59 results about "Generative process" patented technology

A method of generating outputs in LLMs including receiving including textual content, defining a context for the documents including identifying a topic or a category, segmenting the textual content into content chunks associated with the topic or category, assigning a tag to each content chunk, identifying selected chunks, adding metadata to the selected chunks indexing the selected chunks, receiving a query, and performing a response generation process including determining if a cache includes information for the query and either retrieving the information from the cache or performing a search of the index to retrieve the information, generating an augmented query by augmenting the query with retrieved information, generating a response from the augmented query, evaluating the response for compliance with criteria, and one of generating a final response and transmitting the final response to the user or performing a fine-tuning process comprising redefining the of the one or more contexts.

A system and method for generating multimedia artifacts with managed scene continuity in visual and multimedia using an AI-based and scene continuity aware media generation platform. The system receives a user or AI agent specification or simulation result(s), selects or trains generative models based on the specification, preprocesses relevant data, and generates scene narrative or frame-specific, sequence specific or broader continuity aware content using the selected or trained model(s). The generated content may be further enhanced using frame interpolation and view synthesis techniques to create smooth transitions or novel viewpoints or to aid in more efficient transmission or viewing or persistence of resultant content. The system enables efficient and customizable generation of high-quality scene continuity aware content for various applications in visual and multimedia production using neuro-symbolic and simulation enhanced compression, representation and generation processes.

A computer program comprising instructions which, when executed by a computer, cause the computer to perform the following operations: Receiving an input document (104), wherein the input document (104) contains content elements including at least one text element; Performing a lossless hierarchical text processing process (106) in which the input document (104) is reproduced as a hierarchical representation (108) in which all text elements from the input document (104) are preserved verbatim and organized according to a structure of the input document (104), the lossless hierarchical text processing process (106) comprising: Prompting a generative language model to generate the hierarchical representation (108), wherein the prompt instructs the generative language model to generate the hierarchical representation (108) without summarizing or omitting any of the text elements; and Performing a retrieval unit generation process (110) in which retrieval units (112) are generated based on the hierarchical representation (108) and stored in a memory accessible by a generative client language model for retrieval-enhanced generation.

An apparatus includes at least one processing device including a processor coupled to a memory, wherein the at least one processing device is configured to modify first data obtained from one or more sources, wherein the modifying includes adding user context data to the first data to generate second data, the second data representing the first data supplemented with a per-user context and, in response to receipt of a query, generate a response to the query using at least one generative language model supplemented by a retrieval augmented generation process based on at least a portion of the second data.

The invention discloses a multi-modal-driven game 3D art resource stylization generation method, and relates to the technical field of artificial intelligence and game development, the method comprises the following steps: S1, obtaining multi-modal input data: receiving and analyzing various combined data input by a user, the data comprises but is not limited to natural language texts for describing roles and scenes, 2D concept original pictures and reference art style picture sets; according to the method, multi-modal input data is combined with feature extraction and semantic understanding, an automatic process from a concept to 3D model generation is realized, manual operation links are reduced, the production cycle of game 3D art resources is remarkably shortened, and the production efficiency is improved by automatically generating the 3D model, performing stylized rendering and optimizing engine suitability. Dependence on multiple professional workers is reduced, development cost is reduced, the reference art style picture set serves as style constraint, style consistency is kept in the generation process, and it is ensured that the generated 3D model and the overall art style of the project are highly unified.

The invention belongs to the technical field of dialogue generation, and relates to a digital intelligent carrier pigeon WeChat intelligent dialogue generation method based on deep learning. Dialogue data are obtained, and semantic anomaly detection is carried out through a regular expression and a BERT model in combination with a public corpus; organizing dialogue data in a mode of constructing a tree structure and introducing a three-dimensional timestamp; for emotion information of emoticons, text and visual features of expressions are combined through a hybrid encoder, a dynamic attention mechanism is adopted, weighting and memory length adjustment are carried out according to importance of historical statements, after user portrait features are introduced, user features and dialogue codes are fused, and therefore, the emotion information of the expressions is obtained. In the generation stage, through a Transform-based model and a Beam Search algorithm, a diversity penalty term is introduced in the generation process, and generation parameters are dynamically adjusted according to a user portrait, so that the diversity and security of generated dialogues are ensured, the precision and coherence of a dialogue generation model are improved, and the technical problems of context coherence and personalized generation are effectively solved.

The invention belongs to the technical field of natural language processing, provides a multi-dimensional confidence fusion large language model uncertainty evaluation method and system, designs a multi-dimensional confidence modeling mechanism, and can perform multi-dimensional confidence modeling on the basis of internal information in a large language model generation process without depending on an external knowledge base. And the reliability of the output content is effectively judged. According to the system, on the basis of a semantic clustering mechanism, a Token-level multi-dimensional confidence modeling method is innovatively introduced, a scoring system fusing factors such as probability centrality, context disturbance sensitivity, generation consistency and language rationality is constructed, the confidence structure of each part of content in model output can be evaluated from the Token level, and the evaluation efficiency is improved. And the discrimination capability of the system on the uncertainty difference in the generation process is obviously improved.

The invention provides a product design generation method, system and device based on a knowledge graph and a text graph model and a medium, and the method comprises the steps: receiving an object input by a user, and converting the input object into a root node of a to-be-constructed knowledge graph; constructing a knowledge graph based on the root node; extracting attribute values of leaf nodes in the knowledge graph, and expanding the attribute values through a large language model to generate image generation cue words; and generating a design corresponding to the object through a text graph model based on the image generation cue word. Through deep combination of the knowledge graph and the text graph model, the problem that a traditional AI generation method only depends on a single cue word and lacks systematic creative expansion capability is solved, and the design generation process is more hierarchical, logical and controllable. Automatic creative divergence is supported, high-quality and controllable cue words are dynamically generated in combination with a knowledge graph, a user is supported to automatically generate new design schemes at different levels, and inspiration exploration is more efficient.

A method for saving prompts on a page of a workspace includes receiving a first input that instantiates a prompt block configured to initiate a generative process to create in-block content. The prompt block is embedded on a first page of a workspace configured to include multiple pages. The first input includes a string of text. The string of text includes instructions that cause a generative artificial intelligence (AI) system to create the in-block content to be displayed on the page. The method includes receiving a second input on the prompt block. In response to the second input, the method includes saving the string of text as a particular prompt. The particular prompt is accessible from the multiple pages of the workspace and the particular prompt is configured to be shared with users of the workspace.

A method for creating in-block content presented in a block on a page of a workspace. The block is configured to initiate a generative process to create in-block content of a particular type. The method includes determining a selection of in-page content based on a location of the block relative to the in-page content and the particular type of in-block content. The method can include causing a generative function to create generative content of the particular type based on the selection of the in-page content. The method can further include populating a block area to present the generative content.

A method for suggesting prompts on a page of a workspace includes receiving an input and displaying a prompt block configured to initiate a generative process to create in-block content in response to the input. The prompt block is embedded as an in-page object on the page. The method includes causing a large language model (LLM) system to create a set of suggested prompts. Each prompt includes instructions configured to create generative content of a respective type by the LLM system. The set of suggested prompts is created based on in-page text content or a relative location of the prompt block on the page. The method includes displaying the set of suggested prompts as a set of control items of the workspace. Each of the set of control items is selectable to input as a prompt for generating content based on existing content of the workspace.

Provided is a method for modifying a machine-learning model. The method includes performing, by a machine learning model, a generative process to predict a first output, generating, via a processor, a latent space based on an input to the machine learning model, determining, via the processor, an intermediate decision parameter based on the latent space, based on the intermediate decision parameter, changing, via the processor, a structure of the machine learning model to generate a modified machine learning model to perform a modified generative process that is conditioned upon the intermediate decision parameter, and generating, by the modified machine learning model, a second output including content associated with the input.

The invention discloses a question and answer interaction method, device and equipment based on artificial intelligence and a medium, and relates to the technical field of artificial intelligence in the professional service fields of finance, insurance, medical treatment, banking and the like, and the method comprises the steps: segmenting a historical dialogue into segments with coherent themes, and constructing the segments into memory units; generating a semantic embedding vector and a time decay weight for each unit to form a memory bank; retrieving related memory units from a memory bank based on user query, comprehensive semantic relevance, theme consistency and time decay weight; the query and related units are combined into a context input generative language model, and a topic consistency constraint is applied during the generation process to generate a topic consistent reply. By optimizing the memory granularity and enhancing the memory representation and timeliness, multi-dimensional accurate retrieval and constraint generation are realized, and the system improves the accuracy, coherence and service consistency of question and answer interaction.

A method for suggesting prompts on a page of a workspace includes receiving an input and displaying a prompt block configured to initiate a generative process to create in-block content in response to the input. The prompt block is embedded as an in-page object on the page. The method includes causing a large language model (LLM) system to create a set of suggested prompts. Each prompt includes instructions configured to create generative content of a respective type by the LLM system. The set of suggested prompts is created based on in-page text content or a relative location of the prompt block on the page. The method includes displaying the set of suggested prompts as a set of control items of the workspace. Each of the set of control items is selectable to input as a prompt for generating content based on existing content of the workspace.

A method for suggesting prompts on a page of a workspace includes receiving an input and displaying a prompt block configured to initiate a generative process to create in-block content in response to the input. The prompt block is embedded as an in-page object on the page. The method includes causing a large language model (LLM) system to create a set of suggested prompts. Each prompt includes instructions configured to create generative content of a respective type by the LLM system. The set of suggested prompts is created based on in-page text content or a relative location of the prompt block on the page. The method includes displaying the set of suggested prompts as a set of control items of the workspace. Each of the set of control items is selectable to input as a prompt for generating content based on existing content of the workspace.

A system and method for generating multimedia artifacts with managed scene continuity in visual and multimedia using an AI-based and scene continuity aware media generation platform. The system receives a user or AI agent specification or simulation result(s), selects or trains generative models based on the specification, preprocesses relevant data, and generates scene narrative or frame-specific, sequence specific or broader continuity aware content using the selected or trained model(s). The generated content may be further enhanced using frame interpolation and view synthesis techniques to create smooth transitions or novel viewpoints or to aid in more efficient transmission or viewing or persistence of resultant content. The system enables efficient and customizable generation of high-quality scene continuity aware content for various applications in visual and multimedia production using neuro-symbolic and simulation enhanced compression, representation and generation processes.

This invention belongs to the field of computer vision technology, specifically a tree-structured video description generation method based on perceptual grammar knowledge. This invention explicitly utilizes semantic information inherent in language, using dependency structure analysis tools to transform sequential sentences into a syntax tree structure. By analyzing the connections between parent and child nodes in the tree, the dependency grammar structure within the sentence is explicitly modeled. A perceptual context attention network is used to model the contextual information generated along different paths during the generation process. Simultaneously, reinforcement learning and iterative generation training methods are introduced during the training phase to further improve model performance. Qualitative and quantitative experiments demonstrate that the model has the ability to generate more accurate and semantically richer descriptions.

The invention discloses a self-adaptive writing method and system based on support style feature extraction and template generation, and particularly relates to the technical field of self-adaptive writing. According to the method, the writing system has the active recognition and selection capability on the target style in the generation process, style drifting caused by context accumulation in long text and multi-round generation is avoided, meanwhile, a fixed semantic structure and a variable expression mode are combined based on a template constraint model automatically generated in a dominant style interval, and the method is more efficient and efficient. On the premise of ensuring logic integrity and expression normalization, the flexibility and adaptability of the writing process are improved, and a closed-loop self-adaptive regulation and control mechanism is formed by performing literary and physical feature consistency evaluation on the generated text and dynamically adjusting the template slot position weight.

The invention relates to the technical field of font generation in the field of computer mode recognition, in particular to an attribute-controllable font generation method based on a diffusion model. Preparing data including font images and font attributes; performing data preprocessing based on the font image and the attribute to obtain a standardized data set; a generation network based on a pre-training potential spatial diffusion model is constructed, a content encoder is introduced to extract font content features, an attribute encoder maps attribute vectors, an attribute control adapter is designed to integrate attribute information into the generation process, and effective decoupling of font content and style attributes is achieved; and finally, generating a high-quality font with a specified style by inputting a font content image and a target attribute vector, and supporting style conversion between any fonts based on an attribute value. The problems of high design cost, single style, inflexible attribute control and the like in the traditional font generation technology can be solved.

The present invention relates to a device and method for generating context-aware motion using text and emotion embeddings. An electronic device according to one embodiment of the present invention includes a memory and a processor connected to the memory, and the processor receives input information including a text prompt instructing the content of a character's action and emotion information to be assigned to the action for generating three-dimensional motion from a user terminal, and vectorizes the received text prompt and the emotion information, respectively, to generate a text embedding for the content of the action and an emotion embedding that determines the nuance of the action, and generates a motion sequence based on the text embedding using a Generative AI model, but reflects the emotion embedding as a condition of the generation process to assign an emotional context to the character's action, and can generate the motion sequence with the assigned emotional context as final motion data.

The invention discloses a Chinese text style migration method based on a diffusion model. According to the method, through multi-level style feature capture and accurate control, text style migration is carried out on the basis of ensuring original semantic content. The method comprises the following specific steps: firstly, preprocessing an input Chinese text, and mapping the input Chinese text to a continuous vector space by utilizing a pre-training language model to generate an initial representation of the text; then, multi-level style deviation items are introduced in the diffusion process, and vocabulary, syntax and semantic style levels of the text are controlled respectively; and in each time step, fine control of the style characteristics is realized by adjusting the control parameters of the style deviation item. Then, in a denoising stage, semantic embedding of the text is optimized by utilizing a semantic preserving mechanism, and the generated text semantics are ensured to be consistent with the original text; besides, the method is also combined with a DPM-Solver-v3 acceleration algorithm, so that sampling steps in the diffusion process are reduced, and the generation efficiency is remarkably improved. Finally, through multi-task joint training, style migration and semantic preserving tasks are balanced, and the generation effect is optimized. According to the method, the semantic consistency of the text can be kept while the style characteristics are finely controlled in a multi-level mode, the natural and smooth text conforming to the target style is generated, and the method is widely applied to the fields of text generation, sentiment analysis, advertisement copywriting generation and the like. In addition, the efficiency of the generation process is improved, the calculation overhead is reduced, and the method is suitable for large-scale text style migration tasks in practical application.

The invention discloses an AI teaching interaction method based on multi-modal emotion perception and generative strategy optimization, and the method comprises the steps: generating a comprehensive emotion state vector in a robust manner through multi-modal emotion analysis with uncertainty estimation and time sequence fusion based on attention; contexts such as teaching content semantics and cognitive load are combined; a generative model with strategy elements as a vocabulary is utilized, a composite interaction strategy sequence is generated through dynamic combination, and cognitive load constraint is carried out in the generation process; optimization is carried out through reinforcement learning integrating off-line evaluation and safety exploration, and strategy rehearsal is optionally carried out through a digital twin model. Deep understanding of student states and highly personalized and safe strategy generation and optimization in a complex and real teaching scene are realized, and the intelligent level and teaching effect of an AI teaching system are significantly improved.

Exact and interpretable computation is integrated into a large language model (LLM)- based generative process using an LEM in combination with a symbolic architecture that performs arithmetic, or other mathematical or domain-specific processing. Hidden states of a ELM are processed to control this symbolic architecture, and inputs to the symbolic architecture are automatically extracted from already generated words. A selection is made during an autoregressive generative process as to whether to use the ELM output or the output of the symbolic architecture to extend the word sequence. Because the use of the symbolic architecture is part of the autoregressive process, the computed result is incorporated into the generation of future words in the result. Advantages of this approach include reduced computation and higher accuracy, for example, by avoiding arithmetic or computation errors manifested in the generated word sequence.

The invention discloses an ancient language machine translation method for syntactic perception and knowledge enhancement, and belongs to the technical field of natural language processing. Aiming at the difficulties of complex sentence patterns and scarcity of historical knowledge of ancient Chinese, three innovation modules are designed: firstly, a dynamic dependency syntactic analysis module explicitly models an ancient Chinese syntactic structure by utilizing probability analysis and a graph network; secondly, a retrieval enhancement generation module accurately extracts related knowledge fragments from the high-quality ancient language corpus, and semantic comprehension is enhanced; and finally, the stream grammar constraint decoder fuses the source language syntax and the target language generation process through a double-stream mechanism and a stream mechanism, and loyalty and smooth modern text output is realized. The three modules are deeply fused, so that the accuracy and interpretability of ancient language translation are effectively improved, and a reliable technical path is provided for ancient book digitization.

The invention relates to the technical field of artificial intelligence, and discloses a music generation model training method, which comprises the following steps: S1, extracting harmony, rhythm and curve form theoretical characteristics of an input music sequence; s2, constructing a dynamic music theory constraint field, and fusing the spatio-temporal evolution law of the theoretical characteristics; s3, designing a gating constraint generator, and injecting the constraint field state into a music generation process; s4, establishing a dynamic optimization objective function containing theoretical constraint terms; s5, generating and screening a theoretical boundary sample through reverse disturbance; s6, implementing a double-loop iteration mechanism of generation training and theoretical evolution; s7, realizing generation capability migration of the cross-music theory system; the S1 comprises the following steps: extracting chord features by adopting a two-way circulation network; and coding the rhythm mode through depth separable convolution. According to the method, self-adaptive balance of theoretical guidance and artistic innovation is realized through a constraint field and chaotic scheduling, and the dual problems of stiff generated music and rule conflicts are solved.

The present application discloses a processing method, device and medium for question prompts of a large language model, and relates to the technical field of natural language processing. The method includes: obtaining the original text, extracting the entities of the original text, and extracting logical relationships based on the entities, the logical relationships including causal relationships and employment relationships; reconstructing the question according to the entities and logical relationships, and generating restated question prompts; based on the question prompts, using an adaptive random replacement strategy, calculating the semantic similarity and task type between the question prompts and the preset vocabulary, dynamically selecting a replacement mechanism, and generating multiple candidate prompts; calculating the global entropy value of the candidate prompts, and performing multi-source knowledge base verification on the candidate prompt with the smallest global entropy value, and outputting the smallest candidate prompt if the verification passes. The present application optimizes the generation process of a large language model when processing wrong premises or fuzzy problems through the above method, thereby effectively reducing the occurrence of hallucination output.

The invention discloses a multi-modal large language model generation method and system based on visual information backtracking, and belongs to the technical field of artificial intelligence and multi-modal information processing. The implementation method comprises the following steps: 1, acquiring an image token sequence and a text token sequence; 2, obtaining the real attention of the model on the input image; screening the ROI set by using overlapping penalty; 3, generating statement sequence token prediction distribution and obtaining hierarchical normalized entropy; 4, when the normalized entropy of the hierarchy is greater than an entropy threshold and the screened ROI set injection is triggered for the first time, performing feedforward network residual injection; 5, the ROI set is injected into a feed-forward layer of the multi-modal large language model, visual information in the model is updated, and then attention-enhanced visual features are obtained; compared with the prior art, the method has the advantage that the phenomenon that the image is inconsistent with the fact in the image-based generation process due to strong priori brought by text corpora in the pre-training process of the model is solved.

The invention relates to a computer-implemented speech synthesis method comprising the steps: generating at least one simulated natural speech sequence, the generation comprising providing at least one physical speaker attribute as input to a generative artificial intelligence model; output from the generative artificial intelligence model forming said at least one simulated natural speech sequence, the generative artificial intelligence model having been previously trained on the basis of a training dataset comprising at least one natural speech sequence, each natural speech sequence being associated with at least one physical attribute of the corresponding speaker; extracting a set of speech features from each generated simulated natural speech sequence; and configuring a speech synthesis engine according to the extracted set of speech features. Figure 1

The application relates to the technical field of space automatic layout generation and generation process control, and discloses a multi-room consistency layout generation method and system based on generation process constraint modulation. The method comprises the following steps: acquiring space structure information of a target residential space; constructing a multi-room generation state model; constructing and updating a multi-room consistency constraint set in a layout gradual generation process; performing consistency constraint modulation processing on a candidate layout action in a current generation step, so that each room shares a consistency control condition in the generation stage; and outputting a multi-room overall layout scheme satisfying the multi-room consistency constraint set. Compared with the prior art in which each room is independently generated and then subjected to rule checking or scoring screening to realize consistency, the application cooperatively controls consistency relationships such as layout density, furniture scale and space blank proportion in the generation stage, so that the structural rationality, stability and generation efficiency of the multi-room overall layout result are improved.