Disclosed are various embodiments for intelligent text recognition based upon a selected pattern detection mode. First, text can be identified in an image. A pattern detection mode can be selected by a user or autonomously. In some instances, the pattern detection mode can be selected based at least in part on a user account. Next, the text can be parsed for occurrences of a pattern associated with the selected pattern detection mode. A list of occurrences of the pattern can be generated from the text and presented to a user. In some instances, a user can train a model to learn a new pattern.

Disclosures herein teach applying a set of sections spanning a down-sampled version of an image of a road-scene to a low-fidelity classifier to determine a set of candidate sections for depicting one or more objects in a set of classes. The set of candidate sections of the down-sampled version may be mapped to a set of potential sectors in a high-fidelity version of the image. A high-fidelity classifier may be used to vet the set of potential sectors, determining the presence of one or more objects from the set of classes. The low-fidelity classifier may include a first Convolution Neural Network (CNN) trained on a first training set of down-sampled versions of cropped images of objects in the set of classes. Similarly, the high-fidelity classifier may include a second CNN trained on a second training set of high-fidelity versions of cropped images of objects in the set of classes.

Aspects of the disclosure relate to providing improved optical character recognition (OCR). An OCR evaluation platform may generate a script for evaluating OCR performance. The platform may generate modified resources by executing OCR applications to modify original resources. Based on executing the script, the platform may generate comparative analysis information based on comparing the modified resources to the original resource. The platform may generate metric scores based on the comparative analysis information. The metric scores may be used to generate visual representations of the performance of different OCR applications. The platform may generate weighted scores representing the performance of different OCR applications. The platform may identify a preferred OCR application for performing a particular operation. The platform may store correlations between preferred OCR applications and corresponding operations. The platform my cause execution of preferred OCR applications when performing corresponding operations, based on the stored correlations.

The present disclosure provides systems, methods, and computer program products for generating a digital representation of a system from engineering documents of the system comprising one or more schematics and a components table. An example method can comprise (a) classifying, using a deep learning algorithm, (i) each of a plurality of symbols in the one or more schematics as a component and (ii) each group of related symbols as an assembly, (b) determining connections between the components and the assemblies, (c) associating a subset of the components and the assemblies with entries in the components table; and (d) generating the digital representation of the system from the components, the assemblies, the connections, and the associations. The digital representation of the system can comprise at least a digital model of the system and a machine-readable bill of materials.

A method receives an electronic image and uses the image as an input to a neural network. Based on a determination that the image represents a document, the method uses the image as an input to another neural network to identify a portion of the document containing an identifier. The method extracts the identifier by performing character recognition on the identified portion and determines whether the identifier is valid by using a validation API to determine whether the identifier is associated with a valid account at an institution. Based on a determination that the identifier is associated with a valid account, the method authorizes a transaction associated with the identifier. Based on a determination that the identifier is not associated with a valid account, the method denies the transaction. The first neural network classifies the electronic image into one of multiple valid document types and an invalid document type.

An information processing device includes: an obtaining unit that obtains a first classification condition for classifying a document by use of image information of an image formed on the document; an acceptance unit that accepts a second classification condition for classifying the document, the second classification condition being defined by a user; and a classification unit that applies the first classification condition and the second classification condition to the image information based on a predetermined rule of a degree of priority, and classifies the document.

The text recognition method includes: acquiring a first high-frequency feature map and a first low-frequency feature map of a target image; performing an M-level convolution process on the first high-frequency feature map and the first low-frequency feature map by M cascaded convolution modules to obtain M pairs of target high-frequency feature map and target low-frequency feature map of the target image, where M is a positive integer; merging the M pairs of target high-frequency feature map and target low-frequency feature map to obtain a target feature map of the target image; determining a probability map and a threshold map of the target image based on the target feature map, and calculating a binarization map of the target image based on the probability map and the threshold map; and determining a text area in the target image based on the binarization map, and recognizing text information in the text area.

Quality associated with an interpretation of data captured as unstructured data can be determined. Attributes can be identified within the unstructured data automatically. Subsequently, sentiment associated with each of the attributes can be determined based on the unstructured data. Correctness of the unstructured data, and thus the interpretation, can be assessed based on a comparison of the attribute and associated sentiment with structured data. A quality score can be generated that captures the quality of the data interpretation in terms of correctness and as well as results of another analysis including completeness, among others. Comparison of the quality score to a threshold can dictate whether or not the interpretation is subject to further review.

An inter-word score calculation apparatus calculates a degree of relatedness between words included in an amount of data from at least one document. The inter-word score calculation apparatus includes a memory storing document data from the documents, term list data wherein predetermined terms are written and a processor. The processor performs a combination process of amplifying an amplification candidate word, which is a word corresponding to a term in the term list data and included in the document data, creating an amplified word, and adding the amplified word to the document data creating processed document data, calculate the degree of relatedness between words included in the processed document data using a predetermined calculation method, and when an amount of documents accumulated in the document data is smaller than a first predetermined amount, add the amplification candidate word to the processed document data.

Disclosures herein teach applying a set of sections spanning a down-sampled version of an image of a road-scene to a low-fidelity classifier to determine a set of candidate sections for depicting one or more objects in a set of classes. The set of candidate sections of the down-sampled version may be mapped to a set of potential sectors in a high-fidelity version of the image. A high-fidelity classifier may be used to vet the set of potential sectors, determining the presence of one or more objects from the set of classes. The low-fidelity classifier may include a first Convolution Neural Network (CNN) trained on a first training set of down-sampled versions of cropped images of objects in the set of classes. Similarly, the high-fidelity classifier may include a second CNN trained on a second training set of high-fidelity versions of cropped images of objects in the set of classes.