Abstract: Features are disclosed for automatically identifying a speaker. Artifacts of automatic speech recognition (“ASR”) and/or other automatically determined information may be processed against individual user profiles or models. Scores may be determined reflecting the likelihood that individual users made an utterance. The scores can be based on, e.g., individual components of Gaussian mixture models (“GMMs”) that score best for frames of audio data of an utterance. A user associated with the highest likelihood score for a particular utterance can be identified as the speaker of the utterance. Information regarding the identified user can be provided to components of a spoken language processing system, separate applications, etc.

Abstract: Features are disclosed for automatically identifying a speaker. Artifacts of automatic speech recognition (“ASR”) and/or other automatically determined information may be processed against individual user profiles or models. Scores may be determined reflecting the likelihood that individual users made an utterance. The scores can be based on, e.g., individual components of Gaussian mixture models (“GMMs”) that score best for frames of audio data of an utterance. A user associated with the highest likelihood score for a particular utterance can be identified as the speaker of the utterance. Information regarding the identified user can be provided to components of a spoken language processing system, separate applications, etc.

Abstract: Features are disclosed for automatically identifying a speaker. Artifacts of automatic speech recognition (“ASR”) and/or other automatically determined information may be processed against individual user profiles or models. Scores may be determined reflecting the likelihood that individual users made an utterance. The scores can be based on, e.g., individual components of Gaussian mixture models (“GMMs”) that score best for frames of audio data of an utterance. A user associated with the highest likelihood score for a particular utterance can be identified as the speaker of the utterance. Information regarding the identified user can be provided to components of a spoken language processing system, separate applications, etc.

Abstract: Features are disclosed for automatically identifying a speaker. Artifacts of automatic speech recognition (“ASR”) and/or other automatically determined information may be processed against individual user profiles or models. Scores may be determined reflecting the likelihood that individual users made an utterance. The scores can be based on, e.g., individual components of Gaussian mixture models (“GMMs”) that score best for frames of audio data of an utterance. A user associated with the highest likelihood score for a particular utterance can be identified as the speaker of the utterance. Information regarding the identified user can be provided to components of a spoken language processing system, separate applications, etc.

Abstract: Features are disclosed for automatically identifying a speaker. Artifacts of automatic speech recognition (“ASR”) and/or other automatically determined information may be processed against individual user profiles or models. Scores may be determined reflecting the likelihood that individual users made an utterance. The scores can be based on, e.g., individual components of Gaussian mixture models (“GMMs”) that score best for frames of audio data of an utterance. A user associated with the highest likelihood score for a particular utterance can be identified as the speaker of the utterance. Information regarding the identified user can be provided to components of a spoken language processing system, separate applications, etc.

Abstract: Features are disclosed for automatically identifying a speaker. Artifacts of automatic speech recognition (“ASR”) and/or other automatically determined information may be processed against individual user profiles or models. Scores may be determined reflecting the likelihood that individual users made an utterance. The scores can be based on, e.g., individual components of Gaussian mixture models (“GMMs”) that score best for frames of audio data of an utterance. A user associated with the highest likelihood score for a particular utterance can be identified as the speaker of the utterance. Information regarding the identified user can be provided to components of a spoken language processing system, separate applications, etc.

Abstract: Features are disclosed for automatically identifying a speaker. Artifacts of automatic speech recognition (“ASR”) and/or other automatically determined information may be processed against individual user profiles or models. Scores may be determined reflecting the likelihood that individual users made an utterance. The scores can be based on, e.g., individual components of Gaussian mixture models (“GMMs”) that score best for frames of audio data of an utterance. A user associated with the highest likelihood score for a particular utterance can be identified as the speaker of the utterance. Information regarding the identified user can be provided to components of a spoken language processing system, separate applications, etc.

Abstract: Features are disclosed for automatically identifying a speaker. Artifacts of automatic speech recognition (“ASR”) and/or other automatically determined information may be processed against individual user profiles or models. Scores may be determined reflecting the likelihood that individual users made an utterance. The scores can be based on, e.g., individual components of Gaussian mixture models (“GMMs”) that score best for frames of audio data of an utterance. A user associated with the highest likelihood score for a particular utterance can be identified as the speaker of the utterance. Information regarding the identified user can be provided to components of a spoken language processing system, separate applications, etc.

Abstract: Features are disclosed for generating acoustic models from an existing corpus of data. Methods for generating the acoustic models can include receiving at least one characteristic of a desired acoustic model, selecting training utterances corresponding to the characteristic from a corpus comprising audio data and corresponding transcription data, and generating an acoustic model based on the selected training utterances.

Abstract: Neural networks may be used in certain automatic speech recognition systems. To improve performance of these neural networks, they may be updated/retrained during run time by training the neural network based on the output of a speech recognition system or based on the output of the neural networks themselves. The outputs may include weighted outputs, lattices, weighted N-best lists, or the like. The neural networks may be acoustic model neural networks or language model neural networks. The neural networks may be retrained after each pass through the network, after each utterance, or in varying time scales.

Abstract: Efficient coordination of processing (by desulphurizing) and moving hot metal from a direct smelter, producing hot metal on a continuous basis, to an electric arc furnace or furnaces, operating on a batch basis, is disclosed. The invention includes the use of hot metal storage devices, such as ladles, that are large enough to supply hot metal for a small number, preferably two or three, of electric arc furnace batch operations.

Abstract: Cooling panels (31) are attached to the shell (11) of a metallurgical vessel to form an internal lining of the shell. Each panel (31) comprises a coolant flow tube (36) bent to form inner and outer panel sections (37, 38) of zig-zag formation. Panel mounting pins (43) connected to the outer panel section (38) by connector straps project laterally outwardly from the panel through openings (45) in the shell and tubular shell protrusions (46) surrounding the openings (45). The ends of pins (43) are connected to the outer ends of protrusions (46) by welding metal discs (47) thus forming connections exteriorly of the shell in a way which seals the openings (45). Coolant inlet and outlet connectors (42) for the panel project outwardly through openings (48) in the shell and surrounding tubular protrusions (49) and connections are made by welding discs (51) between the connectors (42) and protrusions (49).

Abstract: Efficient coordination of processing (by desulphurising) and moving hot metal from a direct smelter, producing hot metal on a continuous basis, to an electric arc furnace DIRECT or furnaces, operating on a batch basis, is disclosed. The invention includes the use of hot metal storage devices, such as ladles, that are large enough to supply hot metal for a small number, preferably two or three, of electric arc furnace batch operations.

Abstract: Cooling panels (31) are attached to the shell (11) of a metallurgical vessel to form an internal lining of the shell. Each panel (31) comprises a coolant flow tube (36) bent to form inner and outer panel sections (37, 38) of zig-zag formation. Panel mounting pins (43) connected to the outer panel section (38) by connector straps project laterally outwardly from the panel through openings (45) in the shell and tubular shell protrusions (46) surrounding the openings (45). The ends of pins (43) are connected to the outer ends of protrusions (46) by welding metal discs (47) thus forming connections exteriorly of the shell in a way which seals the openings (45). Coolant inlet and outlet connectors (42) for the panel project outwardly through openings (48) in the shell and surrounding tubular protrusions (49) and connections are made by welding discs (51) between the connectors (42) and protrusions (49).

Abstract: A negative pressure control system for protection from airborne asbestos particles, gases and the like from a work space includes an enclosure formed of portable rigid doorway panels and doors with transoms communicating with the work space and defining an air intake and decontamination chambers, an exhaust unit having a blower and HEPA filter is positioned in the work space and discharges to the atmosphere and maintains a suitable negative air pressure in the work space and a substantial flow of air through the work space and the decontamination chambers. The enclosure may comprise a room divider and/or a temporary room in communication with the work space. The doors are normally closed as by spring action and the transom closures open responsive to air flow towards the work space and closes with the loss of negative air pressure within the work space.

Abstract: A reflex light reflector comprising a laminate of water-insoluble layers including an internal specular-reflecting metal layer sealed between water-insoluble layers and formed in situ by vapor deposition of metal to a sufficient thickness to constitute a continuous opaque electrically conductive film, and a monolayer of microsphere lens elements of a diameter within the range of 10 to 200 microns overlying the metal layer and bonded by resinous bonding material in optical relationship thereto for reflex light reflection, wherein said metal layer is characterized by being multiply fractured in a random pattern resembling the cracking pattern of a dried mud flat such that it consists of a multiplicity of non-overlapping patches of metal, each fractionally separated from others sufficiently to render said metal layer highly resistant to electrolytic corrosion, each said patch having an area size sufficient to underlie from one up to about thirty of said lens elements, with each said patch having at least one dep

Abstract: An elastic joint-spanning waterstop of particular design for use in a method of sealing joints, in particular expansion joints between concrete sections in roads, bridge and parking deck structures against passage of water, dirt, etc., is described. In the method, recesses are formed in the concrete surface on each side of the joint and the specially designed flexible elastic joint-spanning waterstop element is placed therein, and thereafter a layer of rigid-setting grout is then placed over the element. The waterstop element is in the form of a continuous strip having in cross-section a center portion for location in the joint, and lateral joint-spanning web portions. The webs have a number of openings therethrough to admit fluid grout therein. An upstanding rib is placed between the openings and the center section and is spaced therefrom.

Abstract: A method of sealing joints e.g. between concrete sections in road, bridge and parking deck structures against passage of water, dirt, etc., is described wherein recesses are formed in the concrete surface on each side of the joint and a specifically designed flexible elastic joint-spanning strip having openings therethrough is placed therein. A layer of rigid-setting grout is then placed over the strip, the openings in the strip when filled with grout serving to support the upper layer of grout and to prevent flexing and cracking thereof by vehicle traffic passing thereover, as well as to anchor the strip to the concrete bodies.