30 results about "Analytic language" patented technology

A method of, and system for, generating a sentence from a semantic representation maps the semantic representation to an unordered set of syntactic nodes. Simplified generation grammar rules and statistical goodness measure values from a corresponding analysis grammar are then used to create a tree structure to order the syntactic nodes. The sentence is then generated from the tree structure. The generation grammar is a simplified (context free) version of a corresponding full (context sensitive) analysis grammar. In the generation grammar, conditions on each rule are ignored except those directly related to the semantic representation. The statistical goodness measure values, which are calculated through an analysis training phase in which a corpus of example sentences is processed using the full analysis grammar, are used to guide the generation choice to prefer substructures most commonly found in a particular syntactic / semantic context during analysis.

The embodiment of the invention discloses a text abstract automatic generation method and a device based on deep learning, and a storage medium. The method comprises the following steps: constructinga Chinese corpus; preprocessing the corpus text in the corpus to extract a keyword table in the corpus text; analyzing sentence feature vectors between sentences in the corpus text and sentence feature vectors of the sentences, and constructing a feature model after sentence vectorization according to the sentence feature vectors; and training the to-be-processed text by adopting the feature modelafter sentence vectorization to establish a model, and generating a corresponding text abstract. By adopting the method and the device, the omission of valuable contents can be avoided on the premiseof ensuring the abstract generation accuracy.

The invention belongs to the field of artificial intelligence, the invention relates to a text information characterization method and a system, computer equipment and a storage medium, and the methodcomprises the steps: obtaining a to-be-analyzed corpus, carrying out the word segmentation preprocessing of the to-be-analyzed corpus, generating corresponding word vectors based on the obtained segmented words, enabling the to-be-analyzed corpus to be text information, and enabling the text information to comprise at least one statement; obtaining word vectors of segmented words contained in each statement in the to-be-analyzed corpus to obtain a word vector group of each statement, and sequentially inputting the word vectors in the word vector group into the initial sentence vector algorithm model according to a sequence to generate initial sentence vectors of the corresponding statements; and inputting the initial sentence vector into a pre-trained sentence vector model to obtain a final sentence vector of each sentence, the final sentence vector being used for representing text information, and the pre-trained sentence vector model being generated based on a context relationship of the sentences. According to the method provided by the invention, the influence caused by different semantics of words in different sentences can be avoided, and the representation of the text information is more accurate.

The invention relates to a Vietnamese multi-category word disambiguation method based on a combination method, and belongs to the technical field of natural language processing. The method comprises the following steps: firstly, extracting Vietnamese multi-category word fields from a Vietnamese text, and constructing a multi-category word field library; secondly, respectively performing maximum entropy, conditional random field and support vector machine disambiguation modeling on the multi-category word field library; disambiguating a multi-category word field testing material to be disambiguated through the three constructed statistic analysis models, and obtaining part-of-speech tags of multi-category words according to analysis results. According to the Vietnamese multi-category word disambiguation method based on the combination method, the Vietnamese multi-category words are effectively disambiguated, and power support is provided for the subsequent work such as Vietnamese part-of-speech tagging, lexical analysis, syntactic analysis, semantic analysis, information extraction, information retrieval and machine translation; the problem of poor generalization performance due to a single learner is solved.

The embodiment of the invention provides a semantic comprehension processing method and device, equipment and a readable storage medium, and the method comprises the steps: carrying out the word segmentation of a to-be-analyzed statement, and setting corresponding labels for a word segmentation result, wherein the labels comprise part-of-speech labels used for representing universal part-of-speechand dictionary labels used for representing special part-of-speech; substituting the label into the to-be-analyzed statement to obtain an updated to-be-analyzed statement, and matching the updated to-be-analyzed statement with a preset matching rule to obtain a matching result, wherein the matching rule comprises an intention and a rule, and the rule at least comprises an N-tuple formed by connecting segmented words through separator marks; representing keywords in the rules by the dictionary labels corresponding to the keywords, and representing non-keywords in the rules by the non-keywordsthemselves. The rule in the embodiment of the invention supports multiple matching modes, and can meet the generalization requirement of the semantic comprehension rule.

The invention relates to an analytic language text type analysis and conversion method, which comprises the following steps of: S100, listing programming languages used by all client environments, decomposing the programming languages into 23 language feature definitions, and configuring related language description configuration files for each supported language; S200, reading various features ofthe supported language to an analysis converter; S300, initializing an analysis class corresponding to each feature according to the configuration features, and defining assignment and initializationof an analyzer according to the semantic features; S400, inputting a source code file needing to be converted, reading a source code to become a string, and providing the string for the language conversion processor; S500, creating a source language feature definition main body; S600, creating a target language feature definition main body; S700, using a source semantic feature definition analyzer to uninterruptedly and circularly decompose the ring of the source code until complete conversion is generated; and S800, storing the generated string as a target code file. The method can reduce the development cost, and has the advantages of low failure rate and high controllability.

The invention relates to a maximum entropy based Vietnamese cross ambiguity elimination method and belongs to the technical field of natural language processing. The method comprises the steps of firstly performing disambiguation modeling on Vietnamese cross ambiguity field corpora in a formed Vietnamese cross ambiguity field library to obtain a Vietnamese maximum entropy cross ambiguity elimination model; and randomly selecting test corpora from the Vietnamese cross ambiguity field corpora and performing disambiguation through the established Vietnamese maximum entropy cross ambiguity elimination model to obtain a disambiguated parameter sequence. The method effectively eliminates the ambiguities of Vietnamese cross ambiguity words and provides powerful support for work such as lexical analysis, syntactic analysis, semantic analysis, information extraction, information retrieval, machine translation and the like; at present, no related Vietnamese cross ambiguity elimination reports are discovered; and the method achieves a very good effect.

The embodiment of the invention discloses an information processing method and device, equipment and a storage medium. The method comprises the following steps: at least two candidate hot words in a to-be-analyzed phrase being extracted; calculating a first heat degree of each candidate hot word based on a Bayesian average method; calculating a second heat degree of each candidate hot word based on the Newton cooling law; performing weighted calculation on the first popularity and the second popularity of the same candidate hot word to obtain comprehensive popularity of the corresponding candidate hot word; and sorting at least two candidate hot words in the to-be-analyzed phrase according to the comprehensive heat of each candidate hot word, and determining the hot words in the to-be-analyzed phrase from the sorting result according to a set sorting sequence. According to the embodiment of the invention, the hot degree of each entry is accurately calculated by analyzing the network information, so that the hot words are mined.

The invention provides a method, device and equipment for constructing a causal relationship determination model. The method comprises the steps of obtaining an original corpus set, wherein the original corpus set comprises at least one first candidate text; screening a second candidate text in the original corpus set according to a target event causal relationship template; for each second candidate text, determining a statement vector of the second candidate text according to the word vector of each word and the word vector of each character in the second candidate text; and training a to-be-trained causal relationship determination model based on the statement vector of each second candidate text to obtain a trained causal relationship determination model. According to the invention, firstly, the words and the characters are used as representation forms for expressing semantics at the same time, the semantic expression accuracy is improved, and then the text with the recessive causal relationship can be effectively recognized by analyzing the semantics.

The invention provides an intention recognition method, system and device based on deep learning and a storage medium. The method comprises the steps of acquiring and preprocessing the statement information to be analyzed; performing word segmentation on the statement information by using a word segmentation tool; weighting all the segmented words by using a pre-trained word vector model; obtaining a mood classification corresponding to the statement information by using a pre-trained first neural network; obtaining sentence pattern classification corresponding to the sentence information by using a pre-trained second neural network; obtaining expression classifications corresponding to the statement information by using a pre-trained third neural network; and predicting the intention category of the statement information by using a pre-trained fourth neural network according to the mood classification, sentence pattern classification and expression classification of the statement information. According to the method, the feature information of the statement text information in the corresponding field does not need to be extracted, and the intention type to be expressed by the statement can be accurately determined from the statement text information by analyzing the word vector of each segmented word in the statement text information.

The invention discloses a regular intention recognition method based on natural speech processing. The method comprises the following steps: acquiring language information input by a user; processing a natural language; performing semantic analysis by using NLP; summarizing language information; semantic sentiment analysis; establishing an intention recognition regular model; extracting the regular keywords; judging a user intention according to the regular keyword; creating an intention preset model; and enabling a user to select the intention. According to the method, the user selects the intention, the establishment of the intention recognition regular model is more convenient, and the language information can be recognized and judged only according to the regular keyword information in the language information, so that the whole judgment and recognition process is more convenient, and the intention recognition result is more convenient to obtain; during intention result judgment, when different judgment results exist, the user can conveniently select the intention to be expressed by himself or herself according to the language information by creating the intention preset model, so that the intention recognition and judgment result is more accurate.

A method for syntactic analysis of natural language is disclosed. The present invention points out some serious technical loopholes in the two internationally leading natural language syntax analysis devices recognized by the computer science community today - Berkeley Parser (Berkeley Parser) and Stanford Parser (Stanford Parser), and aims at these loopholes, A technical solution to the problem is given. The invention establishes a set of brand-new mathematical models for characterizing sentences, and on the basis of this, proposes a set of computer syntax analysis methods. The present invention organically unifies the three aspects of lexical analysis, syntactic analysis and semantic analysis in computer natural language processing through technical means, and strengthens the mutual constraints among these three aspects, thereby improving the effect of computer dissolving structural ambiguity . The invention has high technical difficulty, strong comprehensiveness, wide application range and very large calculation amount, conforms to the natural law of mathematics and computer science, and helps to improve the accuracy of computer syntax analysis.

The invention relates to a maximum entropy based Vietnamese cross ambiguity elimination method and belongs to the technical field of natural language processing. The method comprises the steps of firstly performing disambiguation modeling on Vietnamese cross ambiguity field corpora in a formed Vietnamese cross ambiguity field library to obtain a Vietnamese maximum entropy cross ambiguity elimination model; and randomly selecting test corpora from the Vietnamese cross ambiguity field corpora and performing disambiguation through the established Vietnamese maximum entropy cross ambiguity elimination model to obtain a disambiguated parameter sequence. The method effectively eliminates the ambiguities of Vietnamese cross ambiguity words and provides powerful support for work such as lexical analysis, syntactic analysis, semantic analysis, information extraction, information retrieval, machine translation and the like; at present, no related Vietnamese cross ambiguity elimination reports are discovered; and the method achieves a very good effect.

The invention provides a method for constructing a universal query language for multiple storage systems. The method comprises the following steps: designing grammar of a unified query language; compiling a description script of a grammatical rule; using a compiler, taking the grammar rule file as input, and generating an executable grammar analysis code; inputting specific statements of the unified query language, and generating a grammar analysis tree through lexical analysis and grammar analysis processing; analyzing contents in the grammar analysis tree, and translating the grammar analysis tree into a target query language in combination with a target storage system type; executing the target query language to realize query of the target storage system; and converting the query result into a uniform structure to be provided for the query client. Storage system selection is simplified, and the freedom degree of storage system selection is improved; a business system developer only needs to learn one grammar to realize data query of multiple storage systems; the invention has high expansibility, is convenient to transplant, is compatible with an existing storage system, and improves the development efficiency of application software.

The embodiment of the invention provides an interactive data analysis method and device and a related device or equipment, and belongs to the technical field of data analysis. The method comprises the steps that after a to-be-analyzed statement expressed according to a natural language is received, the to-be-analyzed statement is processed according to an analysis tool configured with a Chinese recognition function, a target keyword of the to-be-analyzed statement and a sentence pattern structure composed of part-of-speech identification of the target keyword are obtained, target data are obtained according to the target keyword, and the target data are stored in the analysis tool; and displaying the target data to the matched chart set through the sentence pattern structure. According to the method provided by the invention, data analysis is carried out in combination with artificial intelligence and visualization means, and the threshold of data analysis is also effectively reduced.

The invention relates to a Vietnamese multi-category word disambiguation method based on a combination method, and belongs to the technical field of natural language processing. The method comprises the following steps: firstly, extracting Vietnamese multi-category word fields from a Vietnamese text, and constructing a multi-category word field library; secondly, respectively performing maximum entropy, conditional random field and support vector machine disambiguation modeling on the multi-category word field library; disambiguating a multi-category word field testing material to be disambiguated through the three constructed statistic analysis models, and obtaining part-of-speech tags of multi-category words according to analysis results. According to the Vietnamese multi-category word disambiguation method based on the combination method, the Vietnamese multi-category words are effectively disambiguated, and power support is provided for the subsequent work such as Vietnamese part-of-speech tagging, lexical analysis, syntactic analysis, semantic analysis, information extraction, information retrieval and machine translation; the problem of poor generalization performance due to a single learner is solved.

The present invention provides a keyword extraction method, which includes: performing a word segmentation operation on all the analysis sentences of the extracted text to obtain word units of the analysis sentences; acquiring word features of the word units, sentence features of the word units in the corresponding analysis sentences, and the text features of the word unit in the extracted text; based on the machine learning model established by the machine learning algorithm, using the word feature, sentence feature and text feature of the word unit in each analysis sentence, keyword analysis is carried out for each analysis sentence Fetch operation. The present invention also provides a keyword extraction device. The keyword extraction method and the keyword extraction device of the present invention use the word features, sentence features and text features of the word unit to establish a machine learning model, thereby performing keyword extraction for each analysis sentence operation, thereby improving the accuracy of keyword extraction.

The invention relates to the field of intelligent decision in artificial intelligence, and discloses a semantic emotion analysis method, which comprises the steps of obtaining an importance metric value corresponding to each word in a to-be-analyzed statement; obtaining an implicit expression corresponding to the to-be-analyzed statement through two parallel running recurrent neural network modelsaccording to the importance metric value corresponding to each word in the to-be-analyzed statement; inputting the implicit expression corresponding to the to-be-analyzed statement and the sentence label corresponding to the to-be-analyzed statement into a semantic sentiment analysis classifier; and receiving an emotion analysis classification result of the semantic emotion analysis classifier for the to-be-analyzed statement. By introducing a self-attention mechanism, the importance of each word in the sentence is quantified through an importance measurement value, and then the meaning of the current word in the whole sentence is obtained according to the position of the important word, so that each word in the sentence and the corresponding importance measurement value thereof are fusedin the hidden state of the finally output whole sentence.

The invention provides a program vulnerability analysis method based on executable formal semantics of a Go language, which comprises the following steps of: (1) studying an expanded barkes normal form officially given by the GO language, and providing a method for converting GO [EBNF] into a general barkes normal form in combination with a BNF grammar format supportable by a K framework to obtain GO [BNF]; (2) defining an executable semantic GO [KS] of the Go language based on rewriting logic by applying a K framework according to the characteristics of the GO [BNF] and the Go language; (3) analyzing the correctness of GO [KS] by adopting a method based on test cases, separately executing the test cases in the K framework and the IntelliJ IDEA tool, and analyzing the correctness of semantics by comparing the consistency of the results of the test cases; and (4) based on the K framework, searching vulnerabilities of the Go program by applying GO [KS]. Integrity of semantics of the Go language is improved, and analysis of program vulnerabilities is realized.

The invention discloses a method and device for sentence analysis, which relates to the technical field of language processing, and its main purpose is to solve the problem that in the existing sentence analysis process, the results of semantic analysis are easily interfered by non-keywords, thus leading to accurate sentence analysis Sexual susceptibility issues, improve the accuracy of sentence analysis results. The main technical scheme of the present invention is: obtain basic vocabulary from the sentence to be analyzed; Carry out feature extraction to described basic vocabulary and the sentence to be analyzed respectively, obtain the word feature corresponding to each basic vocabulary and the sentence feature of the sentence to be analyzed; According to the Describe the word features of each basic vocabulary and the sentence features of the sentence to be analyzed, determine the non-key words in the sentence to be analyzed; remove the non-key words from the sentence to be analyzed, so as to eliminate non-key words The sentence to be analyzed after the vocabulary is analyzed. The present invention is used in the process of analyzing sentences.

The embodiment of the present invention discloses a method, device, and storage medium for automatically generating text summaries based on deep learning, wherein the method includes the following steps: constructing a Chinese corpus; preprocessing the corpus text in the corpus to extract the keyword table in the corpus text; Analyze the sentence feature vectors between the sentences in the corpus text and the sentence itself, and construct the sentence vectorized feature model according to the sentence feature vector; use the sentence vectorized feature model to train the text to be processed and build a model to generate the corresponding text summary . By adopting the present invention, the omission of valuable content can be avoided under the premise of ensuring the accuracy of the abstract generated.

The invention relates to an overhaul application structured analysis method based on topology and business feature points, which is technically characterized by comprising the following steps of: establishing a feature point knowledge base corresponding to an overhaul application according to overhaul content in a power operation process; extracting or analyzing a power failure range related to the maintenance application; analyzing feature points related to the maintenance application based on the maintenance content; analyzing possible modes based on the maintenance application feature points and the power grid topology; and performing structured analysis on the opinions of all the rooms in combination with selectable modes. The method is reasonable in design, and the barriers caused by different text writing methods in the natural language processing process are eliminated by conducting standardization processing on the text data; according to the method, the equipment range and the analysis content of the text content are limited and refined, specific text analysis grammar rules can be conveniently adopted, targeted structured analysis can be carried out on various contents of the maintenance application, the labor cost can be greatly saved, and the working efficiency is improved.

The invention discloses a Chinese text feature extracting method with a text mood fusion function. By means of the method, it is achieved that the text feature representation fusing mood features, syntax features and semanteme features is obtained in a lengthened text. The method comprises the steps that firstly, a text word set and a mood word set are constructed, the text word set and the mood word set are transformed into word embedding forms respectively, and corresponding vector models are obtained; secondly, according to the text word embedding represented time step dimensions and feature dimensions, text features are screened, the mood features are fused into the time step dimension of the selected text feature, and the text feature representation which accurately represents the semanteme is obtained. According to the method, the contributions of modal particles to the text semanteme are fully utilized to fuse the mood features, the syntax features and the semanteme features into the text feature representation, and the text feature representation is low in dimension and continuous so that the text semanteme can be better represented, and natural language processing tasks, such as text analysis, language translation and relation extraction, can be better effectively supported.

The invention provides a classical Chinese analysis method and system. The method comprises the steps: obtaining easy-to-confuse keywords in classical Chinese, and building a keyword library accordingto the keywords; obtaining a classical Chinese sample and corresponding sample semantics; determining sample attribute information according to the classical Chinese samples and the sample semanticsin combination with a keyword library; establishing a semantic concept map according to the sample attribute information and a keyword library; obtaining a classical Chinese corpus; comparing the classical Chinese corpus with a keyword library, and determining corpus keywords contained in the classical Chinese corpus; analyzing attribute information of the corpus keywords according to the linguistic data; analyzing the corpus keywords and the attribute information to determine corpus keyword semantics corresponding to the corpus keywords in combination with the semantic concept graph; and analyzing the linguistic data in combination with the linguistic data keyword semantics to determine the linguistic data semantics. According to the method, the classical Chinese corpus questioned by theuser is analyzed through the conceptual graph, so that the words easy to confuse in the classical Chinese corpus are quickly and accurately determined.

The present invention discloses an implicit feature recognition method based on a word vector model. The method comprises the following steps: obtaining a training corpus, preprocessing the training corpus and then constructing a corresponding emotion word dictionary and an attribute word dictionary; for the preprocessed training corpus, using the word vector model to form a total dictionary, and solving the word vector of each word in the total dictionary to form a corresponding word vector matrix, wherein each row of the word vector matrix corresponds to a word vector of one word in the total dictionary; according to the word vector matrix, setting parameter matrices of the input layer to the mapping layer and the mapping layer to the output layer in the word vector model to obtain a trained word vector model; and using the trained vector model to carry out implicit feature recognition on each implicit feature clause in the to-be-analyzed corpus. According to the method disclosed by the present invention, starting from understanding the sentence semantics, the word vectors are used to represent the semantic information of the words, and the word vector model is used to recognize the attribute words of the implicit feature clauses, so that the recognition accuracy is improved.

The application relates to a construction method and device of a field knowledge library, computer equipment and a storage medium. The method includes: acquiring field core words in a current to-be-constructed field; acquiring corresponding to-be-processed text according to the field core words; splitting the to-be-processed text to obtain corresponding to-be-analyzed sentences; dividing the to-be-analyzed sentences to obtain a plurality of to-be-analyzed words; obtaining target center words from the plurality of to-be-analyzed words according to a preset rule; carrying out syntax structure analysis on the to-be-analyzed sentences to obtain target syntax structures corresponding to the to-be-analyzed sentences; and establishing relationships on the target center words, the field core wordsand the target syntax structures, and storing the relationships into the field knowledge library corresponding to the to-be-constructed field. By using the method, accuracy of knowledge extraction can be improved.

The invention discloses a neural network, a training method, an aspect-level sentiment analysis method and device and a storage medium. The neural network comprises the steps that a BERT word embedding module obtains semantic information of an input text and word vectors of all words of a to-be-analyzed statement in the input text; the first full connection layer obtains a semantic feature vector based on the semantic information; the second full connection layer obtains word feature vectors based on the word vectors; a potential Dirichlet distribution module extracts topic distribution of a statement to be analyzed and topic distribution of an aspect emotion pair; the feature fusion layer judges whether the to-be-analyzed statement is matched with the aspect emotion pair or not; a decoding marking module marks the position of the target word in the decoding sequence to obtain a marking sequence; and the output layer outputs an emotion analysis result based on the judgment result output by the feature fusion layer. And multi-task output is realized, semantic information and topic distribution are subjected to feature fusion through the feature fusion layer, vocabularies in a specific field are learned, and the sentiment classification task effect of the model is improved.

The invention provides a Chinese comment-oriented sentiment multi-tendency classification method, which comprises the following steps of: firstly, extracting morpheme words and sentiment words; secondly, constructing a similarity relationship between morpheme emotion variables; and finally, calculating a morpheme emotion close path. According to the method, morpheme emotion variables are regarded as nodes in a directed weighted acyclic graph, directed weighted relation connection is constructed between morpheme emotion nodes to serve as directed weighted link edges, and on the basis of the directed weighted link edges, an effective path meeting a certain weight condition is searched. According to the method, a directed weighted acyclic graph model is combined with emotion tendency analysis, emotion multi-tendency classification of comments is achieved through three steps of extracting various morpheme emotions of the comments, analyzing similarity relationships among the morpheme emotions and calculating a morpheme emotion close path, various attitudes expressed by a user to things are more accurately distinguished, and the opinions of the users on object attributes and features are reflected.