Described are approaches for assigning tasks between machine resources (e.g., AI task performers, AI task validators), human resources (e.g., task performers, task validators), and/or other smart systems to facilitate collaborative text detection, text recognition, and text retrieval in order to optimize system performance along a variety of different selection criteria specifying various performant dimensions, including, but not limited to improving system efficiency, reducing task performer and/or task validator idle time, improving triage outcomes, reducing data processing loads, maintaining client confidentiality, etc., that may be associated with one or more customers.

Methods are presented suitable for clustering computational event logs (2) including a method for calculating a metric distance between characters of different event messages (4) by comparing both characters to a comparative set of characters. Methods are presented for calculating a metric distance between two event messages (4) comprising determining character metric distances between characters in the compared words and word metric distances between the words in the compared events (4), Methods are presented for creating an area (8) in metric space corresponding to a new cluster (6) when a further event message (26) is found in an overlap region (24) of existing clusters (6, 8). Methods are presented in populating and constructing an event table.

In a question generating device, a reader reads a document and generates a document image. An extracting section extracts an original form character part on the basis of a marked character part included in the document image. The original form character part represents an original form of a plurality of candidate character parts that are candidate answers to a question generated from the document image. A setting section sets as the plurality of candidate character parts, the original form character part and at least one character part formed by either or both adding a character to and deleting a character from the original form character part. A selecting section selects a candidate character part of the highest priority from among the plurality of candidate character parts. A generating section generates the question using the candidate character part of the highest priority as the answer.

Embodiments of this application disclose a method for retrieving similar videos performed at a computing device. The computing device obtains video information of a video for which similar videos are to be retrieved, the video information including a video tag and a video title, and trains the video information by using a preset text depth representation model, to convert the video information into a word vector. After selecting, from a video library according to a preset knowledge graph, videos matching the video information, to obtain a first candidate video set, the computing device screens, in the video library, videos similar to the video information according to the word vector, to obtain a second candidate video set and then determines a similar video for the video information from the first candidate video set and the second candidate video set.

A multi-pattern matching algorithm may be provided that includes: a moving step of moving a moving window from the start of a string one byte by one byte; a DF1 checking step of converting the string on a current position of the moving window into an integer value, and of checking whether or not a bit of a related position in a first direct filter DF1 for patterns having lengths larger than 2 bytes is set to 1; a DF moving step of checking one or more direct filters DF when the bit is set to 1 according to the DF1 checking step; a re-moving step of moving the moving window by one byte again when the bit of a related position in the direct filter DF, which has been checked lastly, is 0; and a terminating step of checking whether the moving window is located at the end of the string or not, and of terminating the algorithm when the moving window is positioned at the end of the string.

A document processing system is configured to identify, for each accessed electronic document in a first set of multiple electronic documents, a set of identified multi-word phrases determined to be in ordered text information in the accessed electronic document, each multi-word phrase of the set of identified multi-word phrases including adjacent words in the ordered text information; and determine, for each accessed electronic document in the first set of multiple electronic documents, a selected document type from the first set of document types based at least on an analysis of the set of identified multi-word phrases with respect to multi-word-phrase characteristics identified by a first definition and associated with each document type in a first set of document types associated with a first document-set type.

Systems and methods are described for training machine learning models to detect objects in image or video data. A system may select a first sample set of frames from one or more video files. Indications of a location of an object of interest in each of at least two sample frames may be received, then the system may determine the location of the object of interest across a number of intermediary frames using a tracker. Annotation data may be stored identifying the objects of interest in the sample frames, and the annotation data may be used in training a machine learning model to identify the object of interest in subsequently provided image or video data.

Data from one or more sensors is input to a workflow and fragmented to produce HyperFragments. The HyperFragments of input data are processed by a plurality of Distributed Experts, who make decisions about what is included in the HyperFragments or add details relating to elements included therein, producing tagged HyperFragments, which are maintained as tuples in a Semantic Database. Algorithms are applied to process the HyperFragments to create an event definition corresponding to a specific activity. Based on related activity included in historical data and on ground truth data, the event definition is refined to produce a more accurate event definition. The resulting refined event definition can then be used with the current input data to more accurately detect when the specific activity is being carried out.

Systems and methods are described for identifying at least one sports team depicted in media content, such as image or video data. Features of the media content may be provided as input to a first set of classification models that are each trained to identify at least one type of scene associated with one or more sports, then features of the media content may be provided to a second set of classification models trained to identify at least one object associated with one or more sports. Once a sport depicted in the media content is determined based on the first and second set of classification models, the system may determine a team depicted in the media content based at least in part by comparing aspects of the media content to stored data associated with various teams that play the identified sport.

Efficient handwriting input is performed. A character input device includes: a handwritten character input unit that receives a handwritten character input operation; a controller that outputs a prediction candidate in which a replacement character is replaced, the prediction candidate corresponding to a series of handwritten characters, when the replacement character exists in the series of handwritten characters input to the handwritten character input unit; and a candidate display that displays the prediction candidate.