51 results about "Generative process" patented technology

In a corpus of scientific articles such as a digital library, documents are connected by citations and one document plays two different roles in the corpus: document itself and a citation of other documents. A Bernoulli Process Topic (BPT) model is provided which models the corpus at two levels: document level and citation level. In the BPT model, each document has two different representations in the latent topic space associated with its roles. Moreover, the multi-level hierarchical structure of the citation network is captured by a generative process involving a Bernoulli process. The distribution parameters of the BPT model are estimated by a variational approximation approach.

The process of text generation/creation is automated. The text to be processed is used as seed for the text generation process. The text to be processed can be in any language and can be passed to text generation process through any internal/external application or process, through speech technology or through manual entry. At the first step, word(s) are extracted from the text. Each word is considered as seed and this seed is grown up into different word(s)/sentence(s) lists according to the selected criteria. The generated lists are then processed and combined/jointed through a simple mechanism to generate text. This generated text then can be saved, analyzed, filtered or searched on the internet, intranet, extranet, in database(s) or in user defined data repositories again according to the criteria selected by the user or by some external application or process.

The invention provides an automatic text abstraction method based on a pre-training language model. According to the method, an ultra-large-scale unsupervised Chinese corpus is used for training a complex deep language model; the low-layer network structure of the model can extract and retain the grammar and structure information of a text, and the high-layer network structure can extract and retain the semantics and context information of the text, so that the richer text features and semantic information are provided for an automatic text summary task; a pre-training language model and an Encoder are combined to realize, the text features and the semantic information in the pre-training language model are fully utilized, so that a better semantic compression effect is provided, and the performance of an automatic text abstract is improved; a pre-training language model and a decoder are combined, not only the semantics in an original text are considered in the text generation process, but also the semantic information of vocabularies is also considered, so that the readability of the generated text and relevance with the original text are improved, and the performance of an automatic text abstract is improved.

A method tracks and analyzes dynamically a generative process that generates multivariate time series data. In one application, the method is used to detect boundaries in broadcast programs, for example, a sports broadcast and a news broadcast. In another application, significant events are detected in a signal obtained by a surveillance device, such as a video camera or microphone.

The invention provides an image subtitle generating method based on a novel attention model. The image subtitle generating method mainly includes the steps of establishing a significance forecasting model; selecting a significant image region; forming significance and text perception attention, generating subtitles; analyzing the attention state, wherein different parts of input images are focused in the subtitle generating process through a novel subtitle structure, particularly, significant parts and parts needing to be combined with the context in the images are given through the significance forecasting model; images are extracted through a convolutional neural network, and corresponding subtitles are generated through the recurrent neural network; through attention model extension, two attention paths are created in the significance forecasting model, wherein one attention path importantly concerns the significant region, and another attention path importantly concerns the context region; the two paths jointly cooperate in the subtitle generating process, excellent subtitles are generated step by step, and a further contribution is provided for the innovative solution of image subtitle generating.

The present invention includes according to picture text stored image information establishing picture and information corresponded relation library; according to picture text display text display image information inputting query criterion, utilizing said query criterion to proceed matching operation in relation library, if matching successfully then feeding back matched information corresponded picture border information and picture text display zone information; according to searching picture border information to determine displaying area only including searching picture, generating displaying regional cover layer capable of covering except searching image display area external picture text; outputting cover layer to displaying terminal, only displaying searching picture. The present invention can high efficiency displaying assigned area in a picture, saving new picture generative process and data transmission process, and processing resource.

A system and method of monitoring and diagnosing on-line multivariate process variable data in a process plant, where the multivariate process data comprises a plurality of process variables each having a plurality of observations, includes collecting on-line process data from a process control system within the process plant when the process is on-line, where the collected on-line process data comprises a plurality of observations of a plurality of process variables, performing a multivariate statistical analysis to represent the operation of the process based on a set of collected on-line process data comprising a measure of the operation of the process when the process is on-line, where the representation of the operation of the process is adapted to be executed to generate a result, storing the representation of the operation of the process and the set of collected on-line process data, and generating an output based on a parameter of the representation of the operation of the process, where the parameter of the representation of the operation of the process comprises one or more of a result generated by the representation of the operation of the process, a process variable used to generate the representation of the operation of the process and the set of collected on-line process data.

The invention discloses an individual abstract system based on user interest modules. The individual abstract system consists of a Web information retrieval unit, a user interest unit and an individual abstract unit. In the individual abstract system, the user interest modules described with hierarchical conceptual structure are established and/or updated by using the conceptual clustering method through analyzing retrieval logs of users; subsequently, the similarity of user interests and sentences in the retrieval results is analyzed according to the user interest modules and the retrieval results, thus getting individual abstracts meeting requirements of users. The individual abstracts got through grading and handling individual sentences takes full consideration of the interests and the features of users so that the generative process of the abstracts are matched according to the interests of users and improving the effectiveness of the abstracts and the satisfaction of users.

The present invention provides a method for integrating crowdsourced annotations. According to the method provided by the present invention, a generalized anti-Gaussian distribution is defined by using a regularization hyperparameter, a spacing distance hyperparameter, an annotator voting weight, and a difference between the number of times that annotators annotate a current prediction item as a corresponding estimate and the number of times that annotators annotate the current prediction item as a subcategory; sampling is performed to obtain an auxiliary parameter; and annotator weights are updated by using the auxiliary parameter, thereby significantly enhancing a discrimination capability of a model. Then, a traditional annotation integration majority voting model and a confusion matrix model are integrated, thereby achieving an objective of more comprehensively describing a data generation process. In addition, an accurate prediction item update value is obtained through sampling, and moreover, running efficiency is also improved.

In accordance with an embodiment, data may be automatically extracted from semi-structured web sites. Unsupervised learning may be used to analyze web sites and discover their structure. One method utilizes a set of heterogeneous “experts,” each expert being capable of identifying certain types of generic structure. Each expert represents its discoveries as “hints.” Based on these hints, the system may cluster the pages and text segments and identify semi-structured data that can be extracted. To identify a good clustering, a probabilistic model of the hint-generation process may be used.

The present invention provides a method, system, and program product for modeling processes such as business processes. Specifically, under the present invention, a process having multiple variations is modeled as a placeholder process, and is associated with an annotation that maps the multiple variations to different contexts of the placeholder process. Thereafter, process variations can be generated based on the different contexts of the placeholder process. Integrity checks can also be performed to determine if any of the process variations has an error so that the generation of any corresponding process will be prevented.

The invention relates to a text abstract automatic generation method and system fused with a pre-training model, and belongs to the technical field of text abstract automatic generation. According to the system, on the basis of a traditional Sequence-to-Sequence model and a Transformer framework, a BERT is used as an encoder to obtain context semantic features of an original text, meanwhile, a convolutional neural network is added to serve as gating, and keyword and key phrase screening is conducted on feature vectors. On the basis of using Transformer Decoder in a decoder stage, a pointer mechanism is added, so that the abstract generation model can better solve the problem of insufficient vocabulary (OOV) in the generation process, the readability of the generated abstract is improved, and the best abstract legal text is generated by adopting a beam search method in a final generation stage.

Embodiments of the present invention allow a system to use data gathered from social networks and other systems to determine an ordered list of desired topics or skills to define a subject matter area and incorporate the order of the topics or skills into a search. To define this subject matter area, embodiments can consider not just the topics or skills that are listed, but those topics or skills that are similar based on a pre-computed topic graph. These considerations can be incorporated into a generated query, so that the query itself accounts for similarity of topics via the topic graph and the order of desired terms. The query generation process can include a claimed skills veracity model that provides differential weighting to claimed skills, based on the skill-sets of users who are deemed to be similar to the user being evaluated.

Provided is a method and system of feature engineering for execution machine learning. The method comprises the steps that (A), a data table taken as a data source is obtained, one line of the data table corresponds to a data record, wherein one row of the data table corresponds to a field; (B), a configuration item used for generating a feature is obtained, wherein the configuration item is usedfor respectively indicating a name, extraction method and source field of each feature; (C), a part of the data in the data table is selected to be recorded; (D), a feature generative process of a feature value of a to-be-displayed feature obtained on the basis of a field value in the part of the data record according to the configuration item is displayed to a user. The method and system of the feature engineering for the execution machine learning have the advantages that the user can conveniently and visually know the feature generative process, and the problems occur in the feature generative process can be known by the user.

The invention discloses a lake-reservoir algal bloom generating mechanism time varying model optimization and prediction method based on a taboo searching algorithm and a genetic algorithm. The method comprises the steps that first, a water bloom generating mechanism time varying model is established; second, an influence factor function model base is established; third, based on the genetic algorithm, water bloom generating mechanism time varying model parameters are optimized; fourth, based on the taboo searching algorithm, a water bloom generating mechanism time varying model structure is optimized, and influence factors are analyzed; and fifth, optimum water bloom generating mechanism time varying model prediction is carried out. According to the method, a time variable is introduced into the water bloom generating mechanism model, the water bloom generating mechanism time varying model is established, so that the method is suitable for simulating a water bloom generating process and can be used for water bloom prediction, and the problem that water bloom prediction based on a data driving model is not accurate enough, and a mechanism driving model cannot predict water bloom is solved.

The invention belongs to the field of computer-generated holography and particularly relates to a rapid computer-generated holography algorithm based on a trigonometric function look-up table. The rapid computer-generated holography algorithm is characterized in that based on computer-generated holography through a point source method, the trigonometric function look-up table deduces an original formula and simplifies the phase position part related to redundancy computation through mathematical approach transformation and trigonometric function identical transformation so as to generate two phase position look-up tables containing depth information. The generative process adopts parallel computing, is high in speed and accuracy, and contains all phase position information of objects (including three-dimensional and two-dimensional objects); in addition, the internal memory usage is less, the repeated utilization factor is high, and the application range is wide. Moreover, the parallel optimization is performed on the algorithm, parallel computing is adopted during the table look-up addressing and computing processes, and the computer-generated holography speed is effectively accelerated.

The invention provides a theme-enhanced chat robot reply generation method and device, and the method comprises the steps: firstly, carrying out the preprocessing and sampling of a dialogue text, andobtaining a dialogue training set; secondly, performing topic extraction on the dialogue text by utilizing the CNN to obtain a text topic; utilizing the text theme, and predicting a reply theme through the MLP to obtain a predicted reply theme; and finally, inputting the dialogue training set into the Seq2Seq generation model, and introducing the text theme and the prediction reply theme into thegeneration model through Gate to assist the generation process so as to obtain the chat robot dialogue reply. The method can achieve the technical effects of reducing the generation probability of high-frequency general replies and enabling the nerve dialogue model to generate more coherent and abundant replies.

The invention discloses a method and device for generating an image by a text with introduced class information, the method for generating the image by the text with introduced class information comprises a training stage and a testing stage, the training stage is based on a generative adversarial network, and a natural language text describing an image, the class label of the text, a corresponding real image training generator and a discriminator are utilized; and in the test stage, a corresponding image is generated in a generator by using the text and the class label thereof. The method and device have the advantages that text semantic image features and class information image features are generated through respective transcoding according to text information codes and class information codes, then the two levels of image features are fused to decode and generate the image, corresponding class information is introduced in the image generation process to enhance the correlation between the generated image and the text, and meanwhile, through the multi-stage generation process in the training process, the higher-resolution image is gradually generated, and the training difficulty of directly generating the high-resolution image is reduced.

The invention discloses a semantic parsing-based structured building specification automatic generation device, which comprises a coding module of building specification domain features, a generationmodule with SNL grammatical constraints and a two-stage generation module, and is characterized in that the coding module of the building specification domain features obtains term semantic vectors according to natural language description; the generation module with the SNL grammatical constraint realizes a generation function with the grammatical constraint; the generation module with the SNL grammatical constraint converts the generation process of the SNL into a selection task of a sentence pattern template and a filling task of sentence pattern details by utilizing the grammatical construction rule of the SNL and the detachable characteristic of the SNL, so that the limitation on the generation process of the decoder is realized; the two-stage generation module provides a function ofconverting a natural language rule into SNL; the two-stage generation module fuses an encoding method and a generation method with grammatical constraints to obtain a two-stage generation model so asto reduce errors caused by overfitting of a single model.

Generative design uses artificial intelligence to automatically create optimal designs from a set of system design requirements. A generative design system receives user-supplied design requirements and generates an optimized geometry by chaining iterative applications of computer aided generative design. In one aspect, the system uses the output of a first generative design process as a requirement of or as part of the design requirements for a second generative process, with or without intervening user input. In another aspect, the system enables designers to incrementally or iteratively specify and edit design requirements such that each resulting geometry is derived from a previously or most recently generated optimized geometry and optionally updated design requirements. The system, by using or reusing outputs and/or data from previous design iterations can produce subsequent designs more efficiently and quickly.

Computer-implemented methods, computer program products and data processing systems for presenting dynamic data pertaining to a process. In one aspect, region information identifying a plurality of user-selected regions of the process, switching information identifying criteria for switching among the user-selected regions, and dynamic data pertaining to the process, are used to generate a dynamic visual representation of the process which emphasizes, at any given time, only one of the user-selected regions. The user-selected region emphasized at any given time is automatically determined according to the criteria for switching. In another aspect, region information identifying at least one user-selected region of the process and dynamic data pertaining to the process are used to generate a dynamic visual representation of at least a portion of the process containing the at least one user-selected region. The visual representation emphasizes each user-selected region within the context of the portion of the process.