56 results about "Semantic mining" patented technology

The invention relates to an automatic standardized filing method based on text semantic mining. The automatic standardized filing method is characterized by comprising the steps: crawling files from a website, and carrying out information extraction, key word extraction and automatic abstract generation on the crawled file and a local file by utilizing text semantics, and finally storing into an informatization system. For the information extraction, a rule set is established by adopting a knowledge engineering method, information points are automatically extracted from the file to form structural data; for the key word extraction, a key word is extracted according to a position and semantics of a word in a text to generate a key word index; for the automatic abstract generation, a content contained by the abstract is firstly set, corresponding information is extracted from the text, the similarity of sentences is calculated, and the texts including the key information in the original file are extracted. By adopting the automatic standardized filing method, business personnel do not need to read a great amount of files, time and labor are saved, and convenience in inquiry and application can be realized.

The invention discloses a multi-strategy fused method and device for identifying a named entity. A first identification result is obtained by utilizing a first identifying model to identify the named entity in obtained linguistic data, the first identifying model can update and extend a corpus in the method, accordingly the newly produced named entity in the linguistic data can be identified, thus a first identifying result has higher accuracy rate, the named entity in the linguistic data is identified by utilizing a multi-identifying-model fused method to obtain a second identifying result, the first identifying result and the second identifying result are fused to obtain a third identifying result, then a semantic mining system is utilized to conduct role assignment on the named entity, the named entity having a role is output, accordingly it is achieved that the named entity is reliably identified under the situations that data is massive, the entity type is diversified and new words emerge in endlessly, and role assignment is conducted on the identified named entity.

The invention discloses an in-text advertisement releasing method and an in-text advertisement releasing system based on deep semantic mining. The method comprises the following steps that an advertisement demand body is built; webpage contents are captured and received, webpage irrelevant to business information is removed according to the advertisement body and the text classification algorithm, the webpage types are judged, and keywords and key sentences are extracted; according to the linguistic rule, the deep semantic mining is carried out on the extracted key sentences, sentences, phases or words with commercial properties and with demands, emotion and attitude are discovered and extracted, and in addition, the advertisement marking is carried out; precise advertisement is embedded in an advertisement marking region through a generating type system, and when users browse the kind of webpage, the advertisement is shown in the specific region. The method and the system provided by the invention have the advantages that the advertisement relevant to the user reading content context demands can be issued in the webpage text contents, which page is suitable for in-text advertisement releasing in the website and the placing region and the advertisement words of the advertisement in the page can be analyzed, and the technical problems in the prior art can be solved.

The invention relates to a Web portrait retrieval method fused with text semantics and visual content, which comprises steps of submitting 'a query string'to a commercial search engine server to realize functions of connecting and downloading based on a HTTP protocol, downloading picture output of the commercial picture search engine and relevant websites to be a local image library, and simultaneously, extracting key tags of original websites to form XML files for post text processing, further, utilizing the AdaBoost face detection technique, mining the high-level semantics of vector models to webpage scripts containing pictures, comparing via using experienced weights and a method of dynamically weighting based on PLSA, dynamically combining visual analysis results of image characteristics and text analysis results of image characteristics via a regulating factor, obtaining rank values of relevancy of images and query, reordering the image output list of the search engine and feeding it back to users. The method has higher precision rate which is greatly increased after fusion of the characteristics.

The invention provides a system and a method for human and machine voice interaction based on Java Map. The system comprises a voice identifying module, a spoken language comprehending module, a dialogue management module, a language generating module and a voice synthesizing module, wherein the voice identifying module is used for receiving voice information inputted by a user, and identifying the voice information into text data, the spoken language comprehending module is used for performing semantic mining on the text data, and converting the text data into a type which can be identified by a machine, the semantic mining is used for integrating contextual information inputted by the user according to a storage and utilization strategy of context key semantic elements of the Java Map, and extracting semantic key elements of the identified text, the dialogue management module is used for controlling the dialogue process of the human and machine interaction, the language generating module is used for integrating fragmentary answers to obtain a fluent text which meets the logic language expression type of people, and the voice synthesizing module is used for converting the generated answer text into voice information, and broadcasting the voice information to the user.

The invention discloses a machine learning-based commodity comment data sentiment analysis method. The method comprises the steps of collecting and extracting commodity comment data; preprocessing thedata, wherein the preprocessing comprises text duplicate removal, mechanical compression word removal and short sentence deletion; based on a Jieba word segmentation method, performing text word segmentation on the preprocessed data; building a sentiment analysis model: generating word vectors based on NNLM (Neural Network Language Model) training, and establishing a semantic network; and based on an LDA topic model, performing semantic mining, and generating a topic in an unsupervised manner. The unsupervised sentiment analysis method is realized; and a result shows that the sentiment analysis mode can effectively analyze comment sentiments of users.

The invention discloses a distribution network earth fault type identification device and method based on deep learning. Semantic mining is carried out on fault recording data by using a deep learningtechnology, a fault identification model with self-learning capacity is constructed by using incremental learning and deep reinforcement learning technologies, automatic identification for distribution network earth fault type is realized, the restriction that only route selection positioning is carried out in traditional earth fault analysis is broken through, and more abundant fault processingdecision-making information is provided; a positioning method based on a transient state zero-sequence current similarity principle is used, and compared with a traditional experience analysis accident processing method, the method has higher accuracy and wider adaptation; fault type and fault location result are synthesized, the decision basis is reasonable, the method is of more pertinency for solutions, and an accident processing plan can be arranged more reasonably.

The invention discloses an automobile public praise semantic sentiment analysis system. The system comprises a real-time monitoring module, a semantic mining module and a system display module, the real-time monitoring module is used for acquiring automobile user comment data in real time in a whole network according to a crawler technology; the semantic mining module is used for mining theme keywords related to the automobile from the comment data and performing emotional tendency analysis on the theme keywords to obtain emotional analysis data; the system display module performs statisticalanalysis on the whole vehicle model, key attributes and product competitiveness details according to the emotion analysis data to form a visual chart; according to the invention, related comment corpora of the user are analyzed by using an artificial intelligence algorithm; the technical problem that in the prior art, no system specially analyzes the evaluation of the automobile by the user, and the evaluation of the automobile by the user cannot be quickly known is solved, the evaluation of the automobile by the user can be quickly known, the system can be in a favorable position of knowing the user in competition, and the system can be widely applied to the automobile industry.

The invention discloses a method for marking marketing calls by means of a semantic mining algorithm and a system for governing the marketing calls. The method for marking the marketing calls by meansof a semantic mining algorithm method for marking the marketing calls by means of the semantic mining algorithm particularly comprises the steps that S1, labels of calls are divided into different types; S2, a dictionary base is built, wherein the dictionary base comprises word vectors corresponding to the labels; S3, the labels, belonging to the same type, in the dictionary base are extracted, and a layer of training sample is formed; S4, training is conducted by means of multiple layers of training samples, and a classification model is obtained; S5, according to the classification model, the types to which the word vectors in the dictionary base belong are marked. The invention further discloses the system for governing the marketing calls by means of the semantic mining algorithm. According to the method for marking the marketing calls by means of the semantic mining algorithm and the system for governing the marketing calls, the marketing calls can be accurately classified, a user can independently select the types of the incoming calls, and the purpose of precisely intercepting the marketing calls is achieved.

The invention discloses an online semantic excavation system of Chinese polysemic words and based on a uniform resource locator (URL). The system utilizes a webpage classification method based on the URL and can conduct semantic excavation on the Chinese polysemic words online. The process includes first constructing a URL classifier through an online URL classification catalogue; then classifying searching results (including webpage URLs and abstracts) of the polysemic words returned by a search engine by means of the URL classifier to obtain initial semantic classification results of the polysemic words; finally clustering the initial semantic classification results according to the webpage abstracts to obtain semantic excavation results of the polysemic words. The semantic excavation system has ideal accuracy and recall rate and is highly applicable to semantic excavation of network popular words.

The invention discloses a comprehensive urban geographic semantic mining method based on multivariate big data. Social text data is a data source capable of best reflecting the cognition of people onurban area functions, so that the functions of urban areas are extracted by utilizing the social text data; based on the bus route data, automatic calculation is performed to obtain the relative rankof the traffic convenience of each region of the city without depending on artificially formulated rules; the urban population density distribution is analyzed from two macroscopic perspectives of thepopulation density index of the working time period of the workday and the population density index of the rest time period. According to the method, the comprehensive urban geographic semantics aredescribed from four different indexes including urban area functions, urban traffic convenience distribution, building functions and population density indexes; and in combination with the informationmined by the four indexes, different types of query requirements of different types of users can be met, and people can be better helped to comprehensively understand cities.

The invention discloses a method for constructing a spatial semantic database based on BIM and GIS data integration. The method comprises the following steps: analyzing spatial data of BIM and GIS models; establishing mapping rules of object model related attribute description fields of IOT equipment and BIM and GIS model object attribute description fields, and constructing a relational database;extracting spatial semantic information, establishing a spatial semantic mapping relation table, and constructing a graph database; constructing a time sequence database according to the time sequence data transmitted by the IOT equipment; storing spatial semantic information in a graph database, storing relational data and static business data in a relational database, storing Internet-of-thingsdata in a time sequence database, and reading data interfaces of all the databases through a visualization engine. According to the method, a data storage standard of a BIM + GIS model is established, and fusion application is realized, so that a digital twin system application mode and a technical framework integrated with key technologies such as ontology, semantic mining, Internet-of-things, big data analysis and the like are constructed.

The invention discloses an N-gram-based semantic mining method for increment of a topic model. The N-gram-based semantic mining method comprises the steps of: (1) expanding an Author-Conference topic model, wherein a word space is expanded from Unigram to N-gram; (2) calculating a prior probability parameter of a current model according to linear weighting of posterior probability in a prior trained model with respect to current input data; (3) calculating a posterior probability value of the current model to the current data by adopting a Gibbs sampling method; and (4) repeating the steps (2) and (3) for training the model in an increment manner with respect to newly input data streams. According to the invention, the N-gram is introduced into the topic model and the property of the topic model for modeling scientific and technical literature is improved according to the semantic features contained by the N-gram; and the topic distribution of historic data is recorded by adopting asymmetric prior probability so as to train the model in an increment manner, and the efficiency of the method is increased.

The invention discloses a binocular image rapid target detection method based on a double-flow convolutional neural network, and the method comprises the steps: carrying out the calibration of a binocular camera, and obtaining calibration parameters; correcting a training image according to the calibration parameters, and training an implicit deep semantic mining network for implicitly learning deep semantic information on the binocular image and training a multi-modal feature hybrid detection network; combining the features output by the implicit deep semantic mining network with the featuresof the multi-modal feature hybrid detection network in a channel series connection manner to form a double-flow convolutional neural network, and training the double-flow convolutional neural networkby using the training image; and obtaining a test image through the binocular camera, correcting the test image, and inputting the corrected image into the double-flow convolutional neural network for target detection to obtain a target detection result. The complementarity of RGB and deep semantic information can be comprehensively utilized, and the method has the advantages of being high in efficiency and more accurate in target detection result.

The invention discloses a depression auxiliary detection method based on an open question and answer text, which is mainly characterized in that the depression auxiliary detection method is a text analysis method suitable for dialogue between a depression patient and a doctor, a model is retrained along with the increase of dialogue data to achieve a self-learning improvement effect, and the depression auxiliary detection method specifically comprises the following steps: text collection: extracting a dialogue text by using a voice recognition technology; text preprocessing: performing text preprocessing according to data features and model design; training a text semantic model, and performing semantic mining on the preprocessed text; predicting a text model and classifying features output by the semantic mining model. The model is self-learned, and the model can be self-iterated and updated along with the increase of the number of recognized samples.

The invention relates to a financial risk prediction method and device based on text pre-training and multi-task learning. The method comprises the steps of obtaining a to-be-processed text; inputting the to-be-processed text into a first neural network model to determine whether the content of the to-be-processed text includes financial risks according to the processing flow of the risk identification task; under the condition that the content of the to-be-processed text comprises the financial risk, determining the risk type of the financial risk according to the processing flow of the risk classification task by utilizing a first neural network model; and determining a risk subject matched with the risk type by using the first neural network model according to a processing flow of the risk subject identification task. According to the method and the device, the problem of poor model performance caused by lack of deep semantic mining is solved through a pre-training language model technology, and the technical problems of limited data volume and poor model performance caused by incapability of information sharing among tasks are solved by adopting multi-task processing.

The embodiment of the invention provides a distribution network first-aid repair work order processing method and device, and the method comprises the steps: importing to-be-processed distribution network first-aid repair work order data and a pre-established expression of a hidden Markov model into a Viterbi algorithm for iterative calculation, and obtaining a calculated state sequence corresponding to a distribution network first-aid repair work order; carrying out convolution operation on the obtained state sequence, classifying and sorting operation results according to different power failure reasons, and obtaining and counting specific reasons used for representing power grid faults in the distribution network repair work order. According to the method, the distribution network first-aid repair work order is used as the recognition model for matching, semantic mining and learning are carried out on the work, and adaptive work order processing is realized, so that the data processing efficiency is improved, the reliability is improved, the active recognition rate of the work order is further improved, and the distribution network repair work efficiency is enhanced.

The invention relates to a dynamic expanding knowledge graph inference method oriented to China Mobile intelligent customer service, and belongs to the technical field of artificial intelligence. Themethod comprises the steps of constructing a ternary knowledge graph representing learning; executing a knowledge graph Top-k query technology under a limiting technology; conducting single-business multi-round dialogue scene knowledge inference; conducting cross-business multi-round dialogue scene new-semantics mining. By means of the dynamic expanding knowledge graph inference method oriented tothe China Mobile intelligent customer service, natural language information with effective response interaction is provided for a user, and the requirements for multi-round dialogues with the intelligent customer service are met.

The invention discloses a pseudo-correlation feedback extended query method based on a question-answering system. According to the pseudo-correlation feedback extended query method based on the question and answer system, mature semantic mining modules, such as attention mechanisms, in the question and answer system are used for reference, so that the model can truly understand the search intention of a user, and extended lexical items are selected according to interactive semantic information of query and documents. Compared with a traditional model, due to the fact that semantic interactioncharacteristics are added, the effect of selecting the extended lexical items is remarkably improved. In addition, a neural network based on a paired loss function is further added to understand statistical characteristics of lexical items, and word frequency, inverse document frequency and the like are used for correcting the possible semantic drift problem of the semantic model. Practice provesthat compared with a traditional pseudo-correlation feedback algorithm, the method has higher sorting accuracy and better robustness, and can be applied to various search scenes.

The invention discloses a content semantic mining method for a non-structured big data stream. The method comprises the steps of S1: extracting a text link, a label attribute and a semantic tendency keyword in the big data stream, and correspondingly defining text nodes, label nodes and content nodes; S2: constructing a text node set containing the text nodes and a label node set containing the label nodes, calculating and outputting weight values from the text nodes to the label nodes and weight values from any label node to other all label nodes; S3: according to the text node set, the label node set, the weight values from the text nodes to the label nodes and the weight values from any label node to other all label nodes, performing semantic classification on the content nodes and constructing different content node classification sets; and S4: according to the text node set and the content node classification sets, performing weighted small-world network clustering calculation on the text nodes to obtain a text node cluster set.

The invention discloses an urban functional area identification process based on space-time semantic mining, which comprises documents, words, basic functional units, space-time data, a topic model, document topic distribution and unit function distribution, and is characterized in that firstly, hidden functions of an area are tried to discover through the topic model, compared with a text theme mining, the basic function units are equivalent to the documents in a corpus, space-time data in the basic function unit is similar to words in the document, unit function distribution obtained after passing through the topic model is equivalent to document topic distribution, and the used city space-time data is representative Sina microblog position sign-in data. Each piece of sign-in data comprises user information, space coordinates of sign-in positions, publishing time, publishing texts and the like. Dynamic activity modes of people can be reflected from different angles, meanwhile, POIs in a research area are obtained from a Baidu map, and function recognition of the area is achieved.