40 results about "Word language" patented technology

A problem to be solved is to robustly detect a pronunciation variation example and acquire a pronunciation variation rule having a high generalization property, with less effort. The problem can be solved by a pronunciation variation rule extraction apparatus including a speech data storage unit, a base form pronunciation storage unit, a sub word language model generation unit, a speech recognition unit, and a difference extraction unit. The speech data storage unit stores speech data. The base form pronunciation storage unit stores base form pronunciation data representing base form pronunciation of the speech data. The sub word language model generation unit generates a sub word language model from the base form pronunciation data. The speech recognition unit recognizes the speech data by using the sub word language model. The difference extraction unit extracts a difference between a recognition result outputted from the speech recognition unit and the base form pronunciation data by comparing the recognition result and the base form pronunciation data.

A speech recognition system provides a subword decoder and a dictionary lookup to process a spoken input. In a first stage of processing, the subword decoder decodes the speech input based on subword units or particles and identifies hypothesized subword sequences using a particle dictionary and particle language model, but independently of a word dictionary or word vocabulary. Further stages of processing involve a particle to word graph expander and a word decoder. The particle to word graph expander expands the subword representation produced by the subword decoder into a word graph of word candidates using a word dictionary. The word decoder uses the word dictionary and a word language model to determine a best sequence of word candidates from the word graph that is most likely to match the words of the spoken input.

A language model comprising a plurality of augmented-word n-grams and probabilities corresponding to such n-grams. Each n-gram is comprised of a sequence of augmented words. Each augmented word is comprised of the orthographic representation of the word together with a tag representing lexical information regarding the word, such as syntactic or semantic information. Also disclosed are a method of building such a language model, a method of automatically recognizing speech using the language model and a speech recognition system that employs the language model.

The invention relates to a method for displaying Mongolian on a GNOME desktop system platform of an LINUX system. The method comprises steps of building a Mongolian processing system engine in a Pango system processing word language in the GNOME desktop system, registering a name of the Mongolian processing system to the Pango system executing word langue processing, forming an interface between the Mongolian processing system engine and a word langue processing module of an operation system, generating a Mongolian processing module based on rules and structures of an Open Type font in the Mongolian processing system engine, constructing an font section engine to select and replace the Open Type Mongolian font, and finally obtaining correct Mongolian display results after font selecting replacement. Mongolian display and intelligent input thereof on the basis of the Unicode in the Linux operation system are realized by the method, and the Mongolian display and the intelligent input method thereof can be used together with Chinese or other language input methods which are loaded and can not affect original functions and applications thereof.

The invention relates to a voice recognition text error correction method in a specific field, wherein the method comprises the following steps: firstly, performing statistics by using correct field corpora to obtain a character and word level language model and a pinyin language model; then, receiving a text sequence to be subjected to error correction, and performing clause processing on more than one sentence; determining suspected wrong words by using a word, word and pinyin language model; determining a candidate word list of the suspected wrong words according to a language model vocabulary and a pronunciation-prone dictionary; and finally, substituting candidate words into the original text sequence, and selecting and outputting the most reasonable sentence in combination with macroscopic and microcosmic scores. Basic units with different granularities and dimensions such as characters, words, pinyin and initial and final consonants are selected to construct a language model, and word segmentation error interference caused by wrong characters is reduced; isolated character disorder is processed by adopting a word language model, and continuous recognition errors caused by pronunciation deviation is distinguished by adopting the pinyin language model; and candidate sentences after the wrong words are replaced are comprehensively evaluated by macroscopic and microcosmic scores, and the smoothness degree of the replaced sentences are measured.

The invention provides a character entry method and device, wherein the character entry method specifically comprises the steps of: presetting binary relation data between character sequences corresponding to a language consisting of letters and a language consisting of pinyin and / or strokes; receiving user entry; and analyzing the user entry by using the binary relation data, and generating a language including letter group words and a character output of word language mixing, wherein the character output consists of the pinyin and / or strokes. According to the invention, under multiple language infiltration scenes of similar Chinese and English mixed entry, a candidate item according with user demands is obtained.

A computer-based information handling system comprises a processor for executing an application program, audio output device, and video display device including a display screen for displaying a word in a language for audible playback. A pointing device controls a cursor movable on the display screen in response to a user operating the pointing device. A memory is provided for storing a digital recording of the word for audible playback and a rollover region is associated with the word for playback, defined at a position on the display screen overlapping a position of said word for playback on the display screen. The rollover region is configured to cause audible playback of the word in the on-screen language when at least a portion of the cursor is moved over the rollover region. In further aspects, a method, computer-readable medium, language instruction system, markup language document, and method for developing a language instruction system are also provided.

A problem to be solved is to robustly detect a pronunciation variation example and acquire a pronunciation variation rule having a high generalization property, with less effort. The problem can be solved by a pronunciation variation rule extraction apparatus including a speech data storage unit, a base form pronunciation storage unit, a sub word language model generation unit, a speech recognition unit, and a difference extraction unit. The speech data storage unit stores speech data. The base form pronunciation storage unit stores base form pronunciation data representing base form pronunciation of the speech data. The sub word language model generation unit generates a sub word language model from the base form pronunciation data. The speech recognition unit recognizes the speech data by using the sub word language model. The difference extraction unit extracts a difference between a recognition result outputted from the speech recognition unit and the base form pronunciation data by comparing the recognition result and the base form pronunciation data.

The invention discloses a generation method of video indexing data and a device. The generation method of the video indexing data comprises the following steps: obtaining video content and text content which is relevant to the video content, classifying the text content, selecting a proper pinyin language model and a proper word language model according to a classification result, segmenting voice data of the video content and classifying speakers, selecting a proper acoustic model according to a speaker classification result, generating a pinyin gridding according to the selected acoustic model and the selected word language model and a first pronunciation dictionary according to the text content, obtaining a word gridding according to the pinyin gridding, the word language mode corresponding to the text content, and a second pronunciation dictionary, recalculating a confidence coefficient of the word gridding to obtain a new work gridding according to the pinyin gridding and the word gridding, and finally combining the new gridding with original video content to obtain the video indexing data. According to the video indexing data, a user can conveniently and accurately retrieve the relevant video content through text keywords.

The invention discloses a speech recognition method, device and apparatus. The method comprises the following steps: receiving voice from a user; obtaining a hot word language model, wherein the hot word language model is a language model obtained by training according to hot words provided by the user; decoding the voice by using the hot word language model and a preset main language model. According to the method, at least the hot word recognition accuracy can be effectively improved.

The invention discloses an address identification device, comprising units as follows; a key word deletion judging unit that judges whether or not a key word in an input address image is deleted; an integral address identification unit that identifies integrally the address region among the key words when the key word deletion judging unit judges that the key word in the input address image is deleted; a word language address identification unit that identifies the word language for the input address image when the key word deletion judging unit judges that the key word in the input address image is deleted; a reliability judging unit that judges reliability of the address identified by the integral address identification unit; wherein when the reliability judging unit judges that the address identified by the integral address identification unit is not reliable, the word language address for the input address image is identified by the word language address identification unit.

The invention belongs to the technical field of image recognition, and particularly relates to a text positioning method and system based on the visual structure attribute. Based on the visual attribute of a text, by means of color polarity difference transformation and edge neighborhood tail end bonding, abundant closed edges are detected so that abundant candidate connection elements can be obtained, then character stroke attributive character and text colony attributive character screening is conducted, the connection elements belonging to characters are extracted from the candidate connection elements, and then the final text is positioned through multi-channel blending and repeated connection element removal. The method is high in robustness and can be adapted to the situation that multiple word language categories are mixed, or various font styles exist, or arrangement directions are random, or background interference exists and other situations, the positioned text can be directly provided for OCR software for recognition, and OCR software recognition rate can be increased. The text positioning method and system based on the visual structure attribute can be applied to image video retrieval, junk information blocking, vision assisted navigation, street view positioning, industrial equipment automation and other fields.

The invention relates to a network translation inquiry system embedded in a webpage and a method thereof, which can solve the problem of inconvenient operation of translation explanation inquiry on a network inquiry system through the technical means comprising the following steps of: towing a cursor by a user to select a translation word after triggering the webpage in which the system of the invention is embedded, transmitting the translation word selected by the user and browser information to an appointed servo end, recognizing the word language system of the translation word and the operation language system of a network browser on the appointed servo end, translating and converting the translation word into the operation language system according to the word language system to inquire a translation result and returning the translation result to a client end to display, thereby achieving the technical effect of simplifying the network inquiry of translation explanation.

The invention discloses a multi-language instant translation system based on big data processing and manual intervention. The system comprises a parallel corpus and a third-party translation engine. A translation method of the system comprises the steps that a user inputs a to-be-translated language into the parallel corpus, wherein the to-be-translated language is a word language; the parallel corpus retrieves the to-be-translated language, and whether the parallel corpus can directly translate the to-be-translated language is judged; if the parallel corpus can directly translate the to-be-translated language, internally recorded information is retrieved, and then a retrieved translation result is output; if the parallel corpus cannot directly translate the to-be-translated language, the to-be-translated language is input into the third-party translation engine to be translated, and a translation result is output; and the translation result is modified manually, the information obtained after modification is fed back to the parallel corpus, and the information in the parallel corpus is updated continuously. According to the system, through combination of the parallel corpus and the third-party translation engine in combination with manual modification, the translation effect is more intelligent, and meanwhile the translation engine has a learning function.

The invention relates to a word language managing technique and a computer data structure technique, in particularly a method for constructing a word language structure tree. The method includes steps as follows: a rule for converting the word language to a space position and converting; a rule design construction of a word language structure tree code information; a method for synthesizing and managing the word language structure tree code information; a method for analyzing and identifying the word language structure tree code information; processing the word language through the computer or other devices that can calculate and store for obtaining the word language tree junction. The method can manage word language simply, directly and efficiently, the user can use the word language tree for reaching same effect when the user manages the word language. The word language tree has certain regularity that can increase efficiency greatly in processing the word language tree. The method has strong practicability.

Given a language with all words in a fixed length, and a set of regular expressions composed only from characters in the alphabet of the language or the “?” sign (any single character), the method of the invention defines a data structure that is used to efficiently find the set of matching regular expressions for a given query word. The method may be adjusted by appropriate selection of a control variable to vary the storage space required and the search time necessary to complete the query. Specifically, the method of the present invention provides a space versus time trade-off between the storage space required for the data structures of the present invention and the amount of time to search those data structures to determine the matching set of regular expressions.

The invention discloses a text detection and recognition method and system combined with text classification. The method comprises the following steps: acquiring all target text line boxes in a target picture; cutting and extracting all the target text line frames to obtain a text graph; sending the text graph into a text direction classification model, and carrying out correction identification to correct the text graph in any direction to the same horizontal direction so as to obtain a text correction graph; sending the text correction graph to a character language classification model, and carrying out character language category recognition to obtain a character language category image; and sending the character language category image into a language text recognition model corresponding to the language category, and performing recognition to obtain a final text content. According to the invention, the problem that text detection in the prior art cannot detect texts with any shapes and complex scenes is solved; the conditions that the text is reversed and the direction is not positive cannot be detected; and the problems of high time cost and low efficiency caused by the fact that multi-language text regions need to be sent into a plurality of models for recognition are solved.

The invention discloses a sign language recognition and conversion system and method based on deep learning and big data. The system comprises an image acquisition module, an image recognition module,an information matching module, a content arrangement module, a text output module and a voice output module. The method includes: collecting a human body image sequence; extracting face key point coordinates and hand key point coordinates in each frame of image of the human body image sequence; searching natural language morphemes most matched with the face key point coordinates and the hand keypoint coordinates in a sign language action database, and calculating matching values; filtering the natural language morphemes according to the repetition condition and the matching value between the adjacent morphemes; converting the reserved natural language morphemes into characters and displaying the characters on a screen; and searching voices corresponding to the characters according to the character language database, and playing the voices. According to the system, the sign language image sequence can be conveniently and quickly converted into characters and voice of other languagesto be output, the meaning of the sign language can be understood more easily, and the communication efficiency is improved.

The invention discloses an RTL graphical description method which overcomes the defect that three expression manners including a data path, a structural graph and a logistical graph do not involve conversion of an RTL language and a graphical expression manner, creatively invents a method for uniqueness conversion between the RTL language and a signal flow graph, gives full play to advantages including visualization of a graphical language and concreteness of a word language and brings great convenience to teaching and researches.

The invention discloses a text language type switching method and device, equipment and a storage medium. The method comprises the steps of receiving typewriting input of a user; in response to the typing input, displaying a first text field; receiving a first input of a user; in response to the first input, determining a to-be-switched text field of the first language type in the first text field; receiving a second input of the user; and in response to the second input, switching the to-be-switched text field to a target text field of the second language type. According to the method, a large number of unnecessary editing operations of the user can be avoided, so that the editing experience of the user is improved.

The invention discloses an address identification device, comprising units as follows; a key word deletion judging unit that judges whether or not a key word in an input address image is deleted; an integral address identification unit that identifies integrally the address region among the key words when the key word deletion judging unit judges that the key word in the input address image is deleted; a word language address identification unit that identifies the word language for the input address image when the key word deletion judging unit judges that the key word in the input address image is deleted; a reliability judging unit that judges reliability of the address identified by the integral address identification unit; wherein when the reliability judging unit judges that the address identified by the integral address identification unit is not reliable, the word language address for the input address image is identified by the word language address identification unit.

The present invention provides a training system and training method based on the function of the visual macrocell-dorsal pathway, including a consistent motion detection module, a visual search module and a visual tracking module. The consistent motion detection module is used to train the user in consistent motion and record the step transition ratio; the visual search module is used to train the user's visual search ability; the visual tracking module is used to train the user's time tracking ability. The present invention is used for training multiple cognitive processing processes responsible for the visual macrocell-dorsal pathway, improving the various abilities of the visual magnolia-dorsal pathway, and then improving its language cognitive ability. This training system can also effectively avoid the impact of the text differences between Chinese and Western Pinyin languages ​​on the training effect, and can directly compare the training effects under different language conditions, helping to determine the cause of dyslexia.

The invention relates to the field of artificial intelligence, discloses a certificate picture generation method, device and equipment and a storage medium, which are used for improving the accuracy of generating a certificate picture conforming to a real scene. The certificate picture generation method comprises the steps of obtaining a sample certificate picture, wherein the sample certificate picture comprises sample text data and sample background data; using a picture similarity comparison algorithm for forming certificate background data and certificate character data based on the sample certificate picture, and the certificate character data comprising character language data and font style data; writing the certificate text data into the random position of the certificate background data to generate an initial certificate picture; preprocessing the initial certificate picture to generate a plurality of preprocessed certificate pictures; and adopting a preset random scaling function to perform multiple times of random scaling on the plurality of preprocessed certificate pictures, so that a plurality of target certificate picture groups can be generated. In addition, the invention also relates to a blockchain technology, and the sample certificate picture can be stored in the blockchain.

An example embodiment may involve a software application executable on computing devices of a remote network management platform containing a computational instance associated with a managed network. A text string may be received, and characters of the string may be categorized among a plurality of symbol script families. A respective likelihood of the string corresponding to each family may be determined, and a respective probability of the string being in each language of each given family may also be determined. The respective probabilities for the languages of each given family may be weighted by the likelihoods of the given family, and then weighted sums of the probabilities for each language may be computed. The maximum of the weighted sums may correspond to the language of the text string. The respective probabilities may be determined according to hybrid N-gram and word language models for each family.

The invention provides an ID mapping method based on a regulation and control cloud platform, and the method comprises the steps: obtaining the equipment information of regulation and control cloud and D5000 of a power system, and carrying out the classification of the equipment information of the regulation and control cloud and the D5000 of the power system according to the types; on the basis of classification, sequentially utilizing a pre-constructed Chinese equipment word bank to carry out word segmentation processing on equipment in the power system regulation and control cloud and equipment in D5000, and constructing text information corresponding to the equipment in the power system regulation and control cloud and the equipment in D5000 by considering the occurrence frequency of word segmentation; based on the text information corresponding to the equipment in the power system regulation and control cloud and the text information corresponding to the equipment in the D5000, determining a relationship between the equipment in the power system regulation and control cloud and the equipment in the D5000; through the cloud platform and the D5000 which are matched in a mappingmode, efficient collection and storage can be well conducted, and information exchange sharing and analysis synchronization are achieved; according to the method, the difference of character languagesis fully considered, word segmentation is performed on equipment names by establishing a Chinese word bank, and the semantic difference is reduced.

Embodiments of the invention disclose a symbol reasoning intention recognition method and apparatus, and an electronic device, which can solve the problem that a machine cannot recognize complex instructions such as intentions, meet the user demand that a user hopes that the machine can quickly recognize own intentions, and improve user experience. The symbol reasoning intention recognition method comprises the steps that firstly, a target intention of a user is obtained, and the target intention comprises characters, languages, voices or scripts input by the user; secondly, at least two layers of symbol generation models are utilized, the target intention is decomposed layer by layer, decomposition is stopped until an undecomposable execution task is obtained, the symbol generation models comprise a natural semantic understanding model or a long and short term memory model, and the execution task can be recognized and executed by a machine; and finally, generating an executable instruction according to the non-decomposable execution task, the executable instruction being an instruction which can be executed by a machine and generates an effect.

The invention relates to a search text processing method and device, computer equipment, a storage medium and a computer program product. The method comprises the following steps: acquiring a search text for searching commodities; the method comprises the following steps: performing error correction on commodity words extracted from a commodity corpus to obtain a commodity word bank, and performing word segmentation processing on a search text to obtain a word sequence; phrases formed by the independent words in the word sequence and the adjacent words of the independent words are used as potential wrongly-identified words in the search text; on the basis of the pinyin editing distance, candidate words used for correcting the potential wrongly-identified words are searched; using the commodity corpus after error correction to train the language model to obtain a commodity word language model, and determining statement smoothness of potential wrong words and candidate words; and when the statement smoothness of the potential wrong words and the target candidate words meets a replacement condition, replacing the potential wrong words in the search text with the target candidate words to obtain an error correction text. The method is suitable for a commodity search scene.

The embodiment of the present invention discloses a monitoring method, device and electronic equipment based on a wearable device, which relate to the field of artificial intelligence and can effectively improve the monitoring efficiency of a monitor. The monitoring method includes: the server receives the sound file uploaded by the wearable device and recognizes the voice in the sound file, and generates a text file corresponding to the voice, wherein the text file records the voice corresponding to the voice text language and time stamp; send the text file to the terminal device held by the listener; the terminal device receives the text file sent by the server, sends a download request to the server to download the sound file at the specified time, and sends the specified time file sent by the server The sound file of the time is downloaded or cached locally and then played. The present invention is suitable for monitoring servers and terminal equipment.

The invention relates to a network translation inquiry system embedded in a webpage and a method thereof, which can solve the problem of inconvenient operation of translation explanation inquiry on a network inquiry system through the technical means comprising the following steps of: towing a cursor by a user to select a translation word after triggering the webpage in which the system of the invention is embedded, transmitting the translation word selected by the user and browser information to an appointed servo end, recognizing the word language system of the translation word and the operation language system of a network browser on the appointed servo end, translating and converting the translation word into the operation language system according to the word language system to inquire a translation result and returning the translation result to a client end to display, thereby achieving the technical effect of simplifying the network inquiry of translation explanation.