49 results about "Word selection" patented technology

The invention provides a large language model optimization generation method based on optimal cue word selection, and belongs to the technical field of large language model optimization generation. The technical problem to be solved is to provide the improvement of the large language model optimization generation method based on optimal cue word selection. According to the technical scheme, a selected data set is preprocessed, a sequence is formed by splicing questions and answers, and a spectral clustering algorithm is adopted to divide the sequence into multiple groups of question and answer pairs with similar characteristics; sampling representative samples from each cluster, and based on professional knowledge, language logic and semantic association, constructing high-quality cue words which can guide and generate a large model aiming at a specific problem; a large-parameter online model is used, a diversified and targeted cue word set is generated for each cluster with reference to the created high-quality cue words, generated questions, candidate cue words and answers of the candidate cue words are integrated, and a data set used for fine tuning of the cue evaluation model is constructed; the method is used for generating the language model.

A text entry apparatus, storage medium and method where users type by tapping their fingers in place, called TapGazer, is presented. Users can tap anywhere as long as the identity of each tapping finger can be detected with sensors. The ambiguity between different possible input words is resolved by selecting target words with gaze. If gaze tracking is unavailable, ambiguity is resolved by selecting target words with additional taps.

The embodiment of the invention discloses a keyboard input method for a three-dimensional space, head-mounted display equipment, a medium and a product. According to one specific embodiment, the method comprises the steps that in response to detection of typing operation acting on an associated entity keyboard, a candidate word list is determined according to at least one key position corresponding to the typing operation, an input cursor is displayed in a window corresponding to the typing operation, and the window is displayed in a three-dimensional space of the head-mounted display device; determining a spatial dimension of a window corresponding to the typing operation; in response to determining that the spatial dimension represents a two-dimensional window, generating candidate word column position information according to two-dimensional position information of the input cursor in the window; according to the candidate word column position information, a candidate word column is displayed in the window, and at least one candidate word in the candidate word list is displayed in the candidate word column. According to the embodiment, 3D space input can be supported, and the word selection box can be dynamically focused near the input box.

The present application proposes a learning-based SQL automatic generation method, which includes: obtaining a database to be queried and preset target constraints, and constructing a vocabulary required for generating SQL statements according to the database; generating a finite state automaton according to SQL grammar rules; inputting the encoding of the specified initial word into the constructed word selection probability model, obtaining the SQL statement output based on the encoding of the initial word, and calculating the actual benefit value of each word in the SQL statement; calculating the benefit estimation value of the next word through the constructed word selection benefit estimation model and constraint encoding network, and updating the parameters of each constructed network model according to the difference between the actual benefit value and the benefit estimation value; repeating the above steps until the network model converges, and generating the target SQL statement through the trained word selection probability model. The method can automatically generate SQL statements that meet the current constraints, improve the accuracy of generating SQL statements, and reduce costs.

This invention discloses a spoken language understanding method based on multi-view expert fusion and interest lexical selection, belonging to the field of natural language understanding and semantic parsing technology. The method includes: extracting hidden state sequences from the input utterance using a shared encoder; constructing a multi-view expert fusion module containing utterance view experts, block view experts, and lexical view experts to generate multi-granularity expert features; generating intent aggregation features and slot aggregation features through a task decoupling gating mechanism; performing interest lexical selection on the intent aggregation features, calculating lexical importance scores, generating weighted intent representations, and predicting multi-intent labels; and inputting the slot aggregation features into a decoder with diagonal mask constraints to generate a position-aligned slot label sequence. This invention enhances the model's dynamic focusing ability on key intent signals and can be applied to intelligent dialogue systems, virtual assistants, and vertical domain semantic parsing scenarios.

The embodiment of the invention provides a static memory and a read-write method thereof. The static memory includes: a memory cell; a word line; the word selection signal driving circuit comprises a first switch sub-circuit, a gradual change sub-circuit and a second switch sub-circuit, and the first switch sub-circuit is configured to write a first selection signal of an operation signal line into a first node under the control of an effective signal of a first control signal line; the gradual change sub-circuit is configured to enable the first level line and the second level line to be conducted based on the signal of the first node under the condition that the first control signal line provides an invalid signal, and discharge a first selection signal of the first node, so that the first node generates a gradual change second selection signal; and the second switch sub-circuit is configured to provide a slowly changing second selection signal to the word line under the control of an effective signal of a second control signal line. According to the invention, the noise margin of the read operation can be improved, and the read stability is improved.

An embodiment of the present application discloses a text information retrieval method and device. According to the first technical field to which the text information to be retrieved belongs, a plurality of words are determined, and a seed word library is constructed based on the plurality of words; semantic generalization is performed on the seed word library, and the words after semantic generalization are screened based on a first preset word selection condition to obtain a first-level generalized word library; text information retrieval is performed based on the first-level generalized word library, the number of text information corresponding to each word in the first-level generalized word library is determined, and the words in the first-level generalized word library are screened based on the number of the text information to obtain a second-level generalized word library; semantic generalization is performed on the second-level generalized word library, and the generalized words are screened based on a second preset word selection condition to obtain a third-level generalized word library, so as to perform text information retrieval based on the third-level generalized word library. Through the above method, the retrieval efficiency is improved.

The invention provides an identity authentication system and method based on a natural language, and the system comprises a processor, a user login module, a text input module, a background collection module, a feature recognition module, a text completion module, a keyboard fingerprint generation module, and a language feature vector generation module. The user login module, the text input module, the background acquisition module, the feature recognition module, the text completion module, the keyboard fingerprint generation module and the language feature vector generation module are respectively connected with the processor. According to the method, input feature recognition, interest word selection, word association, sentiment analysis and other technologies are adopted to distinguish people and machines, the two functions of user login and man-machine verification are integrated, the method can be used for system login, website login, APP login and other aspects, a user does not need to memorize complex passwords or carry out complex picture and character recognition to log in an application system, and the user experience is improved. Identity verification and human-machine verification are unified into one process, and the user experience is improved.

The invention discloses a photographing query method and a related device, and relates to the technical field of information query, the photographing query method is applied to an electronic device, the electronic device is provided with a camera, a display screen and a photographing word selection key, and the photographing query key is a physical key used for triggering photographing word selection. The photographing query method comprises the following steps: when a first key event triggered by a user operating a photographing word-taking key is detected, starting a camera, and displaying a preview picture of the camera on a display screen; when a second key event triggered when the user operates the photographing and word taking key is detected, controlling the camera to shoot, and directly displaying a picture shot by the camera on the display screen; and when a selection operation of a user on information in the picture shot by the camera is detected, querying the selected information to obtain related information of the selected information, and outputting the related information. According to the photographing query method disclosed by the invention, the information query operation is simplified, the user operation convenience of information query is improved, and the user experience is further improved.

The invention discloses an AppStore keyword popularity search scale processing method which comprises the following steps: S1, connecting to a database, and executing query to obtain popularity and search volume data; s2, preprocessing the acquired data, wherein the preprocessing comprises data traversal and normalization processing; s3, acquiring a relationship between popularity and exposure by using the data acquired in the S2; s4, generating popularity data according to the relationship obtained in S3, and calculating corresponding exposure; s5, adjusting coefficients of popularity data of different intervals, and correcting a calculation result; s6, checking the monotonicity of the exposure, outputting a result, and inserting the result into a database; and S7, closing the database. Compared with the prior art, the AppStore keyword popularity search scale processing method has the advantages that direct and accurate information is provided in a selected word putting scene according to key indexes, and powerful data support is provided for word selection.

Techniques for altering default language, in system outputs, with language included in system inputs are described. A system may determine a word(s) in user inputs, associated with a particular user identifier, correspond to but are not identical to a word(s) in system outputs. The system may store an association between the user identifier, the word(s) in the user inputs, and the word(s) in the system outputs. Thereafter, when the system is generates a response to a user input, the system may replace the word(s), traditionally in the system outputs, with the word(s) that was present in previous user inputs. Such processing may further be tailored to a natural language intent.

The present application relates to the field of Internet technology, and discloses a method, device, equipment and storage medium for cracking verification codes by clicking on rare characters. The method includes determining the detection frame of each target rare character based on a target detection algorithm, and aligning the initial slider with each detection frame respectively; rotating the initial slider, calculating the graph vector similarity between the initial slider and each target rare character, and determining the target click position and the target click order in a preset background image according to the graph vector similarity. Through the above method, the present application generates an initial slider in a preset background image, aligns the initial slider with each target rare character in the preset background image, rotates the initial slider, and calculates the graph vector similarity between the initial slider and the target rare character during the rotation process, and then determines the target click position and the target click order according to the graph vector similarity, thereby improving the accuracy of cracking verification codes by clicking on rare characters on the target web page.

To estimate important words and highlight display them in a document of a specific field.SOLUTION: A document display support system includes: a database in which selection object words and non-selection object words are registered; a learned word selection model which has been subjected to mechanical learning to estimate whether a certain word is a selection object word or not; a text pre-processing section for cutting out each word from a received display object document; a word classification section for classifying each of the words into one of a selection object word, a non-selection object word, and an indefinite word, on the basis of the database; a text post-processing section for imparting a prescribed attribute to a predetermined word in the display object document and generating output data; and an output section for outputting the output data. The word selection model, when estimating a label indicating that the indefinite word classified by the word classification section is a selection object word, classifies the indefinite word into a selection object word, and the text post-processing section imparts the prescribed attribute to the classified selection object word.SELECTED DRAWING: Figure 2

This application relates to a method, apparatus, device, and medium for generating advertising text in the field of computer technology. The method includes: acquiring the title of an advertised product and its reference advertising text; using a text generation model's encoder-decoder network to decode the title time-step by time to obtain decoded feature information, and determining its generation probability distribution mapped to a first reference vocabulary referenced by the encoder-decoder network; using a text generation model's encoder to obtain encoded feature information of the comprehensive text composed of the product title and its reference advertising text, and determining its key feature information relative to the decoded feature information as a pointer probability distribution; calculating and determining a matching weight based on the encoded feature information and the decoded feature information, and smoothly weighting the generation probability distribution and the pointer probability distribution with the matching weight to obtain an optimized probability distribution; and determining the lexical units of the target advertising text based on their word selection in the first reference vocabulary or the comprehensive text. This application can generate high-quality advertising text.

The present invention provides a document display assistance system which estimates and highlights significant words in a document of a specific field. The system comprises: a database in which selection-target words and non-selection-target words are registered; a learned word-selection model having been applied with machine learning for estimating whether a word is a selection-target word; a text pre-processing unit which segments words from an accepted display-target document; a word classification unit which classifies, based on the database, the word into any of a selection-target word, a non-selection-target word, and an indeterminate word; a text post-processing unit which generates output data by imparting a predetermined attribute to a predetermined word in the display-target document; and an output unit which outputs the output data. If a label is estimated indicating that the indeterminate word classified by the word classification unit is a selection-target word, the word selection model classifies the indeterminate word into a selection-target word and the text post-processing unit imparts the predetermined attribute to the classified selection-target word.

The present specification provides a text word selection method and device, wherein the text word selection method comprises determining initial text containing a to-be-selected word region and a candidate word set corresponding to the to-be-selected word region; performing text processing on the initial text based on the to-be-selected word region and the candidate word set to obtain target text; and performing word segmentation on the target text and inputting the segmented target text into a mask language model to obtain a target vector of a target segmented word in the segmented target text, wherein the mask language model adopts a sparse attention mechanism; determining a target probability value of each candidate word in the candidate word set relative to the target vector based on a word table, and determining a target candidate word based on the target probability value. The text word selection method accurately predicts the answer to the candidate word fill-in by using the mask language model with the sparse attention mechanism and combining the word table.

The invention discloses a large language model fingerprint construction method and system based on maximum activation component coverage, and belongs to the field of artificial intelligence model verification. The invention aims to solve the problems of blind cue word selection and low model component coverage efficiency in existing fingerprint construction. The method comprises the following steps: acquiring a candidate sensitive prompt word set; for each cue word in the set, determining a key component set which is activated by the cue word in the original model and is composed of an attention head, a feedforward network and the like; selecting a target cue word set by adopting a maximum coverage selection strategy (such as a greedy algorithm), and maximizing the weighted coverage rate of key components commonly activated by the set; and finally, forming a model fingerprint by the target cue word set and the expected output thereof. According to the method, the widest coverage is realized by using the least cue words, the discrimination ability, verification efficiency and reliability of fingerprints are remarkably improved, and the verification cost is reduced.

Disclosed in the embodiments of the present disclosure are a method for keyboard input in a three-dimensional (3D) space, and a head-mounted display device, a medium and a product. A specific embodiment of the method comprises: in response to a typing operation performed on an associated physical keyboard being detected, determining a candidate word list on the basis of at least one key position corresponding to the typing operation, wherein an input cursor is displayed in a window corresponding to the typing operation, and the window is displayed in a 3D space of a head-mounted display device; determining a spatial dimension of the window corresponding to the typing operation; in response to it being determined that the spatial dimension represents a two-dimensional window, generating candidate word bar position information on the basis of two-dimensional position information of the input cursor in the window; and on the basis of the candidate word bar position information, displaying a candidate word bar in the window, wherein at least one candidate word in the candidate word list is displayed in the candidate word bar. The embodiment can support input in a 3D space, and can enable a word selection box to dynamically focus near an input box.

The invention discloses a letter fly input system, and relates to the technical field of computer input. The system inherits the advantages of an existing mainstream input method, a 36-key Dvorak real keyboard serves as a hardware basis, and slender rectangular touch screens are additionally arranged on the left side and the right side to replace traditional control keys and mouse functions; a three-in-one framework of'hardware reconstruction + semantic innovation + modal collaboration 'is constructed on the software level, an existing mainstream input method lexicon and a coding engine are reused, candidate words of multiple input methods are displayed in a partition mode through a 9 * 9 large-capacity word window, and cross-scheme word selection can be achieved without switching modes; a'specific word-cue word-choice box 'linkage mechanism is innovated, a functional module is triggered by a specific word easy to remember an initial, and operations such as application control and website access are realized in cooperation with cue word numbers; and cross-device adaptation of the PC and the mobile phone end and sharing of configuration and lexicon data are supported. According to the system, the problems that a traditional keyboard is poor in portability, switching of multiple input methods is tedious, and the function and portability are contradictory are solved, integration of character input and application control is achieved, the operation efficiency and cross-device experience are improved, and the system is a significant upgrade of an existing input method.

1. Name of the Design Product: Electronic Device with Graphical User Interface Assisted by Large Model for Reading 2. Use of the Design Product: Utilize the capabilities of vivo BlueSky Large Model to assist users in reading e-books, and users can have real-time reading interactions with the large model 3. Design Key Points of the Design Product: Lie in the change state diagram 4. Picture or Photograph Most Indicating the Design Key Points: Change state diagram 5. Human-Machine Interaction Method of the Graphical User Interface: Slide to select words in the text area of the front view, click the "AI Reading Companion" function to call out the change state diagram, and the selected word in the front view will be input as a prompt word into the large model dialog box, and then you can have a conversation with the large model about the context of the book

An electronic device, according to one embodiment of the present disclosure, may comprise: a preprocessing unit which segments an inputted story by word and, according to a word meaning inference possibility, distinguishes first context information and second context information and outputs same; a valid word selection unit which extracts, from the first context information, word information satisfying a predetermined criterion, extracts, among the first context information, word information having a specific linguistic form, and carries out expansion into similar words for the second context information; and a scoring unit which receives output values, of the valid word selection unit, for respective analysis elements of essay text data, calculates the degree of similarity between data being compared and the output values, and thus outputs scoring information of the essay text data.

The invention discloses an entity extraction method for an urban rail transit standard knowledge graph. The method comprises the steps that S1, a cue word set is generated through data preprocessing; s2) according to the cue word set, using a cue word selection module, injecting cue words into an embedding layer of a model layer, and generating the embedding layer with the cue words; s3) carrying out information fusion on the embedded layer with the cue word and the input data passing through the embedded layer; s4) performing matrix and function transformation on the data after information fusion and generating entity extraction data; and S5) applying the entity extraction data to the rail transit standard knowledge graph. According to the method, accurate extraction of the rail transit standard is realized, the labeling cost is reduced, the investment of experts in the field of the rail transit standard is reduced, and rapid construction of the knowledge graph corresponding to the rail transit standard is improved.

The invention discloses an input method, a related device, electronic equipment and a storage medium, and the input method comprises the steps: obtaining a plurality of mapping rules in response to a first character sequence of a selected word on a screen in an input process; wherein the mapping rule comprises a second character sequence and a visual pattern associated with the second character sequence; selecting a target rule from the plurality of mapping rules based on a matching result between the first character sequence and the second character sequence in each mapping rule; and carrying out replacement display on the current cursor based on the visual pattern in the target rule. According to the scheme, the feedback and individuation requirements in the input process can be met at the same time.

A method receives audio data from a call. Services are performed to process the audio data to automatically generate a response, wherein the services include converting the audio data to input text, inputting the input text into a model to automatically generate a text response, and converting the text response to an audio response. Supplemental words are selected based on the input text. The method determines a type of service based on services performed to generate the audio response and determines a position in the response to insert the supplemental words based on the type of service. The supplemental words are provided for insertion in the call at the position to supplement the audio response.

The invention discloses a mobile robot vision book checking method and system, and the method comprises the steps: obtaining a spine image, recognizing a book name text in the spine image, and obtaining a book name recognition result; determining at least one word to be corrected based on the book name recognition result and a pre-constructed candidate word bank; performing recall operation on the word to be corrected to obtain a candidate set; calculating the comprehensive score of each candidate word in the candidate set; and based on the comprehensive score, selecting the candidate word with the highest comprehensive score to replace the word to be corrected so as to output a corrected book name. After the spine image is obtained and the book name text is recognized, the word to be corrected is determined based on the pre-constructed candidate word bank, and then intelligent correction of the book name recognition result is realized through the processes of candidate word recall, comprehensive score calculation and optimal candidate word selection. Text recognition errors caused by factors such as spine deformation, non-standard fonts and partial shielding in the visual recognition process are effectively overcome, and the accuracy and robustness of book name recognition are remarkably improved.

The application discloses a text processing method, comprising: collecting original text, and performing data preprocessing on the original text to obtain a list of candidate words; obtaining an anchor word in response to a selection input of a user according to the list of candidate words; processing a preset initial model according to the anchor word to obtain a theme word extraction model corresponding to the original text; and extracting theme words from the original text according to the theme word extraction model to obtain the theme words. The text processing method can preprocess the original text to obtain the list of candidate words, further generate the anchor word according to the selection of the user in the list of candidate words, and then extract the theme words from the original text according to the theme word extraction model, so that the text information is effectively screened with a purpose by referring to the anchor word given by the user.

The embodiment of the invention provides a webpage word cutting function implementation method and device, equipment and a storage medium. The webpage word cutting function implementation method comprises the steps that in response to word selection operation on a pdf page, a corresponding word selection event object is captured, event information is packaged in the word selection event object, and the event information comprises target text content; analyzing event information in the word selection event object to obtain event parameters; and calling a target word cutting function interface, and realizing a target word cutting function according to the event parameters. According to the technical scheme provided by the embodiment of the invention, the text content selected in the pdf page is packaged into the event object with the readable attribute. According to the method, the event object is analyzed to obtain corresponding event parameters such as text content, and finally the event parameters are input into the word cutting function interface, so that the corresponding word cutting function is realized, and the problem of capability failure or degradation of the word cutting function when facing a pdf page is solved.

This relates to an information processing system capable of generating images that visualize the impression of items, services, etc., that are associated with scent. [Solution] The information processing system comprises an acquisition unit, a display word selection unit, a color determination unit, and a display word placement unit. The acquisition unit acquires text data relating to the impression of the subject, information on the type of scent of the subject, and information on the display type of the communication image. The display word selection unit uses an empathy word library containing multiple empathy words to select a display word from among multiple words included in the text data that is the same as an empathy word included in the empathy word library and will be placed in the communication image. The color determination unit uses relationship information between the type of scent and the color imagined from that scent to determine an image color to color the display word from the information on the type of scent of the subject acquired by the acquisition unit. The display word placement unit places the display word colored with the image color within the display type.

AI-driven definition generation system and method disclosed herein guides an artificial intelligence (AI) engine to provide contextually relevant definitions of words given in a passage. The AI-driven definition generation system includes an online learning platform such that a user can select any word given in a passage on the online learning platform to instantly receive a contextually relevant definition of the selected word. The definition is generated as per the reading level of the user and the context of the surrounding passage. The AI-driven definition generation system achieves this by integrating user profile information, analyzing the surrounding text of selected words, and using an AI engine guided by a content generation system.

1. The name of the design product: large model assisted reading graphical user interface of electronic equipment. 2. The use of the design product: vivo blue heart large model capacity combined with knowledge base to assist users in reading electronic books. Users can interact with the large model in real time. 3. The design points of the design product: in the change state diagram 3. 4. The picture or photo that best indicates the design points: change state diagram 3. 5. The use of the graphical user interface: text area sliding word selection in the main view, click "AI reading companion" function to call out the change state diagram, the selected word in the main view will be input as a prompt word into the large model dialog box, and then the user can dialogue with the large model about the context content of the book. Books can be selected as a unified pre-knowledge base for reading and answering questions. 6. The human-computer interaction mode of the graphical user interface: long press and click AI reading companion in the main view to open change state diagram 1; click the input text below change state diagram 1 to call out change state diagram 2; click the back key in change state diagram 2 to open change state diagram 3; click the back key in change state diagram 3 to open change state diagram 4.