39 results about "Sentence processing" patented technology

The invention provides a criminal identification and forecast method. The method adopts a data pre-processing method in data mining; aiming at criminal information such as data, street address, criminal police zone, week, criminal type, criminal description and sentence processing, attribute reconstruction, feature extraction and feature selection are performed, the correlation between the criminal information is mined, a characteristic factor with maximum difference is generated, and the correlation between the characteristics factor and a criminal result, namely the criminal type is generated; and then a model integrating Gaussian Naive Bayes, a neural network, Logistic regression, regularized regression, K neighbor, random forest, a support vector machine and an XGBoost learning algorithm is built to obtain an element classifier based on a weighted voting classifier having highlight classification and favorable clustering effect, reconstructed data is analyzed, processed and identified, a criminal condition of a city in future is forecasted, an individual criminal map of the city is drawn, and the effects of promoting and regulating city public security and management are further achieved.

A software system that automates the use of computing systems by storing English language sentences using synonyms and symbol words to allow user access to computer resources without having to know the technical details or the specifics associated with a body of knowledge. Synonym sentences allow users to get to computer resources using different words stored in different sentences that open the same computer resource (i.e.: a document) to significantly improve a first time request for information. Sentences learned (stored) in text or SQL databases or used in a software agent sentence or search engine sentence are automatically parsed and directed to the appropriate knowledge repository to automated computer processes or to get user information without having to know any technical details required to navigate through computer systems or computer data. Multiple words or sentences can be linked together with computer resources to further automate job tasks.

The invention provides a query sentence processing device and a query sentence processing method. The query sentence processing device comprises a rule definition module 102, a syntactic analysis module 104 and a rule processing module 106, wherein the rule definition module 102 is used for setting a rule for processing self-defined elements; the syntactic analysis module 104 is used for analyzing an expression to acquire the self-defined elements in the module 104; and the rule processing module 106 is used for processing the self-defined elements according to the rule and forming query sentences by using a processing result of the elements. By a preset rule, the elements of the expression are converted into an identifiable form of the current database, so that the expression can be applied to any database.

A method and apparatus for translate terms base on context are provided in embodiments of that present invention, The method comprises the following steps of: performing sentence processing on a fileto be translated; extracting a term in the file to be translated and a sentence in which the term is located by using a maximum forward matching algorithm combined with a term library selected in advance by an interpreter; and obtaining a term definition corresponding to the term from the term library; and performing sentence processing on the file to be translated. Extracting a corpus whose sentence matching degree with the term is greater than a preset threshold value, sorting the corpus according to the matching degree from high to low, and filtering out the corpus not containing the term;Obtaining a corpus interpretation corresponding to the term in the corpus by a word alignment method; The corpus definition is combined with the string editing distance algorithm to filter the term definition, and the final term definition is obtained. The embodiment of the invention extracts the definition of the terms in the corpus through the word alignment method, improves the term prompting function of the computer-aided translation, and can effectively improve the translation efficiency of the interpreter.

A neural network system includes one or more computers including a memory to store a set of documents having textual elements; and a processor to partition the set of documents into sentences and paragraphs; create a segment vector space model representative of the sentences and paragraphs; identify textual classifiers from the segment vector space models; and utilize the textual classifiers for natural language processing of the set of documents. The processor may partition the set of documents into words and sentences. The processor may create the segment vector space model representative of sentences, paragraphs, and words, and documents.

The invention relates to a method of processing minority characters into Chinese language, and in particular relates to a united labeling method for syntax of Tibet language and semantic roles. The united labeling method comprises the following steps of: a) distinguishing a simple sentence and a compound sentence; b) labeling semantic roles; c) recognizing a predicate; d) classifying verb semantics; e) labeling a syntactic structure; f) editing and revising semantic role labeling results. According to the united labeling method, the syntax of Tibet language and semantic features are extracted, on the one hand, semantic role information such as a performer, a receiver, time, a place and a way expressed in the sentence can be labeled by directly utilizing grammatical labels of the Tibet language; on the other hand, a syntax analytical process can be reacted upon by the predicate semantic role labeling result so that the influence of the syntax labeling which is not well-determined can be reduced, and accordingly the performance of a sentence processing system can be improved.

The application relates to a SQL sentence processing method, a SQL sentence processing device, a computer device and a storage medium. The method includes steps of acquiring a database sub-parameter of every initial SQL sentence; calculating according to the database sub-parameter to obtain performance consumption ratio of every initial SQL sentence; extracting an initial SQL sentence of which performance consumption ratio is more than the preset threshold value, and forming a core SQL sentence set; placing the core SQL sentence set in an old version environment and a new version environment to carry out performance analysis, and acquiring a low-performance SQL sentence, wherein the low-performance SQL sentence is the initial SQL sentence of which performance in the new version environmentis lower than that in the old version environment, and which is located in the core SQL sentence set; promoting performance of the low-performance SQL sentence. By adopting the method, the time for analyzing and positioning the SQL sentence with reduced performance in the new version environment is saved, thus the version upgrading cycle of a relationship database system is shortened.

The invention relates to an SQL sentence processing method and system. The SQL sentence processing method comprises following steps: obtaining original SQL sentences; filtering redundant fields of SQL sentences in order to obtain first SQL sentences of redundant fields; replacing variable value of the first SQL sentences in order to obtain second SQL sentences; performing structured processing of the second SQL sentences in order to obtain structured SQL sentences; and relating structured SQL sentences in order to obtain SQL sentence prototypes corresponding to original SQL sentences. The above-mentioned SQL sentence processing method and system can bring convenience to users for search and reduce data amount of auditing.

The invention provides a SQL (structured query language) sentence processing method and device. The method includes following steps: distributing SQL sentences into corresponding queues according to acquired SQL sentences whose execution time is longer than or equal to a first preset value within a preset time period and execution time number of the SQL sentences; acquiring average execution time of the SQL sentences in each queue; processing selected SQL sentences whose average execution time is longer than or equal to a second preset value and/or the execution time number is larger than or equal to a third preset value. In the technical scheme, the SQL sentences with large execution time number and serious time consumption are optimized after the SQL sentences with the average execution time longer than or equal to the second preset value and/or the execution time number is larger than or equal to the third preset value are found out, so that system running efficiency is greatly improved, and user experience is improved.

The invention relates to a sentence expression method based on a Chinese sentence meaning structural model and a topic model, and belongs to the technical field of computer science and natural language processing Chinese analysis. The method includes the steps that sentence meaning structural analysis is conducted on a sentence to obtain a sentence meaning structure of the sentence; elementary item words and general item words in the sentence are extracted, and the topic model is used for analysis to obtain an elementary item knowledge base and a general item knowledge base; according to words under a topic and a theme in the sentence meaning structure, the knowledge bases obtained in the previous step are used for expanding sentence content to obtain a sentence expression result. A new thought train is provided for solving the feature sparse problem of sentence expression, the sentence classifying effect is effectively promoted, and the sentence expression method has high theoretical value and an important practice function.

A system for contextualising an unstructured stream of text, representative of spoken word, including a grammar processor, a sentence processor, a frequency processor, a summer and an emotion processor. The unstructured stream of text is processed and outputs an audio file total for each matched phrase, word and proper noun, determined from the unstructured text, to a data significance processor. The data significance processor receives and audio file total for each name, proper noun and matched real phrase, determined from the unstructured text, and outputs a list including the names, proper nouns and matched real phrases in order of contextual significance.

The invention discloses a document writing assistant implementation method and system and electronic equipment, and the method comprises the steps: inputting a search item which should be contained ininformation to be searched in a document editing interface, wherein the search item at least comprises a keyword or a word or a sentence; converting the search items into word vectors, searching sentence vectors matched with the word vectors for from a pre-established database, wherein each sentence vector is arranged in an independent data unit of the database, and the data unit at least comprises sentence text information, sentence vectors, sentence outcomes and reference information carried by sentences; and in the document editing interface, returning corresponding sentence text information, sentence vectors, sentence expressions and reference information carried by sentences in the data unit for editors to select. Through the word vector model, sentences and words are converted intoreal number vectors to be stored and matched. Compared with the prior art, through dictionary or regularization expression matching, the retrieval result is more accurate.

The invention discloses a simplified and traditional Chinese input method with no need of remembering codes, which is characterized in that during input of individual characters, a Chinese character is split into 280 base roots corresponding to 27 code elements respectively and naturally inputted according the sequence of from left to right, from top to bottom, from inside to outside, from middle to lateral, from horizontal stokes to vertical stokes and from throw-away stokes to press-down stokes. Traditional Chinese can be directly inputted by converting fixed base roots of tails of the simplified Chinese into universal base roots, and inputting without remembering codes is realized. Sounds, forms and meanings of the 280 character-forming base roots are dependent on 'sounds, forms and meanings' of letters on a keyboard for association reasoning, unique positioning is realized, remembering base roots and keys of the keyboard is not needed, and rote learning and practicing of codes and keys of the keyboard can be avoided by inputting conventionally. Existing complex and fragmentary basic stoke splitting codes are avoided while integral character and sentence processing is realized, integral characters are used as fixed codes during inputting of phrases and sentences, and intermediate links such as splitting thinking and the like are omitted.

The invention provides electronic equipment and a sentence processing method thereof, which can generate a law for acquiring an attribute of an input sentence only aiming at key words having great contribution degree for obtaining the attribute of the input sentence. The sentence processing method comprises the following steps: storing N training sentences, wherein attributes of a part of phrases in each of the N training sentences are marked; dividing the N training sentences into M training sentence sets according to the attributes marked for the part of phrases; aiming at each word obtained through word segmentation processing for the training sentences, judging whether a specific word whose relation between statistical results of the same word in each training sentence set satisfies a specified condition exists, and determining the specific word as the key word; for the determined key word, generating the rule based on the part of phrases in each training sentence and the attributes marked for the part of phrases. And thus the processing efficiency can be improved and the calculation resource can be saved on the premise of not affecting the accuracy degree of the obtained attribute.

The invention discloses a character-learning method. Characters are classified in terms of properties and respective categories to produce character cards group by group; the characters, which can be explained with pictures, have the corresponding pictures. The learning courses comprise single-character reasoning, multi-character reasoning and multi-sentence reasoning; wherein the single-character reasoning comprises the steps that the teacher matches partial character cards in the same group with the pictures at first, with leaving at least one character card each time, therefore, the child matches the remaining character cards with the remaining pictures in accordance with the reasoning. The multi-character reasoning is to learn the characters in a short sentence after the child can read a certain quantity of single characters, and the child can reason the remaining characters in accordance with language habit only if the other characters in the short sentence are readable for the child. The multi-sentence reasoning comprises the steps of teaching the child to read a long sentence or a short sentence, telling the child another sentence that has synonymous, similar or antonymous meaning with respect to the former sentence, and causing the child to find out new characters corresponding to the meaning of the sentence. The invention has the characteristics of enlightening the thought of the child and training the logic reasoning ability of the child, and the efficiency thereof is remarkably superior to the efficiency of the traditional spoon-feed learning.

The present invention discloses a threshold value classification SLM method and system for reduction of an OFDM (orthogonal frequency division multiplexing) system peak-to-average power ratio (PAPR).The method comprises the steps of: generating a phase factor rotation sequence by employing two sets of different phase factor sets, performing weighted processing for a binary sequence after mappingand then performing inverse fast Fourier transform to generate two sets of original alternate signal sequences, performing sentence processing of the two sets of the original alternate signal sequences, adding the signal sequence with PAPR values in the two sets being larger than a threshold value and the signal sequence with the PAPR values in different sets being larger than the threshold valueand performing normalization processing to generate a new set of alternate signal sequence; and finally, selecting an alternate signal sequence with the minimum PAPR from the three sets of the alternate signal sequences for transmission. The threshold value classification SLM method and system for reduction of the OFDM system peak-to-average power ratio can obtain the PAPR performance similar to the original SLM method and can greatly reduce the calculation complexity so as to improve the performances of the OFDM system.

The invention provides a clustering analysis method for commodity evaluation. Specifically, the method comprises the following steps: extracting specific numerals whose semantics are closest to each other in commodity evaluation corpus according to a point of interest proposed by a user; carrying on the sentence processing to all evaluations of the commodity, and selecting the sentences which contain the key words close to the semantics of the point of interest in these sentences; classifying these sentences emotionally and then clustering the selected sentences densely makes it easier for potential buyers to understand the particular functionality of the product from a finer-grained perspective. Because this method does not define the possible concerns of the data users in advance, the query of the users is no longer limited, and the requirements of the users are better satisfied.

The invention discloses a sentence processing method, which comprises the following steps of when an original sentence is received, performing a feature extraction operation on the original sentence to obtain a first feature, and inputting the first feature to a conditional random field model; determining a candidate antecedent set, and performing the feature extraction operation on the candidateantecedent set and the original statement to obtain a second feature, and inputting the second feature to a preset classifier; according to the output result of the conditional random field model, determining the target missing bits of the original statement, and according to the output result of the preset classifier, determining the target antecedent, and filling the target antecedent in the target missing bits of the original statement. The invention also discloses a sentence processing device and a computer readable storage medium. The invention can effectively improve the accuracy and flexibility of missing bits and antecedent words.

The invention discloses a method and system for evaluating the readability of a news reading text. The method includes: performing sentence processing on the news reading text to be evaluated to form a sentence list; performing syntactic analysis on the sentences in the sentence list to obtain each The syntactic structure information of a sentence; the sentence matching number of each sentence is calculated based on the syntactic structure information of each sentence; the sentence matching number fractal dimension of the news reading text to be evaluated is calculated based on the sentence matching number of each sentence, and calculated The ratio of the sentence number fractal dimension to the length of the above sentence list is to obtain the modified sentence number fractal dimension; based on the modified sentence number fractal dimension, the readability of the news reading text to be evaluated is evaluated by fusion method . The invention integrates the modified sentence number fractal dimension with the existing readability formula, and improves the classification accuracy and F1 value of the original readability formula.

The invention discloses a dynamic query sentence processing method and device. The method includes the steps that a configuration file is set, wherein the configuration file is used for defining mark key words needed for query sentences and should comprise at least one configuration item, and each configuration item comprises a mark key word name and an expression of the mark key word name; after receiving the query sentences and query parameter sets transmitted by clients, the mark key words are searched for according to marks in the query sentences, and the corresponding mark key word names and the query expressions of the mark key word names are found from the configuration file; if the marks corresponding to the mark key words exist in the query parameter sets, the original query sentences are replaced according to the determined mark key word names and the query expressions of the mark key word names, the matched marks in the original query sentences are converted into query expressions, and if the marks corresponding to the mark key words do not exist in the query parameter sets, the marks in the original query sentences are removed; through the mode, new query sentences are obtained.

The invention discloses a consumer policy awareness mining method and system based on big data text mining. The method comprises the steps that policy text data of a specific field and consumer comment text data are collected on related government websites and social platforms; preprocessing the collected text data, wherein the preprocessing comprises duplicate removal, noise reduction and short sentence processing; performing deep mining on professional vocabularies based on machine recognition and an expert knowledge corpus, and constructing a policy corpus; and carrying out policy awareness mining analysis on the consumers based on the consumer comments and in combination with the policy corpus. According to the method, network public opinion information is fully utilized, consumer comment texts are mined from the perspective of consumer policy awareness, and an expert knowledge corpus for consumer policy awareness mining is established, so that the text word segmentation effect is improved to a greater extent, and the text word segmentation efficiency is improved by combining natural language processing and other technologies. The policy perception state of the consumer is more systematically and accurately mined and analyzed, and suggestions are provided for policy supply optimization and policy accuracy and benefit improvement.

The invention discloses a formalized encoding method for an action description natural text, which comprises the following steps of: (1) reading the action description natural text, dividing the action description natural text into sentences, and storing the sentences into a sentence set L; (2) taking any sentence li from L, performing word segmentation to obtain a word set W, and performing part-of-speech tagging to obtain a part-of-speech set P; (3) based on W and P, carrying out dependency syntactic analysis to obtain a dependency relationship set Rel; (4) based on Rel, extracting main components of the sentence li, and storing the main components into a component triple set E; (5) traversing E, and if part description or transformation description of the component triad does not have amatching result in the dictionary, changing the description into closest description of the matching result in the dictionary; (6) based on a dictionary matching method, carrying out formalized coding on each component triple in the E, and storing the formalized codes into Mi; and (7) finishing processing all sentences in the L to obtain a formalized code set F of the action description natural text. According to the method, formalized coding of the action description natural text can be efficiently and accurately realized.

The invention provides an attribute-level sentiment analysis method and system based on a three-dimensional target matrix, and belongs to the technical field of data processing, and the method comprises the steps: processing each comment sentence in a training set into a required input format; on the basis of the format, constructing a three-dimensional target matrix capable of representing all attributes of the corresponding comment sentences at a time; inputting the processed training set into a model for supervised learning training, obtaining a first matrix vector representation at the last layer of the model, inputting the first matrix vector representation into a pooling collection layer to obtain a second matrix vector representation, and obtaining a final third matrix vector representation through a linear transformation layer and activation operation; every time the model is trained, the trained model is verified on the verification set, and an optimal model is obtained; inputting comment sentences to be analyzed into the optimal model, and decoding the sentiment polarity representing each attribute.

The invention discloses a sentence-level parallel corpus making method and system realized based on a man-machine collaboration concept, and relates to the technical field of sentence-level parallel corpus making. The method comprises the following steps: loading a Chinese and English contrast article to a source data textbox; respectively scanning and identifying Chinese and English sentences in the source data textbox; automatically copying the currently recognized Chinese and English sentences to a Chinese and English sentence textbox; checking the current Chinese and English sentence in the Chinese and English sentence textbox manually to obtain a manual judgment result, and inputting the manual judgment result into the man-machine cooperation system; and entering next sentence processing. Through the man-machine cooperation system, human initiative and computer regularity can be integrated. Finally, perfect fusion of speed and accuracy is achieved.

The invention discloses a word and sentence processing method and device and a computer storage medium, and the processing method comprises the steps: obtaining a to-be-processed stroke sequence of ato-be-processed word and sentence, and inputting the to-be-processed stroke sequence into a trained language model; generating a representation vector of the current to-be-processed stroke by using the language model and taking the forward to-be-processed stroke and/or the backward to-be-processed stroke of each current to-be-processed stroke as context information; and determining a representation vector of each character according to the representation vector of the to-be-processed stroke. Through the mode, the representation vector of the character is obtained according to the representation vector of the stroke to be processed of the character, so that the representation information of the character can be determined by utilizing the semantic information in the character.