17 results about "Context based" patented technology

Various embodiments of the present invention provide methods, apparatus, systems, computing devices, computing entities, and / or the like for performing context-based document classification prediction using a hierarchical attention-based keyword classifier machine learning framework. Certain embodiments of the present invention utilize systems, methods, and computer program products that perform context-based document classification prediction using at least one of techniques using contextual keyword classifications, techniques using attention-based keyword classifier machine learning framework, techniques using a greedy matching indicator, and / or the like.

The subject technology relates to context-based response formulation for retrieval-augmented generation systems. An example method facilitating context-based response formulation for retrieval-augmented generation systems includes supplementing a query provided to a machine learning model with supplemental entity data, resulting in an augmented query, where the supplemental entity data is determined based on context information associated with the query and expected entity types associated with a determined intent of the query. The method can further include determining an estimated degree of error associated with a document retrieved by the machine learning model in response to the augmented query, and facilitating, in response to the estimated degree of error being lower than a threshold degree of error, generating a response to the query based on the document.

A method, non-transitory computer-readable storage medium and system is disclosed for using context-based dictionaries to search through multimedia data using input that specifies tags, words, phrases, descriptions, environments, emotions, sentiments, multimedia objects or content, or other relevant attributes. The system retrieves original content, analyzes and processes it, and presents to the user synchronized multimedia content and text content that is automatically tagged for searching. The system creates dictionaries containing word definitions and information that have been customized according to context; in addition, the system creates textual and non-linguistic attributes that enable and enhance searching functions; moreover, it enables modification of the dictionary entries as well as its searching functions through a feedback loop that may include input from human users and artificial intelligence programs; furthermore, the system may be used to create or modify a linguistic or a multimedia instantiation of a story.

A self-supervised learning framework for empowering instance discrimination in medical imaging using Context-Aware instance Discrimination (CAiD), in which the trained deep models are then utilized for the processing of medical imaging. An exemplary system receives a plurality of medical images; trains a self-supervised learning framework to increasing instance discrimination for medical imaging using a Context-Aware instance Discrimination (CAiD) model using the received plurality of medical images; generates multiple cropped image samples and augments samples using image distortion; applies instance discrimination learning a mapping back to a corresponding original image; reconstructs the cropped image samples and applies an auxiliary context-aware learning loss operation; and generates as output, a pre-trained CAiD model based on the application of both (i) the instance discrimination learning and (ii) the auxiliary context-aware learning loss operation.

A method of generating a hint of a large language model includes receiving (101) a textual user input for controlling a lighting system, obtaining (103) contextual information indicative of one or more contextual characteristics of a context of the user, a subset of configuration data related to the context of the user and / or an example subset of lighting control behaviors related to the context of the user is selected (105) based on the contextual characteristic. The method further includes generating (107) a hint of the large language model and communicating (109) the hint to the large language model. The prompt includes the textual user input and the selected one or more subsets. The prompt excludes configuration data that is not selected a portion of the configuration data subset, and excludes examples that are not selected a portion of the example subset.

This paper discloses the automated application of context-based image editing and personalized visual photograph preferences using artificial intelligence. A method for context-based insertion of three-dimensional (3D) objects into a photograph includes: analyzing the scene of the photograph using a machine learning model to determine the scene's contextual properties. The method also includes: selecting 3D assets from a set of 3D assets based on the contextual properties. The method further includes: posing the 3D assets using geometric information determined from the scene such that the posed 3D assets conform to one or more of the photograph's viewpoint, orientation, or surface geometry. The method also includes: adding 3D elements associated with the scene context, which are selected and positioned according to the contextual properties. The method further includes: generating an output photograph that includes the posed 3D assets and the 3D elements integrated into the photograph.

Techniques are described for maintaining contextual data to support content-related actions. In an example, a system stories second content at a source. The source is associated with first content. The system sends, to a device, an object that indicates the first content. From the device at a first time, the system receives first data indicating a first request for the second content and including source information that indicates the source. From the device at a second time, the system receives second data indicating a second request for the second content, the second data including the source information, the first data and the second data received at a frequency indicated by the object. The system determines that the requests are associated with the first content based on the source information included in the received data, and stores third data indicating a presentation of the first content by the device.

A large model calling method and system based on hierarchical setting of a knowledge graph, comprising the following steps: hierarchically setting a knowledge graph into a domain knowledge layer, an entity object layer and a context prompt layer; the domain knowledge layer comprises professional vocabulary and instruction data; the entity object layer comprises entities, concepts and events; the context prompt layer comprises subgraphs, attributes and rule chains; a large model is called in the following manner: obtaining a question, completing alignment, if an answer can be generated in the domain knowledge layer or the entity object layer, directly generating the answer; if an answer cannot be generated, calling the domain knowledge layer and / or the entity object layer based on the context prompt layer to generate an answer. By adopting the independent layer mode for the knowledge graph, the application can coordinate information calling and expand within limits, thereby reducing the problem of large model hallucination and improving reasoning ability to a certain extent.

The application discloses a skeleton self-supervised behavior recognition method based on multi-task perception diffusion, comprising the following steps: inputting to-be-processed skeleton sequence data into a knowledge bridging multi-task self-supervised learning network trained by adopting a two-stage curriculum learning strategy, so as to extract intermediate semantic features; taking the intermediate semantic features as self-supervised signals to guide feature learning, obtaining skeleton features aligned with the intermediate semantic features, and then recognizing action recognition results; encoding part of the skeleton sequence data to obtain context features, and outputting future skeleton sequence prediction results based on the context features and the intermediate semantic features. The application converts multi-task optimization from direct confrontation to decoupling cooperation by constructing an independent intermediate semantic bridge, eliminates optimization conflicts and gradient competition, breaks through the bottleneck that a single feature is difficult to balance discriminativeness and structuralness by fusing contrastive learning and diffusion models, and improves training efficiency and stability by adopting a two-stage curriculum learning strategy.

A training method of a conversation model according to various example embodiments of the present disclosure may include identifying a first context, identifying a first response set corresponding to the first context based on a first model, identifying a response subset selected from the first response set based on a gold response corresponding to the first context and training a second model based on the first context information and the response subset.

The application provides a question and answer method and system based on a large language model, which comprises the following steps: determining a current question of a user and a natural language understanding prompt word; inputting the current question and the natural language understanding prompt word into a pre-trained large language model to obtain a question understanding result and a question to be answered output by the large language model according to a natural language understanding strategy; in the case that the question understanding result is a factual question, retrieving an answer from a knowledge graph based on semantic analysis; inputting the retrieved answer and an answer verification prompt word into the pre-trained large language model to obtain an answer verification result; in the case that the verification is reasonable, taking the retrieved answer as the final answer to the current question; and in the case that the verification is unreasonable, generating an answer by the large language model. In the multi-round question and answer task, the application can deeply understand the question of the user based on the context, effectively retrieve and integrate information from different knowledge sources, and thus accurately and efficiently provide an answer to the current question.

The application discloses a semantic deviation evolution prediction method and system, and belongs to the technical field of natural language processing and artificial intelligence. The method maps input data to a unified semantic vector space, constructs a literal semantic vector, determines a standard context based on historical corpus and context conditions, and generates a corresponding context intention semantic vector. The semantic deviation between the literal semantic vector and the context intention semantic vector is calculated, and an initial judgment of semantic deviation triggering state is made by using a judgment model trained by contrast learning. For samples judged as semantic deviation, a predictive generation reasoning is performed under the semantic consistency constraint, a potential semantic evolution path is simulated and generated, and a semantic evolution distribution stability is checked. Finally, the semantic deviation, evolution confidence ranking result and context construction basis are output, realizing quantitative judgment, evolution prediction and explainable analysis of semantic deviation.

This invention relates to the field of medical image processing and data compression technology, and discloses a dual-chessboard autoregressive entropy coding method based on context prior learning for lossless compression of medical images. This method performs lossy compression on an input single-channel grayscale medical image to obtain a lossy reconstructed image and its residual image. A context prior learning network is used to extract features and generate context for the lossy reconstructed image, achieving structure-aware modeling of the residual features. Based on a dual-chessboard decomposition autoregressive entropy model, the residual image is divided into complementary sub-images. Pixel-level probability estimation is performed by combining context prior and coded block information, and a residual bitstream is generated through arithmetic coding. Finally, the bits are merged to obtain the lossless compressed bitstream of the medical image. This invention effectively reduces the bit rate while ensuring lossless reconstruction quality, is suitable for lossless compression of 2D and 3D medical images, and features high compression efficiency and strong robustness.

Disclosed herein are systems and methods for generating a prompt with context based on a list of topics generated for responses from large language models (LLMs). In one aspect, the method includes: obtaining a query from a user; generating and transmitting a prompt based on the query for input into a first and second LLMs; obtaining a first selected portion of the first response from the first LLM and at least a second selected portion of the second response from the second LLM; generating a list of topics using a trained topic analysis machine learning model (MLM) to identify topics from the responses; and generating a prompt for input into the third LLM utilizing at least one topic from the list of topics using at least the selected portions from the first LLM or the selected portions from the second LLM.

This invention discloses a robustness enhancement method for large-scale code models based on context processing and post-training, selecting the enhancement method according to the type of model to be enhanced. For post-training enhancement, adversarial perturbation samples are first generated from the given source code using equivalence semantic transformation, greedy search, and simulated annealing. These perturbation samples are then input into the post-training framework to fine-tune the target large-scale language model. For context enhancement, a backup dataset is generated using backup code. The optimal subset of examples most semantically similar to the current problem is selected by calculating similarity. Dead code is then normalized, and the optimal subset of examples is adaptively selected. Contextual prompts are dynamically adjusted based on task complexity and input into the target large-scale language model. Finally, evaluation metrics and the robustly enhanced target large-scale language model are output. This invention supports multiple mainstream programming languages ​​and both open-source and closed-source models, exhibiting good versatility and scalability.