34 results about "Morpheme" patented technology

The application relates to the technical field of text label marking, and provides a text specification marking method and system based on a rule corpus, which comprises the following steps: analyzing a policy and regulation document, identifying and marking condition morphemes and conclusion morphemes in the policy and regulation document, constructing a logical relationship between the two by using a large language model, and forming a rule corpus composed of structured morpheme pairs; performing semantic embedding on the corpus to generate a semantic vector library; performing multi-label coding on a verification data set based on the rule corpus, and constructing a multi-label training data set; training a deep learning classification model by taking semantic vectors as features and multi-labels as targets, so that a text specification marking model is obtained; and automatically marking target text by using the model. The application significantly improves the accuracy, interpretability and business adaptability of text marking, improves the update quality of a system label data set, and reduces the system maintenance cost.

An inference method and apparatus for a large language model, which method and apparatus are used for improving the inference calculation efficiency of a large language model. The method comprises: receiving an inference task, which carries a morpheme number, wherein the morpheme number is used for identifying a morpheme to be subjected to inference calculation, the morpheme comprising one or more sub-words determined on the basis of input text of a large language model (201); on the basis of the morpheme number corresponding to the inference task, querying a global index tree to determine a target inference node, wherein the global index tree comprises sub-trees corresponding to a plurality of inference nodes in a distributed inference cluster, and a sub-tree corresponding to the target inference node is a sub-tree in the global index tree, the matching degree of similarity of which sub-tree with the morpheme number is greater than a threshold, the matching degree of similarity indicating the number of pieces of reusable key-value cache data during inference calculation (202); and on the basis of the target inference node, executing the inference task (203).

The embodiment of the invention discloses an inference method and device for a large language model, which are used for improving the inference calculation efficiency of the large language model. The method comprises the steps that an inference task is received, the inference task carries morpheme numbers, the morpheme numbers are used for identifying morphemes to be subjected to inference calculation, and each morpheme comprises one or more sub-words determined based on an input text of a large language model; a global index tree is inquired based on morpheme numbers corresponding to the reasoning tasks, target reasoning nodes are determined, the global index tree comprises sub-trees corresponding to multiple reasoning nodes in a distributed reasoning cluster, and the sub-trees corresponding to the target reasoning nodes are sub-trees with the similarity matching degree with the morpheme numbers larger than a threshold value in the global index tree; the similarity matching degree indicates the number of reusable key value cache data in reasoning calculation. And executing the reasoning task based on the target reasoning node.

The invention relates to the technical field of text label marking, and provides a text standard marking method and system based on a rule corpus, and the method comprises the steps: analyzing a policy and regulation file, recognizing and marking a condition morpheme and a conclusion morpheme in the policy and regulation file, constructing a logic relation between the condition morpheme and the conclusion morpheme by using a large language model, and marking the rule corpus; forming a rule corpus consisting of structured morpheme pairs; performing semantic embedding on the corpus to generate a semantic vector library; performing multi-label coding on the verification data set based on the rule corpus, and constructing a multi-label training data set; training a deep learning classification model by taking the semantic vector as a feature and taking multiple tags as a target to obtain a text specification marking model; and automatically marking the target text by using the model. According to the method, the accuracy, the interpretability and the service adaptability of text marking are remarkably improved, the updating quality of the system label data set is improved, and the system maintenance cost is reduced.

Systems and methods evaluate the meanings of whole words bound within the meanings of their parts by analyzing and identifying features and components of words (such as letters and syllables), using sound information, for example phoneme class structure, to find sound correlation and using synonymity to find meaning correlation between words (or word parts) and a data set of known / identifiable sound / meaning associations called morphemes to find morpheme‑word‑parts that optimally describe an input word. Various means are described of compiling morphemes into a data set. The systems and methods further include analyzing the semantic functions of word parts and their modifying interrelationships, and using identifications made by one or more analyses. The systems and methods can evaluate or improve understanding of word meanings. The systems and methods are useful in researching sound / meaning associations in language generally and for improving language processing generally by the incorporation of more semantic information.

The invention relates to the technical field of computers, and discloses a multi-task processing-oriented Soc chip server computing power distribution method, which comprises the following steps of: acquiring execution primitive vectors of at least two tasks concurrently executed on an Soc chip; converting the execution primitive vector into an execution symbol sequence corresponding to each task; based on the execution symbol sequence, identifying an execution morpheme representing a task execution stage, and constructing a task grammar state machine, used for predicting a next morpheme, of each task; analyzing cross-task morpheme association between the execution morphemes of the concurrent tasks; and on the basis of the current morpheme of a certain task, the predicted next morpheme is associated with the cross-task morpheme, and a collaborative scheduling instruction is generated. By converting the execution primitive vector into the execution symbol sequence and identifying the execution morphemes based on the sequence to construct the task grammar state machine, the prediction of the next execution stage of the task is realized.

Provided are a method for segmenting sign language into morphemes, a method for predicting morpheme positions, and a method for augmenting data. A system for analyzing sign language according to an embodiment of the present invention comprises: a recognition unit which recognizes key points of a speaker from a sign language video; and a prediction unit which inputs the recognized key points into an artificial intelligence model, segments the sign language into morphemes, and predicts position information of the segmented morphemes. Accordingly, by recognizing the morphemes of the sign language video frame by frame on the basis of a skeletal model and thereby segmenting the sign language into morphemes and predicting morpheme positions, it is possible to lay the foundations for accurate sign language translation.

A method of processing a natural language may include generating a grammar used for analyzing a natural language by compressing a plurality of languages including common elements so that resources and time required for recognizing a grammar of an input sentence may be reduced. The method of processing a natural language may include separating an input sentence in units of morphemes, recognizing a grammar of the input sentence separated in units of morphemes, and analyzing a meaning of the input sentence based on the recognized grammar. The recognizing of the grammar may include recognizing a grammar that matches the input sentence separated in units of morpheme from among a plurality of pre-stored grammars, and the plurality of pre-stored grammars may include at least one grammar obtained by compressing two or more grammars to which at least one element among a plurality of elements constituting a single grammar is common.

The invention discloses a boundary perception and pre-training enhanced low-resource Tibetan language recognition method and system, and the method comprises the following steps: 1, carrying out the standardization processing of data and texts, and enabling the language knowledge and pre-training model enhanced low-resource Tibetan language voice to be converted into text. Unified modeling of Tibetan language boundaries, normal characters and morphemes is realized under a low-resource condition; remarkable and stable alignment is realized through a tsheg-perception adapter and three-head joint learning; language knowledge is fused into search in a learnable mode through the micro WFST soft constraint, and illegal font and language / case marking errors are greatly reduced. Training data are effectively expanded through teacher semi-supervised and coverage rate active learning, and cross-domain and accent robustness is improved; the end side maintenance cost is remarkably reduced through'adapter + atlas' hot updating; iCR and a boundary F1 are introduced as closed-loop indexes, so that the quality of an optimization target is consistent with that of a Tibetan Chinese character method, and the whole method has the advantages in engineering landing, interpretability and maintainability.

An information processing apparatus outputs answer information corresponding to inquiry information that is input. The information processing apparatus includes a memory and circuitry. The memory is configured to store a plurality of databases each having at least a first field and a second field. The circuitry is configured to: perform morphological analysis on the inquiry information, to divide the inquiry information into morphemes; perform a first matching process based on the morphemes and the first field of each of the plurality of databases, to determine whether to adopt the database as an extraction source from which the answer information is to be extracted; and perform a second matching process based on the morphemes and the first field of the database, which is determined to be adopted as the extraction source, to output, as the answer information, data in the second field corresponding to data in the first field.

The present application relates to a kind of general english tweet preprocessing method and computer equipment, belong to data processing technical field;Solved the problem that a large number of subjective words and non-standard morphemes exist in english tweet, affect tweet preprocessing result and named entity recognition performance;The method of the present application includes: based on multiple field english text, subjective word table is obtained by construction;The non-standard morpheme in the english tweet to be processed is carried out semantic reduction and information extraction, and the tweet text after preprocessing is obtained;Based on the tweet text after preprocessing, double stack structure is constructed to extract clause;Based on subjective word table, using syntax dependency analysis model and tree parent-child level structure, the tweet text after preprocessing is carried out named entity extraction, and the named entity recognition result of english tweet is obtained;The preprocessing result of english tweet is obtained by outputting the tweet text after preprocessing, the clause contained in english tweet and the named entity recognition result of english tweet.

Disclosed in the present application are an animation rendering method and apparatus, and an electronic device. The method comprises: acquiring text information and animation data that are required for animation rendering, wherein the text information is used for describing the animation data; determining speech data and morpheme data that correspond to the text information; converting the morpheme data into weight coefficients of corresponding control points in the animation data, wherein the control points are used for controlling the shape of a model in the animation data; and on the basis of the weight coefficients and the speech data, rendering the animation data.

The embodiment of the utility model provides a morpheme training device. The morpheme training device comprises a morpheme panel and a plurality of morpheme cards, a plurality of base areas are arranged on the surface of the morpheme panel, a hollow area is arranged in the middle of the morpheme card, and the morpheme card is used for being placed on the morpheme panel, so that the hollow area and the base areas form a composite morpheme specified by a task instruction. By adopting the scheme provided by the invention, the morpheme awareness of an individual can be enhanced and improved, so that the vocabulary growth of the individual is promoted, and subsequent Chinese character learning is facilitated.

A system, method, and computer program product for implementing automated digital text optimization is provided. The method includes receiving during execution of an Internet search process, digital textual content comprising embedded words, morphemes, and phrases. A specified format for generating a summary for the digital textual content is identified and probability distribution attributes are generated with respect to target vocabulary of the digital textual content. Content of the digital textual content is replaced with replacement content and an automated digital summary of the digital textual content is generated thereby enabling optimized operational functionality of a recurrent neural network (RNN) encoder-decoder hardware device. A decoded version of the automated digital summary is presented via a specialized graphical user interface.

This invention discloses a method and apparatus for rapid conversion of singing voice timbre based on zero-shot sampling. The method includes constructing a singing dataset containing dry vocals and lyrics; extracting the audio codebook index sequence of the dry vocals using a singing feature decoupler incorporating a Hubert model and residual quantization codebook; introducing a text encoder to extract morpheme features and morpheme index sequences from the lyrics; optimizing the singing feature decoupler through cross-prediction to improve the accuracy of speech content feature extraction; introducing pitch and timbre features to represent prosody; and enhancing the quality of synthesized vocals generated by the generator based on speech content features, pitch features, and timbre features through adversarial training. This achieves rapid conversion of song vocal timbre to the user's timbre while ensuring conversion quality.

Disclosed is a foreign language phrases learning system based on basic sentence pattern unit decomposition, and implemented in a computing device including at least one processor and at least one memory for storing instructions executable by the processor, which includes: a sentence decomposition unit, when a natural language composed of a foreign language is input from a user, for decomposing a compound sentence corresponding to the input natural language into a plurality of basic sentences; a sentence pattern determination unit for checking one of morphemes or words contained in each of the decomposed basic sentences when the compound sentence is completely decomposed by the sentence decomposition unit, thereby determining a sentence pattern for each of the basic sentences; an additional information designation unit, when the sentence pattern for each of the basic sentences is completely determined by the sentence pattern determination unit, for designating some of the morphemes or the words contained in each of the basic sentences as additional information; and an additional information storage unit for matching the additional information with one of the morphemes or the words, which are not designated as the additional information, when the designation of the additional information is completed, thereby storing the additional information in a basic morpheme category included in a pre-stored basic morpheme dictionary table. In addition, various embodiments identified through the present document are possible.

The application discloses a morphological enhancement based tensorized word embedding compression system, which comprises a morpheme segmentation module, a morpheme index and embedding module and a word embedding generation module; the morpheme segmentation module divides each word in a word table of a text task into morphemes; the morpheme index and embedding module firstly generates a morpheme table by counting the division result of the morpheme segmentation module, then defines a morpheme index matrix and a plurality of trainable morpheme embedding matrices, each row of the morpheme index matrix represents the position of the morpheme of a corresponding word in the word table in the morpheme table, and each row of the morpheme embedding matrix represents an embedding vector of a corresponding morpheme in the morpheme table; the word embedding generation module indexes the morpheme vector from the morpheme embedding matrix and performs a tensor product on each word in the word table, adds the results of a plurality of tensor products to generate a word embedding vector; and the application overcomes the problems of a large amount of parameters and storage space occupation in general word embedding technology, and the problem of task effect loss when high compression word embedding is performed.

An information processing apparatus outputs answer information corresponding to inquiry information that is input. The information processing apparatus includes a memory and circuitry. The memory is configured to store a plurality of databases each having at least a first field and a second field. The circuitry is configured to: perform morphological analysis on the inquiry information, to divide the inquiry information into morphemes; perform a first matching process based on the morphemes and the first field of each of the plurality of databases, to determine whether to adopt the database as an extraction source from which the answer information is to be extracted; and perform a second matching process based on the morphemes and the first field of the database, which is determined to be adopted as the extraction source, to output, as the answer information, data in the second field corresponding to data in the first field.

To provide a Japanese evaluation system, a Japanese evaluation device, a Japanese evaluation method, and a program capable of efficiently creating a clear Japanese sentence suitable for foreign language translation.SOLUTION: A Japanese evaluation system 100 includes a morphological analysis unit 212 that divides an acquired Japanese target sentence into morphemes by morphological analysis, a measurement unit 213 that measures the number of times of detection of a predetermined particle in the morphemes of the target sentence, and a setting unit 214 that, when the predetermined morpheme is detected from the morphemes of the target sentence, divides the number of times of detection before and after the predetermined morpheme and sets the divided number of times of detection as the number of times of detection related to the target sentence.SELECTED DRAWING: Figure 4

The invention discloses a sign language translation system and method based on time domain and representation alignment. The method comprises the following steps: extracting sign language video visual features through a visual encoder; generating an alignment path of the visual feature sequence and the sign language morpheme sequence based on a discriminant dynamic time warping algorithm; training a sign language fragment segmentation model by using the aligned path; the sign language fragment features are aggregated into time domain alignment visual representation by using an adaptive time sequence fusion module; performing vision-semantic representation alignment training according to semantic embedding of morphemes; in the sign language translation step, visual features of a video to be translated are extracted through a visual encoder, and morpheme fragments are generated by using a sign language fragment segmentation model; a morpheme-level feature is obtained through the self-adaptive time sequence fusion module and then input into a translation decoder, and a corresponding natural language text is output; according to the method, the problems of speed difference and visual-semantic modal difference of sign languages can be solved at the same time, and translation errors caused by similar actions between sign language primitives are reduced, so that the translation accuracy is improved.

The invention discloses a Mongolian multi-modal sentiment analysis method based on cross-modal information enhancement and fusion. In order to solve the problems that emotion expression is dispersed and implicit clues in multi-modal emotion information are difficult to capture due to Mongolian adhesive language characteristics, a three-level innovation scheme is provided: (1) a cross-modal explicit enhancement layer: converting the implicit emotion clues such as rhythm, expression and the like in audios and videos into emotion description texts by using a multi-modal large language model; the modal information is enriched through a path of implicit, explicit and semantic fusion; (2) a semantic alignment layer based on a Gram matrix: extracting a second-order statistical structure of a text modal as a semantic reference, mapping heterogeneous audio and video features to a unified semantic space, and solving cross-modal alignment difficulty caused by Mongolian morphological change; and (3) a gating displacement fusion layer: a self-adaptive displacement mechanism is adopted, accurate injection of non-text information is realized through fine tuning of semantic positions of text features, and the problem of modal submerging in a Mongolian long morpheme sequence is avoided.

An electronic device for evaluating the specificity of an answer in an employment document according to the present disclosure may comprise: a memory for storing at least one instruction; and at least one processor for executing the at least one instruction. The at least one processor may extract answer data from employment document data, analyze the answer data on the basis of at least four evaluation criteria, and calculate a specificity evaluation score of the answer data. The at least four evaluation criteria may include at least one of sentence length evaluation, keyword evaluation, morpheme evaluation, and personal experience evaluation.

The invention discloses a linguistic prior and progressive cross-modal contrast disambiguation network-based lip language recognition method, which is suitable for the field of computer vision speech recognition and comprises the following steps of: 1, on the basis of a linguistic vision-phoneme mapping rule, constructing a fuzzy text and realizing replacement by replacing similar phonemes in a pronunciation text; and 2, constructing a progressive comparison disambiguation network which is composed of a vision-phoneme disambiguation sub-module and a text semantic disambiguation sub-module, mapping in coding through vision-phoneme feature alignment and text semantic alignment in a mode, and realizing end-to-end ambiguity analysis in decoding. According to the method, through fusion of linguistics rules and a deep learning technology, the whole process optimization of fuzzy sample construction-progressive disambiguation-targeted evaluation is realized, so that the decoupling capability of the model for visual morpheme-phoneme asymmetric ambiguity can be improved, and the lip language recognition precision and accuracy of the fuzzy sample are improved.

The invention relates to the technical field of text label application management, and provides a coding and application method and system constructed based on a rule file and a corpus, and the method comprises the steps: carrying out the file coding of the rule file; performing chapter and section term coding on chapter and section terms in the rule file; constructing a corpus unit; file codes, chapter and clause codes and corpus units are associated and combined to form a traceable code combination, the traceable code combination is applied to standardized marking of business file key information, standard reference traceability is formed through coding, and a business corpus can be formed by combining business practice and generalization mapping to improve the application effect. According to the method, reverse tracing of original rule files, clause positions and morpheme semantics is realized, and forward retrieval can be carried out by utilizing texts which are matched with conditions through codes. According to the method, semantic disambiguation and accurate traceability of the rule text are realized through a multi-level coding system, meanwhile, the generalization understanding capability of an AI technology is compatible, and the accuracy of text marking is improved.

The invention discloses a data processing method and device, a storage medium and electronic equipment. The method comprises the steps of obtaining a target statement group for executing data processing on a target database, wherein the target statement group comprises a plurality of statements to be executed; identifying a target statement carrying a combined morpheme from the target statement group, wherein the combined morpheme is used for indicating that the target statement is a composite instruction statement obtained by combining at least two instruction statements; determining target data to be processed by the target statement from a target database; processing operation corresponding to each instruction statement contained in the target statement is executed on the target data, and a composite processing result corresponding to the target statement is obtained; under the condition that statement processing results corresponding to all the statements in the target statement group are obtained, statement processing results are returned, and the statement processing results comprise the composite processing result. According to the method and the device, the technical problem of relatively low data processing efficiency in a data processing mode provided in related technologies is solved.

A Mongolian multi-modal sentiment analysis method based on cross-modal information enhancement and fusion. To solve the problems of scattered sentiment expression caused by the agglutinative characteristics of Mongolian and the difficulty of capturing implicit clues in multi-modal sentiment information, a three-level innovative solution is proposed: (1) Cross-modal explicit enhancement layer: Use multi-modal large language models to convert implicit sentiment clues such as prosody and expression in audio and video into text descriptions of sentiment, enriching the information of each modality through the "implicit→explicit→semantic fusion" path; (2) Semantic alignment layer based on Gram matrix: Extract the second-order statistical structure of the text modality as the semantic reference, and map heterogeneous audio and video features to a unified semantic space to solve the cross-modal alignment difficulty caused by the morphological changes of Mongolian; (3) Gating displacement fusion layer: Use an adaptive displacement mechanism to fine-tune the semantic position of text features to achieve precise injection of non-text information, avoiding the modality submersion problem in long morpheme sequences of Mongolian.

This invention belongs to the field of big data information technology and discloses an information processing method and system based on big data: The method involves segmenting the sentences of the text to be detected to obtain a second set of sentences; filtering the second set of sentences using a sensitive word library to obtain a first set of candidate sensitive sentences and a third set of sentences; calculating the sentence similarity between sentences in the first set of candidate sensitive sentences and sentences in the sensitive word library, with sentences having a maximum similarity greater than or equal to a first threshold being considered sensitive sentences; recombining the morphemes of sentences in the third set of sentences; filtering the recombined sentences using a sensitive word library to obtain candidate sensitive sentences; and calculating the sentence similarity between candidate sensitive sentences and sentences in the sensitive word library, with the maximum similarity greater than or equal to a first threshold being considered sensitive sentences. 1 The statement is identified as a sensitive statement at that time; the maximum similarity is less than TH. 1 But greater than or equal to TH 2 In such cases, the information is then reviewed manually. This invention improves the detection rate and accuracy of sensitive information.

An electronic apparatus for comparing a similarity between texts on the basis of sub-sentence extraction, according to the present disclosure, may comprise: a memory storing at least one instruction; and at least one processor that executes the at least one instruction. The at least one processor may: perform analysis of dependence between a plurality of morphemes included in at least two pieces of text data; generate a dependent relationship graph for the plurality of morphemes, on the basis of the dependence analysis; extract at least one sub-sentence by using the dependent relationship graph; and calculate a similarity between the at least two pieces of text data by cross-comparing the at least one sub-sentence.

The application discloses a sign language generation method based on a semantic guidance diffusion model, which comprises the following steps: inputting a sign language morpheme sequence and a noise posture sequence into a text encoder and a visual encoder respectively, extracting text features and posture features, inputting the posture features and the text features into a local enhancement module, each local enhancement layer of the local enhancement module being stacked and comprising adaptive sign language graph convolution and time convolution, the posture features being enhanced under the guidance of the text features, finally obtaining local enhanced posture features, converting the local enhanced posture features through a feature converter to obtain converted features, then passing the converted features through a global enhancement module to obtain global enhanced posture features, and inputting the global enhanced posture features into a Fiergi to obtain a predicted sign language posture sequence. The application takes into account the delicacy of local joint generation and the natural coherence of the overall sequence, and a semantic consistency guidance mechanism can effectively ensure that the generated posture sequence is highly consistent with the corresponding sign language morpheme sequence at the semantic level.