40results about How to "Accurate reply" patented technology

The embodiment of the invention discloses a page access method, which comprises the following steps of: judging whether a target page to be accessed is a historical page, if the target page to be accessed is the historical page, acquiring a page resource and page data which correspond to the historical page and are stored during finishing of the access to the historical page, and accessing the historical page according to the acquired page resource and the acquired page data of the historical page. The embodiment of the invention also discloses a page access device. By the technical scheme of the embodiment of the invention, the historical page can be accessed under the condition that the page state during finishing of the access to the historical page is retained, so that the concurrency pressure of a server is reduced, the access operation time is shortened, and the overall page access efficiency is improved.

The invention relates to an intelligent question and answer method and device. The method includes: receiving user input information; acquiring the highest semantic similarity between the user input information and all the questions in a question and answer knowledge base, and obtaining a corresponding initial answer when the highest semantic similarity is greater than a preset similarity threshold; performing abstract semantic recommendation processing on the user input information according to an abstract semantic database when the initial answer includes a plurality of clauses, and the plurality of clauses at least correspond to more than two pieces of first pragmatic information, and then obtaining an abstract semantic expression corresponding to the user input information and the type of an abstract semantic, wherein the abstract semantic database includes a plurality of types of abstract semantics, and each type of abstract semantic has a corresponding second pragmatic information; and extracting a clause of the first pragmatic information corresponding to the second pragmatic information from the initial answer as a final result. The method can improve the accuracy rate of reply answers of the intelligent question and answer system, and improve the user experience.

The invention discloses an intelligent question-answering method and device based on federated learning, equipment and a storage medium, and the method is applied to terminal equipment, and comprisesthe steps: calling a local classification model based on the federated learning to determine the type of a to-be-consulted question when the to-be-consulted question is received; calling a local federated learning-based semantic comprehension model to retrieve each similar question corresponding to the question type to obtain a matching question corresponding to the to-be-consulted question; obtaining a reference answer corresponding to the matched question, and outputting the reference answer as a reply corresponding to the to-be-consulted question, wherein the classification model and the semantic comprehension model are respectively generated by executing federated learning algorithm training based on classification samples and semantic comprehension samples of the multi-party terminalequipment. According to the intelligent question and answer realized based on the federated learning classification model and the semantic comprehension model, the intelligent question and answer efficiency is improved through classification retrieval, and meanwhile, the question and answer accuracy and the security of data of each party are ensured.

Executing the steps of: receiving information of customer orders and information of prospect orders and storing into a received order database 20; dividing orders stored in said received order database 20 based on a reference master having various kinds of information about production materials registered therein, and storing the information of the orders which have been subjected to the division process to a received order division database 21; applying a process development to the orders which have been subjected to the division process, based on a basic unit master and storing into a process development database 22; specifying an optimum production starting date based on the information of orders which have been subjected to the process development and a production pattern stored in a production pattern database, performing loading, and storing results of the loading into a production planning database 23; and creating delivery date answer information, based on said optimum production starting date stored in the production planning database 23.

The embodiments of the present application provide an intent classification method and device based on a business corpus and an intelligent question answering method. An intent consists of a category,a user question and a reply source. In the intent classification, the method comprises the steps of setting an intent category according to the business requirement; setting a user question matchingthe business requirement according to the intent category; and setting a reply source of the intent, the reply source being associated with a corresponding knowledge base, knowledge category or knowledge item in the business corpus, so that the intention and the business corpus are logically independent of each other, and form a corresponding relationship in the form of the user question and the reply source in the intent. After the corresponding relationship is established, the process of training an intelligent question answering system by using the business corpus becomes a process of intention training for the intelligent question answering system, so that the intelligent question answering system has the ability to accurately identify the user's intent from the user's problem after training, the reply to the question is more accurate and the reply friendliness of the intelligent question answering system is improved.

The invention relates to the technical field of network application, and provides a chat expression reply method and device, and a storage medium, and the method comprises the steps: S10, receiving achat message sent by a user; wherein the chat message comprises any one of a text message, a picture message and a voice message; S20, analyzing the chat message to obtain effective information; S30,judging the effective information, and matching an expression reply strategy according to a judgment result to obtain an expression reply result; and S40, sending the expression reply result to the user. In addition, the invention also provides a chat expression reply device and a storage medium. According to the technical scheme provided by the invention, the chat message sent by the user is automatically recognized, the optimized expression reply is intelligently matched, and the expression reply is accurate and reasonable, has diversity and interestingness, and is rich in expressive force.

The present invention relates to Bi-Level optimization method for image deblurringso that the two levels of loss functions alternate model optimization, Bi- Level optimization mechanism is divided into two steps, the first step is to train a basic model with MSE loss condition, the second step is to use two-level loss iteration to fine-tune the model. This is because at the beginning of training,the divergence between the restoration effect and the clear image is large, and the effect of noise can be ignored, but at the end of training, the noise is amplified and the negative effect is more obvious, so the perceptual loss is introduced to suppress the noise and the MSE loss is changed to L1 loss to ensure the structural continuity sufficiently. The invention has the advantages of accuratetexture restoration and matching depth features with pixel values.

The invention discloses an emotional chat type reply generation method with theme perception. The method is characterized in that a generation model composed of an encoder, a theme perception module,an emotion perception module and a decoder is adopted; an external topic common sense introduced into a generation model is fused with two attention mechanisms, semantic and emotion information of a session is effectively captured by using hidden variables under a variational auto-encoder framework, and an emotion reply with topic perception is generated. Compared with the prior art, the method has the advantages that the generated session and the input session are better in topic correlation; according to the method and the device, the reply which better conforms to the preset emotion category can be generated, the reply with emotion can be more accurately generated, the generated reply and the input session can be ensured to be under the same session theme, two attribute requirements ofa conversation system are simultaneously met, and the method and the device have wider practical significance.

The invention discloses an electric sales AI robot based on artificial intelligence. The electric sales AI robot comprises a voice recognition module, a data storage module, an information processingmodule, a voice broadcast module and a reply result feedback module. The data storage module is used for storing the optimal semantic recognition model and providing reading and storage function permission to the outside; and the voice recognition module is used for recognizing user voice information and converting the recognized user voice information into user text information. According to therobot, a sample data set containing semantic keywords and intention labels is established according to the e-commerce service voice of the user intention processed by a real person, and an optimal semantic recognition model used for recognizing the user intention to generate a reply text is established on the sample data set; and the optimal recognition model is continuously updated according to the result feedback of the reply voice so as to improve the accuracy of user intention recognition, so that the questions proposed by the user can be replied more accurately, and the user experience inthe electricity selling process can be greatly improved.

The invention relates to the field of power monitoring systems, and more particularly to a power outage information monitoring system. The system is formed by collecting power outage information by acollection system, transmitting the same to a data processing system for processing, and transmitting data after processing to a monitoring system. The monitoring system consists of a power outage monitoring module, a system management module, a power outage information template maintenance module and a report form statistic module. Efficiency of staffs can be improved, automation of data statistics and analysis is realized, the enterprise manpower costs can be reduced, and thus teams and groups are enabled to have more energy for other business work.

An apparatus for detecting a phase of an input signal includes: a phase detector unit for detecting the phase of the input signal referring to a plurality of clock signals having different phases and for outputting a plurality of phase detection results, wherein each of the phase detection results represents a candidate phase of the input signal and corresponds to one of the clock signals; and a phase decision unit, coupled to the phase detector unit, for determining the phase of the input signal according to the phase detection results.

The invention relates to an automatic control system for silicothermic process semicontinuous magnesium smelting furnace burden calcination. The automatic control system can precisely calculate the deviation and control the temperature output value. The automatic control system comprises a computer (1), a D/A converter (2), a retaining appliance (3), an execution mechanism (4), an A/D converter (5), a sampling apparatus (6) and a transmission signal measuring mechanism (7). The automatic control system is characterized in that closed-loop control is adopted in the automatic control system. The automatic control system has the advantages of avoiding the conditions that a past open-loop control system is generally poor in control performance, the activity degree cannot be adjusted, and the calcination temperature and time cannot be tracked, and achieving action system stability.

The invention relates to a neutral point control device for a plunger pump, wherein the neutral point control device is installed on a tray and connected to the plunger pump on the tray, the plunger pump is equipped with a rotating shaft that is connected to an adjusting mechanism, the adjusting mechanism drives the rotating shaft to rotate to adjust the oil circuit movement direction of the plunger pump, the plunger pump is in an idle state when the shaft is rotated to the neutral position, an energy storage mechanism is included and is connected to the rotating shaft, the rotating shaft drives the energy storage mechanism to run and potential energy is stored in the energy storage mechanism when the adjustment mechanism adjusts the rotation of the rotating shaft, and the release of the potential energy in the energy storage mechanism drives the rotating shaft to return to the neutral position when the adjustment mechanism cancels the adjustment of the rotation of the rotating shaft.The invention has a novel structure, reasonable layout and stable operation, can realize the control of neutral point of the rotating shaft of the plunger pump, has the advantages of convenient installation, accurate accuracy and long service life, and meets current requirements for use.

The invention discloses a license entity extraction method and system based on deep learning, belongs to the field of natural language processing, and aims to solve the technical problem of how to improve the accuracy of license nouns and handling behaviors in an intelligent question and answer service and ensure more accurate reply to questions so as to improve the service quality and the satisfaction degree of a user. According to the technical scheme, the method specifically comprises the following steps: generating a training data set; training a BiLSTM-CRF model for extracting the license nouns and the affair handling behavior entities; and testing the training effect of the BiLSTM-CRF model. The system comprises a generation module, a training module and a test module. The training module is used for training a BiLSTM-CRF model extracted by a license noun and an affair handling behavior entity; the BiLSTM-CRF model is divided into three layers, the first layer is an Embedding layer, and each word in a sentence is mapped into a low-dimensional dense word vector from a one-hot vector by using a pre-trained or randomly initialized embedding matrix; the second layer is a bidirectional LSTM layer for automatically extracting sentence features; and the third layer is a CRF layer for performing sequence labeling of statements.