12 results about "Reserved word" patented technology

The invention belongs to the technical field of data acquisition and transmission, and particularly relates to an intelligent acquisition method and system for clinical test data, and the method comprises the steps: carrying out the text coding of a hexadecimal format on the acquired clinical test data, carrying out the statistics of the initial frequency of each basic character based on the obtained hexadecimal text, and carrying out the calculation of the initial frequency of each basic character based on an inter-class maximum variance method; the basic characters are divided into reserved characters and replacement characters, the run length of each reserved character in the hexadecimal text is obtained, a run length set is formed and used for setting a plurality of replacement schemes for each replacement character, and data belonging to the replacement characters in the hexadecimal text are subjected to replacement operation according to the sequence, the replaced text is obtained, the replaced text is scrambled, and an encryption result of the clinical test data is obtained and transmitted. According to the invention, the security and confidentiality of the collected clinical test data in the transmission process are improved.

The invention provides a large model code member inference method based on code grammar constraint lexical element filtering, which comprises the following steps that: on the basis of a grammar specification of a target programming language, a grammar agreement set is constructed, and each grammar agreement comprises a antecedent condition and a corresponding subsequent lexical element; inputting a to-be-detected source code into the large language model, and obtaining a lexical element sequence output by the model and a prediction probability of each lexical element; performing grammar structure analysis on the lexical element sequence, identifying and rejecting grammar agreement lexical elements which are only determined by grammar rules and are irrelevant to author personality based on the grammar agreement set, and obtaining a reserved lexical element set; calculating member scores based on the prediction probability of each lexical element in the reserved lexical element set; and judging whether the to-be-detected source code belongs to a pre-training data member of a large language model or not according to a comparison result of the member score and a preset threshold value. According to the method, the detection performance of the pre-training code membership of the large language model is remarkably improved.

This invention belongs to the field of data acquisition and transmission technology, specifically relating to an intelligent acquisition method and system for clinical trial data. The method includes: encoding the acquired clinical trial data into hexadecimal text; based on the obtained hexadecimal text, calculating the initial frequency of each basic character using the inter-class maximum variance method; dividing the basic characters into retained characters and replacement characters; obtaining the run length of each retained character in the hexadecimal text and forming a run length set for setting multiple replacement schemes for each replacement character; sequentially performing replacement operations on each data in the hexadecimal text belonging to the replacement character to obtain the replaced text; scrambling the replaced text to obtain the encrypted result of the clinical trial data and transmitting it. This invention improves the security and confidentiality of the acquired clinical trial data during transmission.

An example couples a non-standardized tokenizer to an input of a neural network model with attention. The non-standardized tokenizer is to use a non-standardized vocabulary to convert reserved words of a first sequence of actions logged via use of a device by a first entity to first tokens including word-based tokens that describe actions in the first sequence of actions. The example omits a position encoder of the neural network model with attention. The neural network model with attention is trained, with the omitted position encoder, using a training input including the first tokens generated by the non-standardized tokenizer.

The application discloses a naming method, system, device and medium for modules in a distributed operating system, and can realize the following technical effects: (1) a standardized input is provided for dynamically generating a URL and a two-dimensional code through a structured naming rule; (2) the functional properties, action ranges or system levels of the modules can be more intuitively reflected; (3) a clear and manageable path is provided for the fine classification and expansion of the system under the premise of maintaining the stability of the core architecture through the introduction of a predefined and protected secondary classification mechanism; (4) the global uniqueness of the module naming is ensured in combination with user information, and the naming conflict between the system key modules and the user modules is avoided through the reserved word mechanism of the secondary classification; and (5) the readability, recognizability and processing efficiency of the module names in the user interface, command line operation and network transmission are improved through the standardized naming structure and length limitation.

A device, method, and non-transitory computer readable medium that for analyzing computer source code to detect code smells is disclosed. The method includes inputting, via processing circuitry, the source code and creating, via the processing circuitry, pseudo labels by a proxy task based on a vector of tokens for the source code. In addition, the method includes training, via the processing circuitry, a transformer model on the pseudo labels, as a pre-trained model that outputs a prediction of a value of tokens in the vector of tokens, and applying, via the processing circuitry, the pre-trained model to a plurality of fine-tuning models for respective downstream tasks, where each fine-tuning model is created by training the pre-trained model. The method also includes outputting, via the processing circuitry, from each fine-tuning model, an indication of whether a code smell has been detected in the source code.

A method for controlling operation of a data storage device, the data storage device, and a controller thereof are disclosed. The method can include selecting a block of a plurality of blocks of a non-volatile memory component of a plurality of non-volatile memory components, receiving a data write instruction from a host, generating a plurality of operation instructions corresponding to the data write instruction, and performing data write in a plurality of non-reserved word lines of the block, wherein the block includes the plurality of non-reserved word lines and a plurality of reserved word lines, and each of the plurality of non-reserved word lines includes a plurality of pages, and writing user data to a reserved word line of the plurality of reserved word lines by a single level cell write mode, such that the reserved word line includes a single page.

An example couples a non-standardized tokenizer to an input of a neural network model with attention. The non-standardized tokenizer is to use a non-standardized vocabulary to convert reserved words of a first sequence of actions logged via use of a device by a first entity to first tokens including word-based tokens that describe actions in the first sequence of actions. The example omits a position encoder of the neural network model with attention. The neural network model with attention is trained, with the omitted position encoder, using a training input including the first tokens generated by the non-standardized tokenizer.

The application provides a data processing method and device. The data processing method comprises the following steps: determining a text to be searched, and acquiring a filter field set and a reserved field set; determining an i-th word field and an i-th word segmentation field corresponding to the i-th word field in the text to be searched based on the filter field set; generating a target word unit according to the i-th word field and the i-th word segmentation field in the case that the i-th word segmentation field belongs to the reserved field set; generating a target word unit according to the i-th word field in the case that the i-th word segmentation field does not belong to the reserved field set; i is sequentially increased, and the step of determining the i-th word field and the i-th word segmentation field corresponding to the i-th word field in the text to be searched based on the filter field set is executed until i is increased to k, and an index text of the text to be searched is created according to at least one generated target word unit, wherein i is a positive integer starting from 1 and ending at k, and k is determined according to the text length of the text to be searched.

This invention discloses a text classification method and system based on a dynamic graph attention network, belonging to the field of natural language processing technology. The method includes: character-level word segmentation of the text and construction of an encoded vocabulary; construction of a global position graph using the original text structure, establishing three types of edge connections: co-occurrence edges, same-word edges, and self-loop edges; fusing position encoders to all nodes to generate word embedding vectors; constructing a dynamic graph attention mechanism, using multi-head attention computation to learn multi-level semantic associations in parallel, combining a residual module to alleviate the gradient vanishing problem in graph networks, and employing a hybrid strategy of weighted global max pooling and global average pooling to obtain graph-level representations; and outputting the classification results. This invention effectively solves the technical problem of traditional graph construction methods ignoring the semantic differences of the same word in different positions by preserving character position information and word order relationships, thus improving the accuracy and generalization ability of text classification.

This invention provides a pre-training method and related methods and devices. The pre-training method includes: acquiring a sequence of character information and a sequence of phoneme information corresponding to a training text, and alignment information between the character information sequence and the phoneme information sequence at the whole-word level; combining the alignment information, performing a mixed information sequence at the whole-word level to obtain a mixed information sequence, wherein, during the mixed processing, only one type of information, character information or phoneme information, is retained for the same whole word; and training an initial language model based on the mixed information sequence. Because this invention pre-trains the language model based on a mixed information sequence containing both character information and phoneme information, the language model can learn both pronunciation information and semantic information through training, resulting in a language model with better representational capabilities.

The application relates to a dialogue text normative analysis method and device and a storage medium, and is applied to the technical field of text analysis, and comprises the following steps: storing dialogue texts in an Elasticsearch cluster according to speaking roles, the Elasticsearch cluster is based on Lunce, supports self-defined word segmentation and query plug-ins, on the basis, the dialogue texts are stored according to the roles, the problem that a word segmentation plug-in cannot distinguish and store role texts based on the full-text retrieval technology Lunce in the prior art is solved, and the application sets self-defined system reserved words, sets a query syntax according to the self-defined system reserved words, generates a syntax tree through Antlr4 technology, obtains a query statement that can be recognized by the Elasticsearch cluster through the syntax tree, compared with the construction of the query syntax in the prior art which adopts a common tree type data structure, the construction of the dialogue text query syntax in the application adopts the Antlr4 technology which is commonly used in the industry, and storage, checking and conversion are relatively simple.