Abstract: A computer-implemented method, includes accessing, by a processor, a set of asset documents. The method also includes performing, by the processor, feature extraction on text of each document of the set of asset documents using a feature extraction module to generate a set of features, where each feature of the set of features represents a document of the set of asset documents. The method also includes generating, by the processor, a set of lower-dimensional features from the set of features using a singular value decomposition module. The method also includes generating, by the processor, a set of clusters from the set of lower-dimensional features using a clustering module. The method also includes training, by the processor, a machine-learning model of a classification microservice using the set of clusters generated from the clustering module.

Abstract: A modular loading system may include structural beams and modular beams that can be located in a shipping container. The structural beams can be arranged in a first direction, and the modular beams can be arranged in a second direction that is approximately perpendicular to the first direction. The structural beams can include a fixed contact surface and an adjustable contact surface positioned opposite the fixed contact surface for stabilizing the modular loading system in approximately the first direction. The modular beams can include receiving notches sized to receive a locking bar for stabilizing the modular loading system in a third direction that is approximately perpendicular to the first direction and the second direction.

Abstract: A tiered data storage management system in which environmental, social, and governance (ESG) factors are used to evaluate and manage storage of physical and corresponding digital assets, particularly (but not limited to) storage of digital assets generated from corresponding physical assets, and particularly (but not limited to) situations in which the cost to store a digital representation of an asset exceeds the cost to store the original physical asset for a given amount of time.

Abstract: In some embodiments, techniques are provided for document identification using a multimodal model that has been trained using one-shot learning. In one example, a first method of document image processing includes generating, for each template document image of a plurality of template document images, a corresponding fingerprint of a plurality of fingerprints; and based on the plurality of fingerprints, training a multimodal model. For each template document image of the plurality of template document images, generating the corresponding fingerprint may include detecting a plurality of regions within the template document image, wherein the plurality of regions comprises a plurality of text regions; and filtering the plurality of regions to obtain a plurality of regions of interest, wherein the fingerprint is based on the plurality of regions of interest.

Abstract: Features described herein generally relate to systems and methods for generating and managing tokens for authenticated assets. In some aspects, generating a token is performed according to one or more rules and includes generating metadata that links the token to a representation of an asset. In some other aspects, tracking token access includes monitoring a blockchain for token access events and generating a token based on a modified historian and received feedback. In other aspects, validating a token for an authenticated asset includes matching representations of the authenticated asset.

Abstract: A method of document image processing comprises, based on at least a document page image, generating a plurality of semantic tokens that includes a plurality of word tokens and a plurality of special tokens. Each special token among the plurality of special tokens represents a non-textual semantic element of the document image, and generating the plurality of semantic tokens includes predicting, for each special token among the plurality of special tokens, a token type of the special token. The method also comprises generating, for each semantic token among the plurality of semantic tokens, a corresponding semantic token embedding among a plurality of semantic token embeddings; and applying a trained model to process an input that is based on the plurality of semantic token embeddings and a plurality of visual token embeddings based on at least the document page image to generate a semantic processing result.

Abstract: A process for document processing (e.g., automated package splitting) may involve producing, for each document page of an ordered plurality of document pages, an image of the document page and a representation of text from the document page; generating, for each document page of the ordered plurality, and based on the image of the document page and the representation of text from the document page, an embedding of the document page; and generating, for each document page among a subset of the ordered plurality, a label for the document page that indicates whether the document page is a document first page, based on the embedding of the document page, the embedding of each of at least one document page that precedes the document page in the ordered plurality, and the embedding of each of at least one document page that follows the document page in the ordered plurality.

Abstract: A process for document processing (e.g., automated package splitting) may involve producing, for each document page of an ordered plurality of document pages, an image of the document page and a representation of text from the document page; generating, for each document page of the ordered plurality, and based on the image of the document page and the representation of text from the document page, an embedding of the document page; and generating, for each document page among a subset of the ordered plurality, a label for the document page that indicates whether the document page is a document first page, based on the embedding of the document page, the embedding of each of at least one document page that precedes the document page in the ordered plurality, and the embedding of each of at least one document page that follows the document page in the ordered plurality.

Abstract: In some embodiments, techniques for self-labeling to extract a representative set of samples from a large-scale set of unlabeled documents (e.g., a set that represents a distribution of the large-scale set) are provided. The samples of the representative set may then be used to classify the documents of the large-scale set.

Abstract: In some embodiments, techniques for knowledge-based validation of entities extracted from digital documents are provided. For example, a process may involve selecting, from among a plurality of entities extracted from a digital document, a first set of correlated entities. Selecting the first set of correlated entities may be based on a correlation that is indicated by relative location of the entities of the first set of correlated entities within the digital document or by similarity among tags of the entities of the first set. The method may also include determining, using a knowledge model, that the first set of correlated entities is not valid; and generating, using the knowledge model, a first modified set of correlated entities, wherein each entity of the first modified set of correlated entities corresponds to a respective entity of the first set of correlated entities.

Abstract: In some embodiments, techniques for document entity extraction are provided. For example, a process may involve processing document images to detect a plurality of regions of interest that includes text objects and non-text objects; for each of the plurality of regions of interest, producing a corresponding text string; and processing the text strings to identify entities. Processing the document images may involve applying a text object detection model to the document images to detect the text objects; and applying at least one non-text object detection model to the document images to detect the non-text objects. Prior to processing the document images, at least two object detection models among the text object detection model and the at least one non-text object detection model were generated by fine-tuning respective instances of a pre-trained object detection model.

Abstract: A computer-implemented method, includes accessing, by a processor, a set of asset documents. The method also includes performing, by the processor, feature extraction on text of each document of the set of asset documents using a feature extraction module to generate a set of features, where each feature of the set of features represents a document of the set of asset documents. The method also includes generating, by the processor, a set of lower-dimensional features from the set of features using a singular value decomposition module. The method also includes generating, by the processor, a set of clusters from the set of lower-dimensional features using a clustering module. The method also includes training, by the processor, a machine-learning model of a classification microservice using the set of clusters generated from the clustering module.

Abstract: Technologies are described herein for providing secure data protection and recovery. A virtual data center can be created on backup servers to manage data backup for a specific client. Backups can be performed by a master server in the virtual data center initiating a secure network connection with a client computing device at a random or pseudo random time within a backup window specified by the client. A backup agent on the client computing device can execute a data mover program complied at the client's computing system but received from the master server to retrieve and save backup data at a random or pseudo random temporary location. The saved data can be transmitted and stored on data storage accessible only by a storage server. Backup data can be inspected in an isolated inspection server and be made retrievable through a secure data repository accessible by the client.

Abstract: Technologies are described herein for providing secure data protection and recovery. A virtual data center can be created on backup servers to manage data backup for a specific client. Backups can be performed by a master server in the virtual data center initiating a secure network connection with a client computing device at a random or pseudo random time within a backup window specified by the client. A backup agent on the client computing device can execute a data mover program complied at the client's computing system but received from the master server to retrieve and save backup data at a random or pseudo random temporary location. The saved data can be transmitted and stored on data storage accessible only by a storage server. Backup data can be inspected in an isolated inspection server and be made retrievable through a secure data repository accessible by the client.

Abstract: In one arrangement, a storage system for storing items includes an enclosure enclosing a hypoxic environment, and a pallet shuttle system comprising rows of racks disposed within the hypoxic environment. The pallet shuttle system also includes at least one transfer car and at least one shuttle for carrying storage modules into and out of the racks. The system further includes a loading station outside the hypoxic environment, and an enclosure door to the hypoxic environment. The enclosure door is operable to admit the transfer car of the pallet shuttle system into and out of the hypoxic environment, and to close the hypoxic environment upon passage of the transfer car through the door. Optionally, the system includes paneling separating the enclosure into at least two separately climate-controlled zones such that at least one of the racks is in each of the at least two separately climate-controlled zones.

Abstract: Methods and systems for managing, tracking, and displaying power usage information in connection with colocation datacenters are described. In particular, one or more current transformers determine an amount of current associated with breaker positions of a breaker panel. The current readings may then be used to construct one or more virtual circuits. Access to the current and power readings associated with the virtual circuit is provided to clients via a communication network. Additionally, the current readings may be utilized to track actual current and/or power use in conjunction with one or more client allocated amounts of power.

Abstract: A tamper resistant box includes: a front panel; a side panel attached to the front panel at a first fold line; a back panel attached to the side panel at a second fold line; a side top flap attached to the side panel at a third fold line; and a front top flap attached to the front panel at a fourth fold line. In some examples, the side top flap can define a tapered slot extending towards the third fold line from an edge of the side top flap at a location distal from the third fold line, and a widest portion of the tapered slot can define a tapered slot width. In further examples, the front top flap can define a top slot having a top slot width that is less than the tapered slot width.

Abstract: A method of remediating a media asset comprising a magnetic tape includes preliminarily cleaning the media asset, treating the media asset, and finally cleaning the media asset. Treating the media asset includes baking the media asset, determining whether adjacent layers of the tape are stuck to each other, and re-baking the media asset. If adjacent layers of the tape are stuck to each other, the method includes submerging the magnetic tape in a cleaning bath for a predetermined period of time, unwinding the magnetic tape from a supply reel to a take-up reel at an unwind speed while drying the magnetic tape, and rewinding the magnetic tape onto the supply reel at a rewind speed before re-baking the media asset.

Abstract: A technique for organizing pathology samples includes sorting the samples into bins defined at different physical locations of a sorting apparatus, based on a set of designations applied to the samples. With such an arrangement, unorganized pathology samples may be grouped so that all the same accession numbers are grouped together for efficient filing or re-filing with reduced error. An apparatus suitable for use according to this technique includes a support structure and multiple dividers defining multiple sets of bins on the apparatus. A different set of bins is provided on the apparatus for each of the set of designations applied to the samples.

Abstract: An improved technique for managing offsite data storage avoids the need for tapes or other media arriving at an offsite storage location to be mounted. The improved technique includes generating a list of data objects stored on each piece of media of a first set of media sent from a first site to a second site. A reply list is generated, which identifies any data objects that did not arrive at the second site. The missing data objects are stored on a second set of media being prepared for shipment from the first site to the second site. The second set of media typically contains not only the missing data objects, but also additional data objects.