58results about "Semantic analysis" patented technology

The invention discloses a telecommunication fraud prevention system and method based on big data and machine learning. The system comprises a mobile terminal, a big data analysis terminal and a fraud interdiction governance terminal, wherein the mobile terminal is used for performing fraud detection determination on current telecommunication data according to predetermined constraint rules when receiving a short message or an incoming call message, a machine learning algorithm is adopted to detect whether the telecommunication data is a telecommunication fraud, and if the detection result is that the telecommunication data is determined as a telecommunication fraud, fraud data information is uploaded to the big data analysis terminal; the big data analysis terminal is used for performing real-time statistics on the fraud data information uploaded and reported from the mobile terminal and sending fraud early-warning information to the fraud interdiction governance terminal according to a bank card account or/and a phone number with the number of received reports exceeding a certain threshold value; and the fraud interdiction governance terminal is used for taking corresponding measures in time to interdict occurrence of a telecommunication fraud event when receiving the fraud early-warning information. The system can unite the mobile terminal, an operator, a public security institution, a bank and other institutions, quick and effective prevention can be realized, and the telecommunication fraud can be cracked down in time.

An object of the invention is an automated analysis system for analyzing at least one of scientific, technological and business information. The automated analysis system comprises means 100 for processing a data amount to accomplish a collection data form, means for structuring the collection data form describing the content of a source document, means for automatically identifying input documents in data warehouses comprising similar structured data forms as said structured collection data form, means for preprocessing and tokenizing said input documents, and the automated analysis system comprises a classification engine using the tokenized and preprocessed documents as input classifying each item of the input documents, and said means for processing, means for structuring, means for automatically identifying, means for preprocessing and tokenizing and said classification engine being configured to perform their tasks in one or more modelling round(s).

Techniques are presented for providing misinformation detection in online content. The described techniques can identify instances of misinformation in online content and pass a misinformation result to the user. A misinformation probability analysis can be performed by applying a syntactic analysis and a semantic analysis to detect misinformation with confidence by applying featurization to a URL, text of content referenced by the URL, and metadata associated with the URL using a feature set, the feature set comprising semantic-based features and syntactic-based features, wherein the semantic features and the syntactic features are selected from the group consisting of: sentiment amplifiers, sentiment continuity disruption features, lexical features, keywords, baseline features, speech act, sensicon features, emotion detection on the obtained text, exaggerated language, strong adjectives, heuristics, bag-of-words, objectivity, colloquial-ness score, and semantic difference.

The invention provides a method for predicting social attention hotspots in network public opinion events based on group analysis, which can predict the network public opinion events which may be developed into the social attention hotspots, provide decision support for processing the public opinion events, collect related data of the public opinion events, construct a public opinion event set, obtainpublic opinion event relation subgraphs and social attention indexes in different time windows, carry out graph representation learning on data in the public opinion event relation sub-graphs in different time windows through a graph embedding technology to obtain feature vectors of corresponding event relations, and enable the feature vectors of the event relations and corresponding social attention indexes to form a two-tuple, construct a time sequence prediction model based on a bidirectional long-short-term memory neural network, input the two-tuples into the model for training iteration until the model converges, input new public opinion events into the model to obtain social attention indexes of the new public opinion events in future time, and select the public opinion event with the maximum social attention index as a predicted social attention hotspot.

The invention belongs to the technical field of natural language processing, and relates to a biomedical relationship extraction method based on a pre-training model and a self-attention mechanism. According to the method, more complex information in the biomedical sentences can be extracted by using the ELMO pre-training model, so that the expression effect of sentence features of the biomedicaltext is improved, and the relationship between biomedical entities is better extracted; position features are added to learn the position relationship between the internal structure of the sentence and the biomedical entity, and the internal correlation between the data and the features in the sentence is better captured by using a self-attention mechanism, so that the task of biomedical relationship extraction is better completed. According to the method, the problem that only simple semantics of sentence sequences are focused in current biomedical relation extraction is solved, the relationextraction effect of regular biomedical texts can be enhanced, and the method has a good effect on relation extraction of non-regular biomedical texts on social media.

The invention discloses a medical institution dialysis level assessment method and device based on semantic recognition, equipment and a storage medium. The method comprises: a dialysis effect feedback instruction triggered by a dialysis patient is received, dialysis data, information of a medical institution and a dialysis sufficiency evaluation report are obtained from a dialysis data managementserver according to the dialysis effect feedback instruction, and the dialysis data and the dialysis sufficiency evaluation report are displayed on a user interface; receiving dialysis effect evaluation information made by the dialysis patient; extracting at least one keyword from the dialysis effect evaluation information; determining dialysis data contents corresponding to the keywords in the dialysis data; and sending each keyword, the dialysis data content corresponding to each keyword, the information of the medical institution and the dialysis sufficiency assessment report to a medicalinstitution assessment server. In this way, the patient can assist relevant departments in examining the dialysis level of the medical institution, and it is guaranteed that the examination result cantruly reflect the dialysis level of the medical institution.

Techniques and systems are described to model and extract knowledge from images. A digital medium environment is configured to learn and use a model to compute a descriptive summarization of an input image automatically and without user intervention. Training data is obtained to train a model using machine learning in order to generate a structured image representation that serves as the descriptive summarization of an input image. The images and associated text are processed to extract structured semantic knowledge from the text, which is then associated with the images. The structured semantic knowledge is processed along with corresponding images to train a model using machine learning such that the model describes a relationship between text features within the structured semantic knowledge. Once the model is learned, the model is usable to process input images to generate a structured image representation of the image.

The invention discloses a data matching method and device and electronic equipment. The data matching method comprises the steps of obtaining to-be-standardized target data and a plurality of standardized data elements; according to at least one of a first matching rule corresponding to the Chinese name of the field, a second matching rule corresponding to the English name of the field and a third matching rule corresponding to the data content, matching the target data with all the standardized data elements to obtain a matching result containing a matching degree value; and determining at least one matching result of which the matching degree value is within a preset range in all the matching results, and determining a standardized data element corresponding to each matching result. Based on the method, the to-be-standardized target data and the standardized data elements are matched through the matching rules corresponding to the field Chinese names, the field English names and the data content, the standardized data elements better matched with the target data can be screened out through the multi-strategy matching mode, and the reliability of the matching result is improved.

The invention relates to a managing and control system based on sensitive content perception. The system includes a sensitive content processing system and a cooperation service processing system, wherein the sensitive content processing system further includes a sensitive information acquisition unit, a sensitive information analysis unit and a sensitive service early warning analysis unit. The sensitive content processing system conducts analysis and selection on service flow data and service work order content data and conducts early warning mark output on corresponding content work order. The cooperation service processing system starts a corresponding cooperation service flow and defines the started cooperation service flow is composed of which specific service steps. According to the invention, based on the sensitive content perception, the system establishes quality management and control based on sensitive content perception, addresses the problems of inefficient cooperation service processing and low client satisfaction, conducts qualitative and quantitative analysis on sensitive information, forms real-time online early warning of sensitive service and guarantees ability of processing sensitive service.

Disclosed are a speech processing method and speech processing apparatus which execute an installed artificial intelligence (AI) algorithm and/or machine learning algorithm to perform speech processing to enable the speech processing apparatus, a user terminal, and a server to communicate with each other in a 5G communication environment. The speech processing method according to an exemplary embodiment of the present disclosure includes converting a user's spoken utterance into a user utterance text, discovering a domain, an intent, and at least one named entity of the user utterance text, determining whether the user's spoken utterance is a complete spoken utterance or an incomplete spoken utterance, generating a spoken query utterance and feeding the generated spoken query utterance back to the user when the user's spoken utterance is an incomplete spoken utterance, and constructing a complete spoken utterance by receiving a user's spoken response utterance which responds to the spoken query utterance.

The invention belongs to the field of software coordinative development, and discloses a public contribution combination request repeatability detecting method based on hybrid similarity. The method includes the steps of calculating the text similarity between a newly-submitted public contribution combination request and a historical public contribution combination request, calculating the variation similarity between the newly-submitted public contribution combination request and the historical public contribution, searching a public coordinative development platform for a historical repeatedcontribution data set, combining the text similarity and the variation similarity by means of a weight calculating method based on a greedy search strategy under the training of the data set to calculate the hybrid similarity of public contributions, and finally obtaining a list of historical public contribution combination requests the most probably repeated with a given public contribution combination request according to the value of the hybrid similarity. The public contribution repeatability can be detected in time, repeated artificial code inspection work is avoided, and the public contribution inspection efficiency is improved.

The invention provides a text processing method and device, and the method comprises the steps: obtaining a question text, and recognizing a question entity in the question text; linking a plurality of candidate entities corresponding to the question entity in a preset knowledge base, and identifying a question type of the question text according to the question entity; generating at least one candidate path corresponding to the question text based on the plurality of candidate entities and the question type, and determining a target path according to the at least one candidate path; and converting the target path into a query statement, and querying an answer text corresponding to the question text in the knowledge base according to the query statement.

The invention provides a text generation method and device based on artificial intelligence, a computer device and a readable storage medium, and the method comprises the steps of coding a preceding language statement to obtain a hidden state of the preceding language statement, and carrying out at least twice decoding on the hidden state of the preceding language statement according to semanticsand rhythm of the preceding language statement to obtain a following language statement matched with the preceding language statement. According to the method, at least two times of decoding are carried out on the hidden state obtained by encoding the preceding language statement according to the semantics and the rhythm of the preceding language statement, so that the generated following languagestatement can meet the limitation of the semantics and the rhythm at the same time, the generated following language statement has poetry, and the requirements of users are met.

The embodiment of the invention provides a cross-platform AI model recommendation system and method, and the system is used for carrying out AI model recommendation according to the demands of a user, and comprises an obtaining module which is used for obtaining a model description text of an AI model from a plurality of AI open platforms; the user module is used for acquiring a demand description text of a user for the required AI model; the recommendation module is used for matching the demand description text of the user for the required AI model with the acquired model description text of the AI model, and generating an AI model recommendation list according to the matching degree; according to the method, the model description text of the AI model is obtained from the multiple AI open platforms, the demand description text of the user for the needed AI model is matched with the obtained model description text of the AI model, and the AI model recommendation list is generated according to the matching degree, so that cross-platform AI model recommendation can be efficiently and accurately realized.

The invention relates to a big data technology, and discloses a topic classification method, which comprises the steps of obtaining a bag-of-words model corresponding to each article in an informationbase, the bag-of-words model being a subject term combination formed after stop words and part-of-speech screening, and the bag-of-words model comprising subject terms and occurrence frequencies corresponding to the subject terms; taking the bag-of-words models corresponding to the articles as topics of the articles in a one-to-one correspondence manner, and inputting the bag-of-words models intoan LDA topic model; judging whether the iterative training process of topic classification of the topics of the articles is converged or not by the LDA topic model according to a preset topic number;and if yes, obtaining classification information of topic classification corresponding to each output article when LDA topic model training converges. Bag-of-words models respectively corresponding to articles are respectively formed after stop words and part-of-speech screening and serve as feature input of the LDA topic model, interference of words without content value and appearing at high frequency is avoided, and noise influence in the topic classification process is eliminated.

The invention provides an association rule-based multi-tag Chinese emotion classification method. The method comprises the following steps of: 1, searching a frequent item set among various Chinese emotion tags by utilizing an association rule algorithm; 2, deducing an association rule among the emotion tags according to the frequent item set; 3, modifying a multi-tag data set by using the association rule so as to obtain a new data set; 4, carrying out classified learning on the obtained new data set by using a Rank-SVM algorithm so as to obtain a new model; and 5, predicting a test data set by using the new model.

The invention discloses a sentence trunk extraction method and device based on dependency syntax and a readable storage medium, and the sentence trunk extraction method based on the dependency syntaxcomprises the steps: obtaining a to-be-analyzed sentence, carrying out the dependency syntax analysis of the to-be-analyzed sentence, obtaining a dependency syntax analysis result, and extracting a target sentence trunk corresponding to the to-be-analyzed sentence to perform semantic recognition on the to-be-analyzed sentence. The method solve the technical problem of low sentence semantic recognition accuracy.

The invention provides a multi-language machine translation method and device, electronic equipment and a storage medium. The method comprises the steps of: determining a to-be-translated source language text; and inputting the source language text into a multi-language translation model to obtain a target language text output by the multi-language translation model, wherein the multi-language translation model is constructed based on a pre-training encoder, and the pre-training encoder is obtained through training by taking a unified coding result obtained by coding a multi-language parallel sentence pair as a target. According to the method, the device, the electronic equipment and the storage medium provided by the invention, unified semantic representation of sentences in different languages can be learned through the pre-training encoder, and then the multi-language translation model is obtained based on the pre-training encoder, so that the machine translation model can learn the semantic alignment relationship more easily, and therefore, the multi-language machine translation effect can be improved, and the generalization performance of the multi-language machine translation model is improved.

The invention discloses a text similarity matching method, electronic equipment and a computer readable medium, and the text similarity matching method comprises the following steps: obtaining an input text which comprises at least one input vocabulary; inputting the input text into a text similarity matching model to perform matching prediction, respectively introducing general context information of the at least one input vocabulary into the input text through a graph neural network, and performing matching prediction according to the introduced input text; and outputting the label of the predicted text matched and predicted by the text similarity matching model. According to the text similarity matching method, the similarity matching task of the short text can be effectively completed,so that the matching accuracy and the matching efficiency are greatly improved.

The invention discloses an emotion early warning method, system and device and a storage medium, and the method comprises the steps: obtaining a message text corresponding to a current conversation; processing the message text corresponding to the current dialogue to obtain a corresponding word vector; inputting the word vectors into an emotion recognition model for classification to obtain an emotion recognition result corresponding to the current dialogue, the emotion recognition model being used for recognizing emotions corresponding to the message text; and generating emotion early warninginformation according to the emotion recognition result corresponding to the current dialogue and the emotion recognition result of the historical dialogue corresponding to the current dialogue. Compared with a traditional keyword matching method, the method has the advantages that the recognition accuracy is high, the real-time performance is high, and timely corresponding processing can be performed according to the prediction result so as to improve the dialogue service quality.

The invention discloses a semantic matching model for knowledge point positioning. The semantic matching model is mainly used for solving the problem of question knowledge point positioning in the field of electricity. The method comprises the following steps: firstly, preprocessing an original textbook to form a corpus; then, using statistics-based semantic matching models TF-IDF, LSI and LDA for coding, afterwards, using a deep learning semantic matching model for enhancing deep semantic understanding, and carrying out BERT coding; then, for the four coding modes, calculating cosine similarity to serve as measurement of semantic similarity; and finally, selecting a specified number of textbook fragments by the user as a final knowledge point positioning result according to the number of times that the textbook fragments appear in the front and the cosine similarity based on the voting semantic matching integration model.

The invention provides a scene rendering method and device, a storage medium and electronic equipment, and the method comprises the steps: receiving a rendering instruction sent by a user through a client, and obtaining a question text and a response text corresponding to the rendering instruction; performing semantic scene analysis on the question text and the response text, and determining a semantic scene corresponding to the rendering instruction; and acquiring a scene picture sequence corresponding to the semantic scene, generating a scene rendering data stream based on the scene picture sequence, and sending the scene rendering data stream to the client, so that the client renders response data by applying the scene rendering data stream. The semantic scene is determined by analyzing the question text and the response text, and the scene rendering data stream is generated based on the scene picture sequence corresponding to the semantic scene, so that the client renders the response scene based on the scene rendering data stream, a neural network does not need to be trained for each scene in the process, the time and cost investment can be reduced, and the scene rendering cost is reduced.

The embodiment of the invention discloses an association word recommendation method and device, an electronic device and a storage medium. The method comprises the following steps: receiving a search term input by a user; obtaining an association word corresponding to the search word in an association word library according to the pre-calculated correlation degree between the search word and each central word in each real-time information material; and recommending the associative words corresponding to the search words to the user. In the embodiment of the invention, not only can more associative words with higher writing value be recommended to the user, but also the real-time requirement of the user on the recommended information can be met.

The invention relates to an annual report risk mining system and method based on phrase vector construction. The system comprises an annual report risk information extraction module, a risk factor mining module, a risk phrase vector construction module and an automatic statistics and visualization module. According to the method, related annual reports can be automatically downloaded according to input listed company stock codes and year ranges, risk part texts in the annual reports are extracted, risk phrases are mined from the risk part texts, and a phrase knowledge base is constructed for further mining risk factors; the method comprises the following steps: performing model training on a phrase knowledge base by a Doc2Vector algorithm, and reasoning representative vectors of risk phrases and risk factors; according to the method, the cosine similarity between representative vectors is calculated, and the display information and co-occurrence information of risk factors in annual reports are automatically counted. Compared with the prior art, the method has the advantages that subjectivity of manual judgment is avoided, meanwhile, a large amount of manual operation is saved, and the method is suitable for carrying out risk mining on a large batch of annual reports.