73 results about "Natural language sentence" patented technology

The invention belongs to the technical field of information, and relates to a semantic guidance pedestrian re-identification method and system based on text prompt. The method comprises the following steps: inputting a training image into a visual encoder to obtain visual embedding; mapping the visual embedding into a text space by using a reverse network to obtain a pseudo token, and integrating the pseudo token into a natural language sentence to obtain a language prompt for an input image; inputting the language prompt into a text encoder to obtain text embedding; training a multi-modal interaction module by utilizing visual embedding and text embedding; and inputting a query picture into the trained multi-modal interaction module to obtain a feature vector fusing vision and text information, and executing similarity retrieval in a pedestrian image database by using the feature vector fusing vision and text information to obtain a pedestrian re-identification result. According to the method, the retrieval performance on an existing pedestrian re-identification data set is remarkably improved.

Various disclosed embodiments are directed to deriving, via a language model, a summary of data by converting or encoding table data into one or more natural language sentences, which are then used as input to the language model for generating the summary. One or more embodiments are additionally or alternatively directed to deriving, via a language model, a response to a user question or command via a chat interface by providing the language model with the generated summary as input. In this way, for example, the language model can use the summary as a prompt or other target context for providing a response.

Automatic code generation for robotic process automation (RPA) that uses a cognitive artificial intelligence (AI) layer to automatically generate computer program code, text, or other items based on an input source is disclosed. Expressions and / or code snippets may be generated for RPA workflows that provide code in a programming language or convert code from one language to another. The input source may include natural language sentence(s), source document(s), pseudocode written by a user for a desired task, etc. These inputs may be used to modify an RPA workflow, generate a new RPA workflow, create a document describing a task, generate an application, etc.

Systems or techniques that facilitate touchless operation of medical devices via large language models are provided. In various embodiments, a system can access, via a microphone associated with a medical device, a first natural language sentence spoken by a user of the medical device, wherein the first natural language sentence requests that the medical device perform an equipment operation. In various aspects, the system can: extract, from an encoder portion of a large language model, an embedding corresponding to the first natural language sentence; identify the equipment operation, by comparing the embedding to a plurality of embeddings respectively corresponding to a plurality of available equipment operations of the medical device, wherein the equipment operation is identified as whichever of the plurality of available equipment operations whose embedding is most similar to the embedding of the first natural language sentence; and instruct the medical device to perform the equipment operation.

Systems and methods for generating an enhanced error message are provided. An example method includes: receiving one or more raw error messages. The one or more raw error messages include one or more stack traces. The method further includes matching at least one raw error message of the one or more raw error messages to one or more error rules from a plurality of error rules. The one or more error rules include regular expression patterns. The method further includes parsing the at least one raw error message, based on the one or more matched error rules from the plurality of error rules; and generating one or more enhanced error messages, based on the at least one parsed raw error messages. The one or more enhanced error messages include one or more natural language sentences. The method further includes embedding the one or more enhanced error messages into a website.

Disclosed herein is computer technology that provides adaptive mechanisms for learning concepts that are expressed by natural language sentences, and then applies this learning to appropriately classify new natural language sentences with the relevant concept that they express.

Method includes populating fake value for each of entities, to generate string of fake entity values that correspond to entities; inserting sentinel token between adjacent fake values included in the string to generate first input data; generating, by natural language generation model, natural language sentences based on first input data, natural language sentences including one or more fake values from the string; analyzing natural language sentences to determine whether any fake value from the string is missing; based on the fake value missing, summarizing, using text summarization model, natural language sentences to generate text summary; concatenating the text summary with the fake value, to generate second input data; and generating, by a next sentence generation model, additional natural language sentence using the second input data, the additional natural language sentence including the fake value. Additional natural language sentence is combined with natural language sentences to generate final natural language sentences.

Generating a semantic synopsis is provided. A dataset comprising structured data is received and semantic analysis of the dataset is performed to determine categories of information included in the datasets. A spatio-temporal analysis of the dataset is also performed to determine time periods and locations to which the dataset applies. A static synopsis of the dataset is then generated in natural language sentences that describes semantic information in the dataset and time periods and locations covered by the dataset. The synopsis can be customized according to a user role and search query data.

Generating a semantic synopsis is provided. A dataset comprising structured data is received and semantic analysis of the dataset is performed to determine categories of information included in the datasets. A spatio-temporal analysis of the dataset is also performed to determine time periods and locations to which the dataset applies. A static synopsis of the dataset is then generated in natural language sentences that describes semantic information in the dataset and time periods and locations covered by the dataset. The synopsis can be customized according to a user role and search query data.

The application discloses a clinical medical examination knowledge base construction and application method, device, equipment and medium, relates to the computer technical field, and includes: converting the collected clinical medical data into corresponding dense vectors based on a preset large language model to obtain a target clinical medical examination knowledge base; performing a configuration operation of a dialogue retrieval mode on the target clinical medical examination knowledge base, and after the configuration is completed, performing vector conversion on a received natural language sentence to be replied based on the preset large language model to obtain a corresponding question vector; by calculating the cosine similarity between the question vector and each dense vector in the target clinical medical examination knowledge base, the most relevant target vector to the question vector is determined; and based on the preset large language model and the target vector, a corresponding target natural language reply sentence is generated to reply to the natural language sentence to be replied. In this way, the retrieval efficiency can be effectively improved, and the convenience of retrieval is further improved.

The embodiment of the invention discloses a statement conversion method and device, electronic equipment and a storage medium, and belongs to the technical field of artificial intelligence. The method comprises the steps of extracting corresponding global constraint information according to semantics of a to-be-converted statement, and determining a first sub-statement corresponding to the global constraint information from the to-be-converted statement; determining global action range information corresponding to the first sub-statement according to the global constraint information, and determining a relationship conjunction between the global constraint information and the first sub-statement according to the first sub-statement; determining nested structure information according to the global action range information, the relational conjunction, the first sub-statement and the global constraint information; when the first sub-statement is not the atomic statement, performing statement decomposition on the first sub-statement to obtain updated nested structure information, and determining an intermediate language statement; and generating a conversion statement corresponding to the to-be-converted statement according to the intermediate language statement. According to the method, the quality of the formalized language statement obtained by converting the natural language statement is improved.

Embodiments of this application disclose a method and an apparatus for generating a natural language sentence describing a workflow, and a storage medium. The method includes: determining a subject placeholder based on a first selection operation performed on a language element library; determining, from a subject thesaurus associated with the subject placeholder a subject occupying the subject placeholder, where the subject corresponds to a resource; determining a predicate placeholder based on a second selection operation performed on the language element library; determining, from a predicate thesaurus associated with the subject, a predicate occupying the predicate placeholder, where the predicate corresponds to a service operation of the resource; and generating the natural language sentence based on the occupied subject placeholder and the occupied predicate placeholder, where the natural language sentence is adapted to be converted into a low-code representation describing a workflow, and the workflow represents that the resource performs the service operation. The workflow may be conveniently described by using the natural language sentence, which reduces implementation difficulty of the workflow, and enriches implementations of the workflow.

A system and method for query selection are provided. The method may include acquiring a natural language sentence, pre-processing to obtain a standard node sequence, constructing a node tree based on the relationship between an index node and other nodes, generating a data query command based on the node tree, querying data using the data query command, and filtering the results. The standard node sequence may include at least an index node and a condition node. The node tree may be used to characterize the index-condition combination. The system may include an acquisition unit, a pre-processing unit, a node tree construction unit, a translation unit, and a querying and filtering unit.

A processor of an electronic device according to an embodiment may receive a prompt sentence from an external electronic device via a communication circuit. The processor may execute a first model for natural language processing to generate a first natural language sentence related to the prompt sentence. The processor may acquire a second natural language sentence by changing a portion of the first natural language sentence. The processor may execute a second model linked to evidence information to determine scores indicating accuracies of the first natural language sentence and the second natural language sentence. The evidence information may be used as a criterion for the second model to evaluate the accuracy of a natural language sentence. The processor may determine the first natural language sentence as a response message to be transmitted to the external electronic device by using the scores.

The invention discloses a text data processing method and system based on mathematical logic and propositional logic. The method comprises the steps of firstly obtaining text data and endowing the text data with weights to generate bottom layer data; thirdly, converting the natural language statement into an operable rule system entry by utilizing a pre-constructed mathematical logic and propositional logic framework, and generating a basic value; a logic item set, a positive control item and a negative control item which are input by a user are introduced into bottom layer data, and weight balancing processing is carried out. And finally, according to the matching degree of the items after weight balancing and the basic values of the items, scores are calculated, and the item with the highest score is selected as response data. According to the method, the response data are ensured to completely meet user requirements and system logic in the error-tolerant field, and the method has high accuracy and reliability.

The invention discloses an electric power standard obtaining method and system based on a knowledge graph, and belongs to the technical field of electric power, and the method comprises the steps: S1, analyzing a natural language statement through a large language model to obtain a user structured intention, and obtaining a knowledge graph node corresponding to the user structured intention in the knowledge graph through knowledge graph embedding; s2, performing multi-modal traversal on the multi-modal index database by taking knowledge graph nodes as traversal anchor points and user characteristics as traversal constraints to obtain a candidate result set; and S3, obtaining weights of candidate result pairs in the candidate result set through a graph attention mechanism, and sorting the candidate result pairs based on the weights of the candidate result pairs so as to obtain an electric power standard. The technical problem that the suitability of the electric power standard is difficult to improve in the prior art is solved.

Provided is a method for identifying subtly biased texts within open corpora, which includes: a natural language sentence collecting step of collecting natural language sentences from text-based datasets including open conversation datasets and web corpus datasets; a bias candidate dataset obtaining step of comparing semantic vectors between words composing the natural language sentences by using a word-embedding model from the collected natural language sentences, thereby obtaining a bias candidate dataset including seed bias data expected to be socially and ethically biased and derived bias data expanded from the seed bias data; and a bias determining step of determining the bias candidate dataset as one of a surficial bias dataset, a subtle bias dataset, and a non-bias dataset by using multiple large-scale language models.

The innovative method (5000) serves for determining an anomalous condition on a surface (10) of / associated to an industrial machine (100); the method comprises the steps of: b) providing (5100) a set of sentence embedding features of a natural language sentence expressing a concept relating to a presence of a predetermined anomaly on a surface, c) receiving (5200) an image of the surface (10) and / or of surroundings of the surface, d) encoding (5300) the image thereby generating a set of image embedding features, e) determining (5400) a similarity value of a similarity score, wherein the similarity score corresponds to a level of similarity between the set of image embedding features and the set of sentence embedding features, f) comparing (5500) the determined similarity value with a predetermined similarity threshold, and g) transmitting (5600) a result, the result being positive if the determined similarity value is higher than the predetermined similarity threshold.

Provided are systems and methods and computer-implemented systems for generating synthetic data and optimizing performance of large language models, by integrating semi-automatic synthetic data generation, prompt optimization and continual learning into a unified system, without full fine-tuning of the model, to improve performance of the large language model. The cold start problem of the large language model is solved through a multi-stage process, including: generating a diversified parameter set and natural language sentences; constructing synthetic user queries using feature descriptions, parameter sentences and constraint instructions; bootstrapping with a small amount of samples with the help of a compiler, and iteratively evaluating instruction-example combinations to optimize the large language model module; and combining real-time data and synthetic data to complete continual validation and optimization.

According to some embodiments, systems and methods are provided including receiving a crash dump file; extracting a call stack from the received crash dump file, wherein the call stack includes one or more functions, the functions having ordered positions in the call stack; converting the extracted call stack to natural language sentences; converting the natural languages sentences to a first call stack matrix; receiving the first call stack matrix and a second call stack matrix at a crash model, wherein the crash model is a Siamese neural network model; determining a similarity score for the first call stack matrix and the second call stack matrix; and determining whether the first call stack matrix and the second call stack matrix represent duplicate crashes based on the similarity score. Numerous other aspects are provided.

The invention discloses a 3D scene semantic segmentation system and method for open vocabularies, and belongs to the technical field of vision and image processing, and the system comprises a point cloud input module which is used for inputting preprocessed point cloud data and forming a point cloud set; the scale point cloud feature extraction module is used for extracting geometric local features of the point cloud data in the point cloud set, then extracting context features of the geometric local features, and finally summarizing to obtain full-scale fusion features; the CLIP language embedding module is used for mapping a natural language statement into a semantic embedding vector through a CLIP text encoder and carrying out normalization processing on the semantic embedding vector; and a vision-language semantic alignment module. According to the method, the natural language tag input by the user can be matched with the 3D point cloud data, support for any semantic description is realized, and a model does not need to be trained again, so that flexible and semantic-rich three-dimensional space understanding is realized.

Disclosed in the embodiments of the present application are a conversion method and apparatus for a natural language sentence, and a storage medium. The method comprises: determining a first mapping relationship between a subject in a natural language sentence and a function block header of a function block, wherein the subject corresponds to a resource in a work cell; determining a second mapping relationship between a predicate in the natural language sentence and a function block name of the function block, wherein the predicate corresponds to a service operation of the resource; and on the basis of the first mapping relationship and the second mapping relationship, converting the natural language sentence into the function block, or converting the function block into the natural language sentence, wherein the function block is presented as a function block type diagram, the function block is adapted to describe a workflow, and the workflow represents the resource executing the service operation. A natural language sentence can be converted into a function block that is presented as a function block type diagram, thereby improving the efficiency of workflow execution.

To provide an evaluation apparatus which narrows down causes for unsatisfactory performance of a language processing model.SOLUTION: An evaluation apparatus has: an acquisition unit for acquiring embedding generated by using an embedding layer included in a language processing model for a natural language sentence included in each of a plurality of training data pieces; a clustering unit for clustering the plurality of embedding; a calculation unit for calculating an evaluation index obtained by evaluating the embedding result, referring to the label included in each of the plurality of training data pieces; and an evaluation unit for evaluating a quality of the embedding layer. Thus, it is possible to optimize the performance of a machine-leaned language processing model.SELECTED DRAWING: Figure 1

Provided are a method and device for providing diagnostic information about spinal diseases using a natural language processing model related to spinal diseases. The method according to an embodiment includes receiving symptom information of a patient with spinal disease from the user terminal, wherein the symptom information of the patient with spinal disease includes a natural language sentence input by the user terminal; determining standardized information about the symptoms of the patient with spinal disease through a natural language processing model related to spinal diseases, based on the symptom information of the patient with spinal disease, wherein the spinal disease-related natural language processing model is a model trained to standardize symptom information of a patient with spinal disease based on chief complaints of a plurality of patients with spinal disease, wherein the standardized information about the symptoms of the patient with spinal disease includes standardized information on a body part where the symptom occurs, standardized information on a symptom type, and standardized information on symptom severity; determining diagnostic information about the patient with spinal disease based on the standardized information about the symptoms of the patient with spinal disease; and transmitting the diagnostic information about the patient with spinal disease to the user terminal.

A sentence generation device capable of generating a sentence in which information elements are included in a desired distribution is implemented. The sentence generation device that generates a sentence composed of a natural language sentence includes an information element distribution requirement acquisition unit that acquires an information element distribution requirement regarding a distribution of the information elements to be included in the sentence, and a sentence generation unit that generates the sentence so as to satisfy the information element distribution requirement using a language model. The generated sentences can also be utilized as training data for AI systems to facilitate automated decision making in diverse machine-learning scenarios.

Various embodiments of this application include methods and apparatus for generating a natural language sentence describing a workflow. An example method includes: determining a subject placeholder based on a first selection operation performed on a language element library; determining, a subject occupying the subject placeholder, where the subject corresponds to a resource; determining a predicate placeholder based on a second selection operation; determining, from a predicate thesaurus associated with the subject, a predicate occupying the predicate placeholder, where the predicate corresponds to a service operation of the resource; and generating the natural language sentence based on the occupied subject placeholder and the occupied predicate placeholder, where the natural language sentence is adapted to be converted into a low-code representation describing a workflow, and the workflow represents that the resource performs the service operation.

The application provides a training method and device of an open information extraction model, comprising: obtaining a target data set with natural language sentences as samples; generating paraphrases of each natural language sentence in the target data set; performing structured knowledge recovery on the paraphrases of each natural language sentence in the target data set to obtain structured knowledge corresponding to each natural language sentence in the target data set; constructing a first data set with the paraphrases and the structured knowledge corresponding to all natural language sentences in the target data set; and training the open information extraction model by using the first data set and the target data set in a noise reduction training mode. The application constructs a syntax-robust training framework based on paraphrase generation and structured knowledge recovery, so that the open information extraction model can be trained on a data set with sufficient and accurate syntax distribution to adapt to real-world scenarios.

Method includes populating fake value for each of entities, to generate string of fake entity values that correspond to entities; inserting sentinel token between adjacent fake values included in the string to generate first input data; generating, by natural language generation model, natural language sentences based on first input data, natural language sentences including one or more fake values from the string; analyzing natural language sentences to determine whether any fake value from the string is missing; based on the fake value missing, summarizing, using text summarization model, natural language sentences to generate text summary; concatenating the text summary with the fake value, to generate second input data; and generating, by a next sentence generation model, additional natural language sentence using the second input data, the additional natural language sentence including the fake value. Additional natural language sentence is combined with natural language sentences to generate final natural language sentences.

To disclose a technique for improving user's convenience.SOLUTION: The computer program causes the computer to function as a first reception unit that receives an input of a first natural sentence, a first supply unit that supplies first natural sentence data representing the first natural sentence to a natural sentence processing unit when the input of the first natural sentence is received, and a first acquisition unit that acquires first object information based on the first natural sentence data from the natural sentence processing unit. The first object information is caused to function as: a first acquisition unit which includes first identification information for identifying the first object for each of M1 screens among the M screens; and a second display control unit which, when the first object information is acquired from a natural sentence processing unit, causes a display device to sequentially display M screens including the M1 screens according to the first object information, wherein the first object is emphatically displayed on each of the M1 screens.SELECTED DRAWING: Figure 6

The computer-implemented method (2000) serves for providing explanatory information regarding anomalies in an industrial machine or plant in the form of natural language sentences; the method comprises the steps of: a) receiving (2100) a plurality of time series sequences deriving from a corresponding plurality of sensors of said industrial machine or plant, b) encoding (2200) said plurality of time series sequences thereby generating a corresponding plurality of time series embedding features, c) mapping (2300) said plurality of time series embedding features thereby generating a plurality of sentence embedding features, d) decoding (2400) said plurality of sentence embedding features thereby generating a natural language sentence, and e) transmitting (2500) said natural language sentence.