally includes storing the re-organized document in a hardware storage system using the processor.

A method comprising receiving an item via a computer input device, comparing the item to a plurality of lexicons using a processor, and determining, using the processor and based upon the comparison, that the item is part of a fragment matching at least one of the plurality of lexicons. The method also comprises storing, using the processor, the fragment in a section of a document determined based on the at least one of the plurality of lexicons, and receiving at least one other item via the computer input device after the fragment is stored in the section.