British Telecommunications public limited company Intellectual Property Department

A dynamic overlay network system comprising: a plurality of decoy resources deployed over an existing network infrastructure, the existing network infrastructure comprising one or more host network machines; wherein the plurality of decoy resources is arranged to run in a plurality of containers; the plurality of containers is arranged to run on the one or more host network machines; and containers of the plurality of containers are arranged to communicate with each other whereby to mimic operational network traffic.

A method of identifying architectural components in a legacy software system, comprising extracting structural and/or functional relationship between software components from the legacy software system, applying predefined semantic markers derived from historical development data to classify software components, and generating an output indicating the classification for use in system refactoring or migration. A system comprising a processor and a data storage module configured to store source code of the legacy software system, predefined semantic markers, extracted relationships, and classification results, wherein the processor extracts relationships, evaluates them against predefined markers, identifies and classifies software components, and generates an output indicating classified components and their roles. The system facilitates structured analysis of legacy software, applying predefined semantic markers from historical development data to automate classification and role assignment in legacy software systems, improving software migration and refactoring efficiency.

A computer-implemented method comprising: receiving a prompt for an artificial intelligence, Al, model, the prompt comprising a sequence of input tokens; obtaining data indicative of a read rate; obtaining a matrix embedding of at least some of the input tokens of the sequence of input tokens; obtaining a positional encoding matrix which encodes the position of each of at least some of the input tokens within the sequence of input tokens; and combining the matrix embedding and the positional encoding matrix to generate an encodings matrix. The encodings matrix encodes only: a first subset of the sequence of input tokens, wherein the first subset comprises fewer tokens than the sequence of input tokens; and a position of each of the first subset of input tokens within the sequence of input tokens. The first subset of the sequence of input tokens is determined based on the read rate.

A computer-implemented method comprising: accessing source code for a software system from a codebase, the software system comprising a plurality of software components, each software component of the plurality of software components having an interface with at least another software component of the plurality of software components; receiving one or more pre-defined inputs for a first software component of the plurality of software components; obtaining a hypothesized model of at least a part of the software system, the hypothesized model defining one or more hypothesized interactions between the software components during processing of the one or more pre-defined inputs by the plurality of software components; obtaining one or more log statements relating to one or more historic executions of at least one software component of the plurality of software components, each log statement comprising a label indicative of the software component that generated it; identifying, using the one or more log statements and for each input from the one or more pre-defined inputs, one or more interactions between the software components; and generating, based on the identified one or more interactions, a respective association score for each of the one or more hypothesized interactions.

A method of managing content delivery to a client device by way of a proxy and a multicast transmitter in a hybrid unicast/multicast content delivery system. A plurality of proxies are used to handle unicast GET requests for content segments from its associated client device. Those requests can be forwarded over unicast directly to a unicast content server before receiving the response over unicast and forwarding that to the client device over unicast. Alternatively, those GET requests can be forwarded to the multicast transmitter where they can trigger a multicast response for delivering the requested content segment on a multicast channel. The multicast transmitter can be configured to send over the multicast channel to connected proxies an advance notification that the content segment will be transmitted over the multicast channel shortly. Proxies receiving the advance notification do not need to send any GET requests for that segment to the multicast transmitter, but instead send a HEAD request directly to the content server, in the knowledge that the content segment will be received over multicast shortly.

The present invention relates to a computer-implemented method and system for semi-automated schema matching in a data processing environment. Schema data from at least one data source is received and normalised into a domain-level abstraction using a predefined transformation rule. At least one schema matching algorithm analyses syntactic and semantic similarities between schema elements to generate candidate correspondences. A confidence score is computed for each correspondence using a machine learning-based confidence model or a fuzzy Boolean-based confidence model, and a subset of correspondences are presented to a user via an interactive interface for validation. The schema matching model is iteratively updated by propagating semantic context from at least one user-confirmed mapping, refining the model to improve future schema matching predictions. Semantic context derived from user-confirmed mapping is used to identify unmapped schema elements that share a base type with confirmed elements, adjusting confidence scores for those unmapped elements. Validated schema mappings are stored in a repository for later retrieval and use. The system further supports re-evaluating schema correspondence based on updated schema structures, ensuring schema mappings remain consistent as schemas evolve.

A computer-implemented method of validating a sensor of a plurality of sensors monitoring the health of a subject. The method comprises obtaining sensor data and reference data for a first sensor of the plurality of sensors and comparing the sensor data with the reference data. A ticket, comprising an assigned status, is generated for the first sensor based on a result of the comparing. The assigned status is validated by comparing the sensor data to additional sensor data from a second sensor of the plurality of sensors. A certificate of the first sensor is updated with the ticket comprising the validated status of the first sensor.

A method of managing content delivery to a client device by way of a proxy and a multicast transmitter in a hybrid unicast/multicast content delivery system. A plurality of proxies are used to handle unicast GET requests for content segments from its associated client device. Those requests can be forwarded over unicast directly to a unicast content server before receiving the response over unicast and forwarding that to the client device over unicast. Alternatively, those GET requests can be forwarded to the multicast transmitter where they can trigger a multicast response for delivering the requested content segment on a multicast channel. The multicast transmitter can be configured to count the number of GET requests received for a segment and compare that to what the multicast transmitter needs, and send an indicator, such as a probability or range of delay values, based on that comparison to the proxies. This indicator is then used by proxies to determine when and how to transmit a GET request for a subsequent segment to the multicast transmitter or to transmit it as a HEAD request to the content server.

A method (300), and a data processing system, DPS (110), for determining an updated software architecture (112) from an initial software architecture. The method comprises receiving (302), at the DPS (110), data (102) specifying the initial software architecture; receiving (304), at the DPS (110), a Software Quality Model, SQM (104), that defines attributes for assessing the quality of a software architecture for a system; receiving (306), at the DPS (110), a set of desired attributes for the updated software architecture (112), the desired attributes being attributes specified in the SQM (104); modelling (308), by the DPS (110), one or more transformed software architectures (320, 330, 340) based on the initial software architecture and one or more architectural design patterns, wherein the one or more architectural design patterns are stored in a memory location (108) that is communicatively coupled to the DPS (110); calculating (310), by the DPS (110), a quality attribute (321, 331, 341) for each of the transformed software architectures (320, 330, 340) using the SQM (104) and the set of desired attributes; selecting (312), by the DPS (110), one of the transformed software architectures based on the quality attributes; and outputting (316), by the DPS (110), the selected software architecture as the updated software architecture (112) for the system.

There is herein described a method of sensing, the method comprising performing an optical sensing operation on an optical fibre, wherein, if the optical sensing operation indicates the presence of a disturbance in the vicinity of the optical fibre, the method further comprises performing a distributed acoustic sensing operation on the optical fibre.