TW200538969A - Handwriting and voice input with automatic correction - Google Patents

Handwriting and voice input with automatic correction Download PDF

Info

Publication number
TW200538969A
TW200538969A TW94103440A TW94103440A TW200538969A TW 200538969 A TW200538969 A TW 200538969A TW 94103440 A TW94103440 A TW 94103440A TW 94103440 A TW94103440 A TW 94103440A TW 200538969 A TW200538969 A TW 200538969A
Authority
TW
Taiwan
Prior art keywords
word
plurality
candidate
words
input
Prior art date
Application number
TW94103440A
Other languages
Chinese (zh)
Inventor
Alex Robinson
Ethan Bradford
David Kay
Van Meurs Pim
James Stephanick
Original Assignee
America Online Inc
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Priority to US54417004P priority Critical
Priority to US11/043,506 priority patent/US7319957B2/en
Priority to US11/043,525 priority patent/US20050192802A1/en
Application filed by America Online Inc filed Critical America Online Inc
Publication of TW200538969A publication Critical patent/TW200538969A/en

Links

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING; COUNTING
    • G06KRECOGNITION OF DATA; PRESENTATION OF DATA; RECORD CARRIERS; HANDLING RECORD CARRIERS
    • G06K9/00Methods or arrangements for reading or recognising printed or written characters or for recognising patterns, e.g. fingerprints
    • G06K9/62Methods or arrangements for recognition using electronic means
    • G06K9/72Methods or arrangements for recognition using electronic means using context analysis based on the provisionally recognised identity of a number of successive patterns, e.g. a word
    • G06K9/723Lexical context
    • GPHYSICS
    • G06COMPUTING; CALCULATING; COUNTING
    • G06KRECOGNITION OF DATA; PRESENTATION OF DATA; RECORD CARRIERS; HANDLING RECORD CARRIERS
    • G06K2209/00Indexing scheme relating to methods or arrangements for reading or recognising printed or written characters or for recognising patterns, e.g. fingerprints
    • G06K2209/01Character recognition

Abstract

A hybrid approach to improve handwriting recognition and voice recognition in data process systems is disclosed. In one embodiment, a front end is used to recognize strokes, characters and/or phonemes. The front end returns candidates with relative or absolute probabilities of matching to the input. Based on linguistic characteristics of the language, e.g. alphabetical or ideographic language for the words being entered, e.g. frequency of words and phrases being used, likely part of speech of the word entered, the morphology of the language, or the context in which the word is entered), a back end combines the candidates determined by the front end from inputs for words to match with known words and the probabilities of the use of such words in the current context.

Description

There are strict restrictions on the appearance, and strict restrictions on the input menu, etc.), and minor issues. Accept text today. Recently, from portable computers, handheld phones, mobile phones, and other user-friendly portable users to edit documents and messages, such as for the simultaneous sending and receiving of e-mail

200538969 发明 Description of the invention: [Technical field to which the invention belongs] The present invention relates to human language using data processing systems, such as handwriting recognition and speech recognition on desktop computers, handheld computers, and personal data. [Previous technology] Due to memory restrictions, size, and correction of text control (buttons, text input on the device is a pick-up handheld computer device has become smaller and the development of personal data assistant to two-way paging technology has been derived for a Small facets to accept text input to messaging systems and especially the " 5 (e-mail) or SMS system. Over the years, portable computers have become smaller and smaller. In making a smaller computer with One size-limiting element in the effort is the keyboard. If quasi-typed keys are used, the portable computer is at least as large as the keyboard. The reduced keyboard has been used on the portable computer, but the reduced keyboard keys are too small to be used. Simple or fast ladder making with sufficient accuracy. Adding a full-size keyboard to a portable computer will also I1 and hinder the true portability of the computer. Most portable computers cannot be operated without being placed on a flat work surface. Allow the user to use both hands to input. One user can move when the stand-up or move-in brain-brain-feeding double-piece is small and moves 5 200538969 It is easy to use a portable computer. Handwriting recognition is a method that has been adopted, which can solve the problem of text input on a small device with an electronic induction screen or tablet that detects the movement of a finger or stylus. In the latest generation of small portable computers, which are personal digital assistants (PDAs), companies try to solve this problem by adding handwriting recognition software to the pda. A user can write on a touch sensor pad or display screen Enter text directly. The recognition software

This handwritten text is converted into digital data. Typically, the user writes text in real time and the PDA recognizes a character in real time. Writing on the touch sensor or display screen creates a data input string indicating the contact point. The handwriting recognition software analyzes the geometric features of the data input string to determine that it matches the-character that the user is writing. The handwriting recognition software typically performs geometric shape recognition to determine the handwriting characters. Unfortunately, Italian. The current hand-powered personal electric small device is sexual; and for these reasons, personal letters are one of the letters of the system and the result of writing the letter is very low. The solution is ancient 4 ☆ There are many problems. For example, even in the brain, the handwriting recognition software is not very accurate; while the memory limit is further increased-further restricting the accurate writing style of handwriting recognition is also related to It is used for training, and different from this handwriting software. By many handwriting or 'graffit 彳, Taichankou requires the user to learn a group of specific pen days. This animal a #

First, the stroke combination is used to simplify the appearance recognition process and to perform the A recognition rate. These pens are very different in their natural way. The final product adoption is the question posed above. Speech recognition is another way to solve the problem of text input. 6 200538969 First, our θ identification system typically includes a microphone to detect and record the speech input. The θ turn is digitized and analyzed to extract a speech sample. Recognition typically requires a powerful system to process the speech input. Some limited voice recognition systems have been used on small devices, such as mobile phones for voice-controlled operations. For voice control operations, a device only needs to recognize several commands. Even for speech recognition based on a limited range of speech, because the sample of Ling will change with different users and different situations, a small device does not have satisfactory speech recognition accuracy. It would be advantageous to develop a more practical system to handle human language input. The system has a user-friendly approach, such as a handwriting recognition system for input, and a natural input for a handwriting or speech recognition system for speaking in a natural way. Input, the system has improved accuracy and reduced computing requirements, such as reduced memory requirements and processing power requirements. [Summary of the Invention] A hybrid method is described here to improve the handwriting recognition and speech recognition on the Behr processing system. In an embodiment ... the front end is used to identify strokes, characters, syllables and / or phonemes. The front end returns candidates with relative or absolute likelihoods that match the input. Linguistics based on the language, such as letters or ideographic languages; words in input, such as words or phrases being used 7 200538969

The frequency of the input word, the possible part of the input word, the type of the language; or the context of the input word, a back end combined with the front end to determine the candidate determined from the word input to match the known word and the Possible uses of these words in the current context. The backend can use wild cards to select candidate words, use linguistic features to predict a to-be-completed word or complete continuation words, present candidate words for user selection, and / or provide additional output, such as characters Auto-accent, auto-capitalization, and automatic punctuation and definition symbols to assist the user. In one embodiment, a single language backend is used simultaneously for multiple input modes, such as speech recognition, handwriting recognition, and keyboard input. An embodiment of the present invention includes a method for processing language input on a data processing system. The method includes: receiving multiple recognition results for a plurality of word components, and processing user input of a word in a language, and A plurality of recognition results and one or more candidate words input by a user who judges the word among the possibility of using the word list. At least one of the multiple recognition results includes multiple candidate word components and multiple likelihood indicators. The multiple likelihood indicators indicate the degree of likelihood that the multiple word components conform to a portion of the user input relative to each other. In one embodiment, the candidate word component includes a stroke from handwriting recognition, characters from handwriting recognition, and phonemes from speech recognition. The language can be alphabetic or ideographic. In one embodiment, determining one or more candidate words includes: eliminating a plurality of candidate word combinations of the plurality of recognition results, selecting a plurality of candidate words from a word list of the language, and the plurality of candidate words. The combination of candidate word components containing the plurality of recognition results, the judgment of the one or more candidate words from the plurality of recognition results and data indicating the possibility of using a word list 8

200538969 or more likelihood indicators to indicate the likelihood entered by a user who matches the word, or to rank the one or more selected words based on one or more likelihood indicators. In one embodiment, a candidate is automatically selected from one or more candidate words and presented to the user. The automatic selection can be performed based on any phrase in the language, word pairs in the language, and word tri grams in the language. It can also perform automatic selection based on any morphology of the language and the grammatical rules of the language. The automatic selection may also be performed based on the text entered by the user of the word received. In one embodiment, the method further includes predicting a plurality of candidate words based on a word that is expected to be automatically selected using an input continuation word. In one embodiment, the method includes presenting the one or more candidate words for a user to select, and receiving a user input to select one of the plurality of candidate words. In one embodiment, the plurality of recognition results of a word component include an indication that any one of the candidate word components has the same possibility to match a portion of the user word input. The data indicating the availability of the word list may include any of the frequency of use of words in the language, the frequency of use of words by a user, and the frequency of use of words in a document. In one embodiment, the method further includes any of automatically accenting one or more characters, automatically capitalizing one or more characters, automatically adding one or more punctuation marks, and automatically adding one or more defined symbols. An embodiment of the present invention includes a data processing system that can recognize a word when the word selection group of the string can use a token 9 200538969

A method for inputting. The method includes: processing a user input of a word in a language through pattern recognition to establish a plurality of recognition results for a plurality of word components individually, and pointing out a plurality of recognition results and a word list. The possibility of using the data to determine one or more candidate words of the user input words. At least one of the plurality of recognition results includes a plurality of candidate word components and a plurality of likelihood indicators. The plurality of likelihood indicators indicate a degree of likelihood that the plurality of word components conform to a portion of the user input relative to each other. The pattern recognition may include handwriting recognition, where each of the plurality of candidate word components includes a day, such as for an ideographic language symbol or alphabetic character; or a character, such as for an alphabetic language. The word can be a one-letter word or an ideographic symbol. The pattern recognition may include speech recognition, where each candidate word component includes a phoneme. In one embodiment, one of the plurality of recognition results of a word component includes an indication that any one of a group of candidate word components has an equal likelihood of matching a portion of the word entered by the user. The set of candidate word components contains all alphabetic characters of the language. The data indicating the possibility of using the word list may include any of the frequency of use of words in the language, the frequency of use of words by a user, and the frequency of use of words in a document. The data indicating the use possibility of the word list may include any one of data indicating the form of the language and data indicating the grammatical rules of the language. The data indicating the frequency of use of the word list may include data indicating the context in which the user has entered the word. In one embodiment, the user input specifies only a portion of a complete combination of word components of the word. The system determines the candidate word. 10 200538969 In one embodiment, the one or more candidate words include a part of the words formed by the candidate word component combinations in the plurality of recognition results and a part of the words contain the candidate word component combinations in the recognition results. . In one embodiment, the one or more candidate words include a plurality of candidate words. The method further includes: presenting the plurality of candidate words for selection, and receiving a user input to select one of the plurality of candidate words.

In one embodiment, the method further comprises: predicting one or more candidate words based on selecting a word based on predicting a continuous word input by a user. In one embodiment, the plurality of candidate words are presented in an order of likelihood that matches the words entered by the user. In an embodiment, the method further includes: automatically selecting a most likely one from one or more candidate words as a recognition word of a word input by the user. In one embodiment, the method further includes predicting one or more candidate words based on predicting a most likely word of a contiguous word input by a user. In one embodiment, the method further includes any of automatically accenting one or more characters, automatically capitalizing one or more characters, automatically adding one or more punctuation marks, and automatically adding one or more defined symbols. In one embodiment, each of the plurality of recognition results includes a likelihood index that is individually related to a plurality of candidate word components to indicate a relative likelihood that it matches a portion of the user input. [Embodiment] 11 200538969 The language and language of the main handheld audio discriminator

Syllable rather than word input

For example, URL, practical knowledge, and number input methods, such as handwriting recognition and speech recognition, can be important alternatives to traditional input methods, especially for small computers, personal data assistants, and mobile phones. Traditional handwriting recognition systems face a challenge that requires more than is available on small electronic devices. The present invention improves the sound input technology on these devices by automatically correcting to reduce handwriting or speech recognition memory requirements and processing power requirements. The present invention uses a hybrid approach to enhance the hand of data processing systems for speech recognition. In one embodiment, a front end recognizes the pen day, word, and / or phoneme and returns a relative or absolute candidate that matches the input. Different candidates can be returned for further processing by a back end. Only one candidate is selected using the front end. The back end combines the candidates identified by the front end to match the known word and the possibility of using the current word. By combining the front-end rain back end, the system has an improved recognition rate and is more user-friendly. One of the system's write and speech recognition inputs is effective and has a low memory / CPU implementation. In the present invention, a "word" means any linguistic string forming a word, word stem, prefix or suffix, abbreviation, slang, emoticon, use Or one or more characters or symbols of a sequence of ideographic characters. In this embodiment, a front end is used to execute samples on the language input such as handwriting, voice input, and so on. Many techniques have been used to compare this target style, such as pen day, handwriting, and phonetic keyboard settings, language memory, recognition, text, writing, recognition, and possible reasoning. Invented from the following text. Audio ID, Mingzhi type input 12 200538969

Stress, etc. Typically'-input matches the target patterns to varying degrees. For example, a handwritten letter may be similar to the characters "a" or "c", "0" or ‘,,,’. Currently available pattern recognition techniques can determine the handwritten letter as < ability of any of these characters. However, a'-identification system is typically forced to report only one match. Therefore, the character with the highest probability of matching is typically reported as a recognition result. In one embodiment of the present invention, several candidates are sent to the backend as possible choices, rather than excluding other candidates in advance to get a possible mismatch, so the backend uses the context to match The language input determines overall more likely candidate combinations, such as a word, a phrase, a word pair, a triplet, or a word that fits before or after a sentence, for example, according to a grammatical structure. For example, different candidate words can be determined from different combinations of character candidates among the words that the user is trying to enter. From the frequency of using the word in the language and the relative or absolute likelihood of matching the candidate character, the backend can determine the word that the user is most likely to type. This is different from the traditional method, which provides a set of independently determined most likely characters that cannot even form a meaningful word. Therefore, the present invention combines a precise word search software with a handwriting recognition (HR) engine or a speech recognition (SR) engine to provide a small electronic device such as a personal digital assistant, a telephone, or a mobile phone. Many specific devices for inputting text and data. Feng Liren Li Xing gave a powerful solution to the persistent problem of Wuyin input. The engine can effectively serve a variety of low-level memories. In addition, the present invention uses a single back-end input type (standard keyboard, handwriting, voice), 13 200538969, and processor requirements.

Figure 1 illustrates a diagram of a system for recognizing user input on a data processing system according to the present invention. After the language input 101, such as handwriting or speech, is received by the style recognition engine 103, the style recognition engine 103 processes the input to provide candidate word components such as characters, phonemes, or pen days and their corresponding parts that match the input 105 Possibility. For example, a character input can match a candidate character list and cause ambiguity. In one embodiment, the obfuscation is tolerated at the front-end level and transmitted to the language non-fuzzy back-end for further processing. For example, a word-based non-fuzzy engine 1 07 compares the word list 1 09 with possible combinations of the characters to establish candidate words and their association possibilities that match the user input 111. Since less frequently used words or unknown words are less likely to match the user input if they are not in the word list 1, 09, these candidate words can be downgraded and have a lower likelihood of matching Even though the results of the pattern recognition engine 105 seem to have a relatively high probability of matching. The word-based non-fuzzy engine 107 eliminates some of the less likely candidate words so that the user is not bothered by a large selection list. Alternatively, the word-based non-fuzzy engine may select a most likely word from the candidate words. In an embodiment, if the output of the word-based non-fuzzy engine 107 has blur, a phrase-based non-fuzzy engine 11 3 further compares the phrase list 115 to check the result. The list can include double hyphens, triple hyphens, and so on. One or more previously identified 14 200538969 words can be combined with the current word to match the phrase. The frequency of use of the phrase can be used to modify the match to create the candidate phrase and match 11 7 The associated month is ambiguous, the phrase based on the phrase-based non-fuzzy engine and the phrase list 115 are predicted to continue in an embodiment. If the # output according to the phrase has ambiguous, then Perform contextual and / or unlikely words / phrases. If the ambiguity cannot be resolved by driving, this option can be presented to the use option 121. After the user selects the word, the phrase list 1 1 5 can be updated to upgrade the user's choice: Add a new word / phrase to the list. Fig. 2 is a block diagram of an identification processing system according to the present invention. Although FIG. 2 illustrates the various elements of the system, it is understood that a mouthpiece system in accordance with the present invention may generally include elements compared to those described in FIG. 2. For example, some systems may not have a nest of components needed to process sound. • Other functions that the system may describe, such as a mobile phone subscription environment. Figures illustrate various feature secrets related to the present invention. In this specification, a skilled artisan will not limit the configuration of the domain / data processing system to the second configuration. The display 203 passes through the appropriate interface circuit. Possibility of choosing words Even if there is no. To use based on previously recognized words. Method analysis of fuzzy engine 113 11 9 Eliminates non-fuzzy automatic language for the user to select a list of words 109 and P-words / phrases and / or input data input by the user. Model data processing is a data section of the embodiment. Or less vowel recognition capability without the uncommunicated circuit in Figure 2. The various meta-relevant meta-analysis solutions are based on the specific structure described in the present invention to the processor 201. 15 200538969 A handwriting input device 202, such as a touch screen, a mouse, or a digital pen 'is connected to the processor 201 to receive user input for handwriting recognition and / or other user input. A speech input device such as a microphone is connected to the processor 201 to receive user input for speech recognition and / or other speech input. Alternatively, a sound output device 205 is connected to the processor as well. The processor 201 receives input from the voice input device 204 and manages output to the display and the remaining eight. The processor 2 0 1 is connected to a memory 2 1 0. The memory includes a combination of temporary storage media such as random access memory (RAM) and a combination of permanent storage media such as read-only memory (ROM), floppy disks, hard disks, or CD-ROMs. The memory 210 contains all software routines and data needed to manage system operations. The memory typically contains an operating system 2m and an application program 220. Examples of applications include word processors, software dictionaries, and foreign language translators. Speech synthesis software is also available as an application. Preferably, the memory further includes a stroke / character recognition engine = Γ for recognizing strokes / characters and / or phoneme recognition in the handwriting turn to identify phonemes in the speech input. The phoneme recognition engine and the stroke / character recognition engine allow any technique known in the field to provide-candidate lists and associations of strokes that match each input =. It has been known that the front-end engine such as the stroke / character two; engine or the element recognition engine 213 for pattern recognition is not appropriate in the present invention. η In an embodiment of the present invention, 'the memory 210 further includes a 16 200538969 language non-fuzzy backend, which may be 6 "" dagger one or more word-based non-puzzling engines 216, and Phrase Rabbit Aunt VII, Non-fuzzy Engine Based on Identification 2 1 7, Non-modelled Pregnancy Based on Precedence 2 1 8, Choice Module 2 1 9 and Others such as Word List 2 1 4 And a y piece of my list 2 1 5 and so on. In this embodiment, the context-based indecent assault fuzzy engine application is helpful to input non-ambiguity user's other postulates' strategies. For example, depending on the selected user location, the user is in the office or at home; the time of day, such as working hours or leisure years, sighs, or recipients, and so on. In a prudent embodiment of the present invention, the components used for a non-fuzzy back end are mostly used in different input forms, such as for handwriting recognition and for speech recognition. The word list 2〗 4 1 4 contains a list of known words in a language. The word list 214 may end in one step—the person ^ step contains frequency usage information of the corresponding word in the language. In one embodiment, the frequency of a word that does not exist in the word list 214 of the language is seen to be zero. Alternatively, a very small frequency of use can be assigned to an unknown word "J °" using the default frequency of use of the unknown word, and the known and unknown word can be processed in a substantially the same way. The word The word list 2 1 4 can be used for Lugu, ~ word-based non-fuzzy engine 2 1 6 and used to arrange, eliminate β > / or choose to recognize the front end according to the style (such as the stroke / character recognition The candidate word determined by the result of engine # 212 or the phoneme recognition engine 213), ΒΒ7 ~, to predict the word based on a part of the user input in order to form the word. None 7 n Ί Similarly, the film The phrase list 2 1 5 may include a y Han ^ including two or more words; a list of shi films ^ and the frequency of use information, and the phrase list may be used by the phrase-based non-fuzzy engine 2 1 7 Used and can be used to predict words to complete the phrase. 17 200538969 In one embodiment of the present invention, each input sequence is processed by referring to a plurality of word modules, each word module contains a Or multiple vocabularies and information about each vocabulary , Including the number of characters in the word to "the word occurrence frequency and the other terms on the same length. Alternatively, information about the vocabulary module or a module in which a specific word is a member is stored with each word, or a module can modify or create a word according to the language style, such as on a specific syllable Put in a distinguishing mark, or build or filter candidate terms based on any other algorithms and / or nearby contexts used to interpret the current input sequence. In one embodiment, each input sequence is processed by a pattern recognition terminal to provide a series of candidate lists, such as pen day, character, syllable, phoneme, and so on. Different combinations of this candidate provide different candidate terms. The non-fuzzy backend combines the candidate's compliance probability and the frequency of use of the candidate word to rank, eliminate, and / or select one or more words as alternatives for the user to choose. Words with a higher frequency of use are highly likely candidates. Unknown words or words that are used less frequently are candidates for low probability. The selection module 2 1 9 selectively presents a number of southerly possibilities from the user-selectable person. In another embodiment of the present invention, the frequency of use of words is based on the use of the user or the use of the word in a specific context, such as in a message or article that the user is editing. As a result, frequently used words become more likely. _ In another embodiment, a word is stored in each vocabulary module, so the word is classified as a file or cluster containing words of the same length. First, each input sequence is processed by searching for groups of words of the same length as the number of inputs in the input sequence, and the selected word is identified with the best coincidence score. If the number of recognized words with the same length as the input sequence is less than a critical number, the system continues to compare the input of N input with the first 1 ^ of each word in the word group of length N + 1 letter. Li continuously searches for longer and longer words and compares the input sequence entered with the first N letters of each word in the group until a critical number of words are identified. Available candidate words longer than the input sequence can be provided to the user as a possible interpretation of the input sequence, which provides a one-word completion form. f During the installation phase, or in the process of receiving text messages or other information, the method of finding this information building case in the data file has been added to the words in the vocabulary. With 曰 一发 ^ 1 exists in the know-how. After searching for α, the day m history breaks into a sub-module as a low-frequency word, and is therefore placed at the end of the list of words associated with the chanting f. During the bet 8¾ process * a specific number of times that a new word was detected, a relatively higher degree was specified by upgrading the scab h subword in the sub-Θ related list, thus increasing information privacy, ^ poor The likelihood that the word will appear in the word list during a long rotation. In the present " 明 之 _-%, sample, for each input sequence, a group of words is formed by identifying a candidate word component that is right and fearing the possibility, and compiling the selected word component ^^ 7 ~ Sub-words to build a candidate word. This "peak type" word is then included in the list of candidate words, and can also be presented in the other place. The word vocabulary has an appendix to the offensive word, which is similar to the offensive word in the state, so when entering the offensive word, the candidate candidates are sorted into order. The rate word is selected by the candidate category one, which is 19 200538969.

Making the exact input of the text include the offensive word will only produce relevant acceptable words in the exact type of stop and, where appropriate, as a suggestion in the word selection list. This feature filters out the appearance of offensive words, which may happen by accident when the user learns that it may be possible to type faster without inadvertently touching the exact position of the expected letter of the keyboard. Therefore, before displaying the exact typed string, use a technique well known in the art, and the software routine responsible for displaying the word selection list compares the current exact typed string and the offending word appendix. If the two are found to match, then Replace the display string with relevant acceptable words. Otherwise, even if an offensive word is regarded as a very low frequency word, when each letter of the word is directly touched, it will still be displayed as the exact typed word. Even though this situation is similar to accidentally typing an offensive word on a standard keyboard, the present invention tolerates less accurate input by the user. This feature can be turned on or off by the user, for example through a system menu option. Those skilled in the art will understand that additional vocabulary modules can be opened in the computer, such as vocabulary modules containing legal terms, medical terms, and other languages. Furthermore, in certain languages such as Hindi, the vocabulary module may use a "template" of a valid sub-word sequence to determine which candidate word component is the previous input and the candidate word is being considered Possible or appropriate. Through a system menu, the user can set up the system so that the additional vocabulary words appear first or last in the list of possible words, such as by special coloring or highlighting, or by which system the vocabulary can be automatically based on The module supplies the directly selected word and automatically switches the order of the words. Therefore, in the scope of the additional application patents, it will be understood that the present invention can be implemented in ways other than those specifically described herein. According to another aspect of the present invention, in the process of using the system by a user, an upgrade algorithm automatically adjusts the vocabulary, and the algorithm is executed every time when the use _select_word Xiao is executed to gradually increase the transparency. Related to the word :: Frequency to upgrade words in that vocabulary. & In one embodiment, the upgrade algorithm increases the frequency values associated with a relatively large number of selected words ' and decreases the frequency values of those words ignored by very small deductions. For a vocabulary module where the relative frequency information is indicated by the consecutive order in which the words appear in a list, the upgrade is completed by moving the selected word upward from the distance between a part and the front end of the list. The upgrade algorithm is best to avoid moving the most frequently used words and very rarely used words away from their original location. For example, the percentage of words in the middle range of the list will increase with each selection. The words between the start and end of the selected word in the language upgrade are effectively downgraded by the value 丨. The list of terms is conserved as a whole, so information about the relative frequency of the terms in the list can be maintained and updated without the need to increase the storage required for the list. The upgrade algorithm increases the frequency of selected words and reduces the frequency of unselected words by 1 as appropriate. For example, 'In the phrase where the relative frequency information is indicated by the order in which the words appear in the-list, the selected word appearing at the IDX position in the list is moved to the (Ιβχ / 2) . Correspondingly, words in the list (position 1_ down to (lDx + i) are moved down one position in the list. When a series of contact points are processed and a word is selected based on the calculated Match the metric score established, and-or more words appear in the list before 21 200538969 of the word selected by the user, the words in the list are downgraded. In the selection above However, words that are not selected can be presumed to be referred to • High frequency, that is, they appear too high in the list. This word of starting position can be downgraded, for example, to (position. Therefore, a The more often a word is considered for selection, the less the number of hierarchies it is moved in. The upgrade and drop are triggered by one of the user ’s actions, or may be performed differently depending on the entry. For example, only There should be

• When the stylus or mouse clicks or drags the expected word to the front of the word, the word appearing above the word in the selection list will be downgraded. Alternatively, a selection word at an upper position in the selection list may have a larger coefficient. For example, the upgrade term starts at IDX (IDX / 3). The change is obvious to those skilled in the art. According to another aspect of the invention, the front end can detect changes in its cognition based on feedback from the back end. As the word is added and the word is selected from the selection list, the candidate order and the expected words contained in each selected word change the possibilities established by the front end. Alternatively, the front end then receives adjustment values for one or more strokes, characters, syllables. Figures 3A and 3B illustrate an example of handwritten recognition output according to the present invention. According to an embodiment of the present invention, the number of degraded places that appear in the selection list is inappropriate to appear in IDX (IDX * 2 + 1), and the level processing may only use the one-touch word selection list of the user ’s loser. The word expected by the Chinese is manually dragged and dropped to be upgraded to a more general position and moved to speech. Many of these system errors and the user's repeated input of the non-components between the components can be used to maintain a self-identity of the self or phoneme Software's non-modular handwriting recognition engine 22 200538969 and -Module 'This module obtains from the handwriting engine all possible correspondences related to each letter entered by the user. This embodiment combines these possibilities with the Word likelihood to predict the most likely word for the user or the word the user is trying to take turns. Any technique known in the art can be used to determine the τ energy sign and the possibility of compliance. For example, ‘Yes, a user may try to enter five characters to enter the five-letter word“ often ”. This user input can appear as in circle 3α

Illustrated. The handwriting recognition software specifies the following characters and the possible output of the characters of the pen day: pen day 1 (301): 0, 60%, a · 24%, V 12%, ,, pen day 2 (302) :, T, 40%, T 34%, 4, 20%, τ 6% pen day 3 (303): V 50%, f, 42%, Γ 4%, i, 4% pen day 4 ( 304): * c, 40%, e * 32% ,, s, 5%, 13% pen 5 (305) :, n, 42% ,,!: 30%, m, 16%, v 12% For example, the probability of stroke 301 is 60%, the probability of stroke 302 is 40%, the probability of stroke 303 is 4 ', the probability of stroke 304 is' c' The probability is 40%, and the probability of pen day 305 is 4, 2%. The letters that the handwriting recognition software considers to be closest to the strokes of the user are grouped together, and the handwriting software module presents the string '0Ucn' to the user, who is not the user's intended input. It is not even a word in English. An embodiment of the present invention uses a non-fuzzy word search module to find an optimal prediction based on these characters, the possibility of matching the character, and the frequency of using the word in English. In one embodiment of the present invention, the combined handwriting module and the non-fuzzy module predict that the most likely word is 'often', which is the word the user is trying to enter.

For example, as shown in Figure 3B, a back-end tool receives all candidates and determines a list of possible words including: ottcn, attcn, ftcn, aftcn, otfcn, atfcn, offcn, affcn, otten, atten, often, aften, otfen, atfen, offen, affen, otter, attcr, oftcr, after, otfer, atfer, offer, affer, otter, atter, ofter, after, otfer, atfer, offer, affer, etc. The possible word can be formed by selecting the character with the highest matching possibility to the lowest matching possibility from the front end judgment. When one or more highly probable words are found, the less probable characters can be left out. To simplify the description, in Figure 3A, it is assumed that the frequency of use of unknown words is 0, and the frequency of use of known words such as often, after and offer is 1. In Fig. A, a coincidence indicator of a candidate word is calculated from the result of the frequency of use and the possibility of coincidence of the candidate word used in the word. For example, in Fig. 3A, the coincidence possibilities of the characters' 0 ,, 'f ,,' t ,, 'e, and' η, "are 0.6, 0.34, 0.5, 0.32, 0.42, The word 'often' is used frequently. Therefore, an indicator matching the word "often" was determined to be 0.0137. Similarly, the indicators for the words "after," and "offer" are 0.0039 and 0.0082, respectively. When the back-end tool selects the most likely word, it will select "often". Note the "indicator" for that word , Can be normalized to sort the candidate terms. In one embodiment of the present invention, one or more inputs are explicit, that is, related to a single stroke, character, syllable, or phoneme, so the probability of meeting each character, etc. is equal to 100%. In another embodiment of the present invention, a 24 200538969 explicit input generates a specific value set from the recognition front end, which makes the non-fuzzy back end only pair the exact character and so on in the corresponding position of each candidate word. In another embodiment of the present invention, the reserved numbers, appropriate diacritics, and accent marks and / or other defining symbols are explicitly entered, and punctuation marks are reserved within and between words.

Figures 4A-4C show a scheme for handwriting recognition on a user interface according to the present invention. As shown in FIG. 4A, the device 401 includes an area for a user to write the handwriting input 407. An area 403 is provided to display the message or article that the user is typing, such as on a web browser, on a note-taking software program, on an email program, and so on. The device includes a touch screen area for writing by the user. ^, 丨, you re-buried the user ’s handwriting input 407 shift This device provides a candidate word list in area 409 for the user

Select. The candidate terms are sorted by the probability of I remaining σ. The device can optionally present the first few candidate words that you can choose from. The user can use a conventional method to select a bullet from the list, or use a numeric key corresponding to the position of the word. Alternatively, the user can select a voice command to select a word, such as selecting a word by uttering the word a or corresponding to the word position number in the list. In the better case,% y t is presented in d trench 1 'the most likely word is automatically selected and presented in & field 403. By saying this, if the user accepts the candidate word ‘If you start writing the result by mistake, the user will choose a wide selection’, and the user will not need to choose. If; candidate replaces the self-word, the device selects < possible word that the user selects as high = select candidate. In another embodiment, the ... and 4 砚 are used as the preset value, indicating that the user g 25 200538969 selects a word that will be output or extended by a subsequent action, and a specified input changes the height Brightness to another candidate word. In another embodiment, a specified input selects a syllable or word for modification or re-entry from a multi-syllable sequence or multi-word phrase that has been entered or predicted.

Figure 4C illustrates a situation where contextual and / or grammatical analysis further assists in resolving one of the ambiguities. For example, in Figure 4C, the user has entered the word "It is an". In terms of grammatical analysis, the device predicts the continuation word as a noun. Therefore, the device further adjusts the order of the candidate words to promote candidate words belonging to a noun. Therefore, the most likely word becomes "offer" instead of "often". However, since an adjective may also be between the noun and the word "an", the device will still present other options for the user to choose, such as "often" and "after". Fig. 5 is a flowchart illustrating a process of user input according to the present invention. At step 501, the system receives handwritten input of a word. A step 503 is followed to establish a candidate character list that may match each character in the handwriting of the word. Step 505 determines a candidate word list from the candidate character list. Step 507 combines the frequency indicator of the candidate word and the possibility of matching the candidate character to determine the possibility of matching the candidate word. Step 509 deletes a part of the candidate words according to the possibility of matching the candidate words. Step 5 1 1 presents one or more candidate words for the user to select. Although Figure 5 illustrates a flowchart for processing handwritten input, it can be understood from this description that speech input can also be processed in a similar manner, in which a speech recognition module creates candidate phonemes for each phoneme in the word. 26 200538969 The high error rate of the recognition system facing the more memory and computer systems and the need for constant low. The relevant report by the knot recognition engine of one embodiment of the present invention is reasonable and can use these phonemes to form white Dynamically revise the speech recognition turn-out. In one embodiment of the present invention, each time an input is received, it is displayed to the user on the display. The candidate words are presented in the order determined by sex. The most likely words will appear. The input of the present explanation of the sequence will start—the new input sequence is in another aspect of the present invention—the indicator is preferably located repeatedly on the text candidate word based on the coincidence metric to start a special Words are presented in the order determined by the specified selection possibility. An input sequence will also refer to and effectively select one of the sequences. The end result and the subsequent data according to the present invention a mixed system of letters, etc. syllable phoneme performed

Command recognition speech recognition technology even solves the problem. In addition, due to today's efforts to correct speech, its adoption is based on the known possibility of using a set of candidate phonemes and monolingual capabilities as well as the use of these output after-words. The department 0 'is presented in a word selection list on candidate words that match the input sequence. Calculating the coincidence of each candidate word is therefore considered to be the top of the list based on the match metric. Selecting this input ends an input sequence and therefore the subsequent columns. Only one candidate word appears at the insertion point where the display is being created. The displayed ones are considered the most likely. With optional input, the user can replace the displayed one or more activations of the displayed selected input with the matching alternative candidate words to present an explanation for the system to actually input a new input sequence. The system starts with a composition level such as strokes, style recognition, such as handwriting recognition, speech 27

200538969 Recognition, etc. to provide ambiguous results and related compliance possibilities Then perform non-fuzzy operations on internal component levels such as words, phrases, word pairs, triplets, etc. The language used by the system to resolve ambiguity can be the frequency of any word in the language, the frequency of the individual words used, the possible phonetic portion of the input word, the language state, and the word being entered Context, ligatures (word pairs) or strings, and any other language or context that can be used to resolve the ambiguity. The invention can be used with alphabetic languages, such as English and Spanish, where the output of the handwriting recognition front end is a letter or pen day and its correlation. A hand-written non-fuzzy operation of a one-letter language can be performed at the word level, where each word lexically includes multiple letters. The present invention can also be used with semantic language, such as the output of the handwriting recognition front end in Chinese and Japanese as pen day and its related possibilities. The handwritten non-fuzzy operation of Italian language can be performed at the root / component or letter level. This non-fuzzy operation can be further operated at a higher level, such as a slice of a two-string, a triple-string, and so on. Furthermore, the grammatical structure of the language is used for the non-fuzzy operation to select the best overall fit of the input. The present invention can also be accompanied by the phonetic or letter expression of the semantic language so that the non-fuzzy operation can be performed on syllables, semantic letters, words, and / or phrases. Similarly, the present invention can also be used for speech recognition, where the output of the speech front end contains phonemes and their associated coincidence possibilities. The candidate can be combined for selection of a word, phrase, two-character string, three-character string, and a string of words that makes it a type of trigram, may be executed, its first-level verbs, Also available. Hierarchical Phonetic Recognition Phoneme String or 28 200538969 One of the idioms best matches.

An embodiment of the present invention is also completed when the user has input only a few predicted time of day. For example, after successfully identifying the first few letters of a word with a high probability, the system can provide a list of words at the rear, where the first few letters are the same as the matching letter. A user can select a word from the list to complete the entry. Alternatively, an indication that one of the words in the list is close to some may prompt the user to be displayed by applying the specified input to one of the list inputs based on the completion of the word; the subsequent pop-up word list display contains The word is a finite word and can indicate further completions in order. Each of the first few characters may have only one high probability candidate, which is used to select the to-be-completed word list. Alternatively, one or more of the first characters may contain ambiguity, so several high-probability combinations of the first few characters may be used to select the to-be-completed word list. The word list for completion can be sorted and displayed according to the likelihood that the word the user is trying to enter. For example, the word used for completion may be based on how often the word is used by the user in the language, in the article the user is editing, in a specific context, such as a dialog box, and / Or sorted by frequency of occurrence in phrases, digraphs, trigrams, idioms, etc. When one or more words in a phrase, double-liga, triple-liga, or idiomatic phrase, etc., immediately precede the word being processed, these phrases, double-liga, triple-hybrid The occurrence frequency of a string or idiom can be further combined with the frequency of the word when determining the ordering of the word to be completed. Words that are not in any of the currently known phrases, double ligatures, triple ligaments, idioms, etc. are considered in unlicensed houses with a very low frequency of occurrence 29 200538969. Similarly, a word that is not in the list of known words is treated as an unknown phrase with a very low frequency of occurrence. Therefore • The input of any word or the first part of a word can be processed to determine the possible input. In an embodiment of the present invention, the back end continuously obtains a candidate list of each word, stroke, and phoneme recognized by the style front end, and uses the list to sort the words to be completed. As the user provides more input, less likely words about completion are eliminated. Used for completion • The word list is reduced in size as the user provides more input until it does not exist or the user selects a word from the list. Furthermore, before the front end of the pattern recognition provides the first word of the continuation word to enter a candidate list, the back end has known phrases, two-character strings, three-character strings, Identifiers, etc. are used to determine the words to be determined, so as to determine a list of words to be completed, such as a phrase, a double ligature, a triple ligament, a idiom. Therefore, the present invention also determines the complete continuation word based on the most entered word of the user. In one embodiment of the present invention, the back end uses any stroke, word 70, syllable or Wild card of phonemes. The list of to-be-completed words according to a part of the input can be regarded as an example of using a wild card for one or more pen characters or phonemes to be input or to be received from the pattern recognition front end. • In the embodiment of the present invention, the front end may not recognize the picture, character or phoneme. The front crawl does not stop the input processing to force the user to re-type the input. On the contrary, the front end can tolerate the resulting merged words, and any of the most updated and updated words with fuzzy input and JL > Wordable user day, pen, messenger, teleport 30

200538969 A wild card stuck to the back end. At a high level, the back end can resolve the ambiguity without forcing the user to retype the input. This greatly enhances the user-friendliness of the department. In one embodiment of the invention, the back end automatically replaces one or more inputs from the front end with a wild card. For example, when no possible words are found from a known word list, the backend can replace the most ambiguous input with a wild card to expand the candidate combination. For example, a list with a large number of low-availability candidates can be replaced by a wild card. In an embodiment, the candidate list is provided by the month J terminal, and thus the probability that the input meets one of the lists will be higher than a critical value. Therefore, a fuzzy input has a low probability of candidates. In other embodiments, the front end provides a list, so that each candidate is more likely to meet the input than one, so it is very unlikely that a fuzzy input is one of the candidates. ‘The system implements wild cards, such as pen pens that are suitable for any letter, giving the same possibility, so it can handle the situation where no valid word is found when the wild card is not used. In one embodiment of the present invention, the backend establishes the same candidate from the combination of candidate strokes, characters, or phonemes provided by ^ ίΙίΤ irtjf J ~; &, difficult words, for example That is, the candidate characters for each character input can be sorted according to the likelihood of being input. The establishment of the candidate word begins to extend to characters that are less likely to match. When the candidate iP + «χ ~, and prepositions are found in the list of known terms, it is less likely to match the element and may not be used to build further candidate terms. In one implementation ', the system displays the most likely words or a list of all candidate words sorted by the calculated probability X. The system can automatically add a general word that should be able to wait for a large number of cases. There are examples of non-compliant numbers that can be added 31 200538969 to help users. This includes, for example, auto accented characters, auto and punctuation, and definition symbols. One aspect of the present invention provides a language back end to be used simultaneously in forms such as speech recognition, handwriting recognition, hard keyboard or touch keyboard input. In another embodiment of the present invention, a language is used to defuzzify the candidate word. After combining a back-end component from the selected input to determine candidate words and their coincidence possibilities, one is used to rank the candidate words according to language characteristics. For example, the second paragraph of B further uses the frequency of words such as the user in the language, an article being edited, an introductory text that requires the input, and candidate words derived from the back-end component. Deblur the candidate. The language backend can also perform a deblurring operation based on strings, triplets, phrases, and so on. The language backend can perform operations based on the context, grammatical structure, and so on. Since the tasks performed by the language backend are the same for various methods such as speech recognition, handwriting recognition, or input using a hard keyboard or touch keyboard, the language backend can be shared among Φ forms. In one embodiment of the present invention, 'a language serves multiple input forms, so when a user combines different to provide an input, only a single language backend is required to support mixed types. In another aspect of the present invention, 'from a specific front end' is considered as a clear candidate word component, and if it is not 100% of the recorded character, it is one of the clear pen and word that the back end will use. Yuanhehe contains its finite words in the corresponding position. The invention also includes capitalization so that the back-end is used as the front-end language back-end on multiple input-control screens. The language, the user, etc. use the conformable one or two ligatures again, and the input side control of the de-ambiguity operation. Multiple input terminals on the screen simultaneously input the form and the input mode. Each input possibility syllable contains a mixture of 32 200538969. First, it uses the candidate set from _ or multiple recognition systems and the sex. Use some of the known characteristics of to solve the betel in. Resolving the ambiguity in the handwriting / speech recognition can improve the recognition rate of the system to promote user friendliness

Although accompanied by the preferred person here, other applications should be clearly understood without departing from the spirit of the invention and the scope of patent applications contained in the following examples to illustrate the invention. Programs for skilled artisans can replace the ones proposed here Or range. Therefore, the present invention should only be limited. [Schematic description] Figure 1 illustrates a system for identifying user input on a data processing system according to the present invention. Figure 2 is for identifying a user according to the present invention. A block diagram of a data processing system in turn; Figures 3A and 3B illustrate an example of non-fuzzy output of a handwriting recognition software according to the present invention;

Figures 4A-4C illustrate a handwriting recognition scheme on a user interface according to the present invention; and Figure 5A is a flowchart of processing user input according to the present invention. [Simple description of component representative symbols] 1 0 1 language input 103 style recognition engine 105, 111 input 33 200538969 107, 113 non-fuzzy engine 109 word list 11 5 phrase list 11 7 match 11 9 analysis 1 2 1 user selection 201 processor

202 handwriting input device 203 display 204 voice input device 205 sound output device 2 1 0 memory 2 11 operating system 212 day / character recognition engine 213 phoneme recognition engine 2 1 4 word list 2 1 5 phrase list 2 1 6 Word-based non-fuzzy engine 2 1 7 Phrase-based non-fuzzy engine 2 1 8 Pre- and post-based non-fuzzy engine 2 1 9 Selection module 220 Application 401 Device 403, 405, 409 Area 407 Handwriting input 501, 503, 505, 507, 509, 5 1 1 Step 34

Claims (1)

  1. 200538969 Scope of patent application: 1. A method for identifying language input in a data processing system, which includes at least the following steps:
    A user input of a word in a language is processed through pattern recognition to generate a plurality of recognition results for a plurality of word components, respectively. At least one of the plurality of recognition results includes a plurality of candidate word components and a plurality of possible components. Sexual indicators, the plurality of likelihood indicators indicate the degree of likelihood that the plurality of word components and a part of the user input are consistent with each other; and from the plurality of recognition results and the use of a list of possible words Sexual data determines one or more candidate words for that word that are available to the user. 2. The method according to item 1 of the scope of patent application, wherein the pattern recognition includes handwriting recognition. 3. The method according to item 2 of the scope of patent application, wherein each of the plurality of candidate word components includes a stroke of day: and the word includes a semantic language symbol. 35
    200538969 4. The method described in item 2 of the scope of patent application, wherein each of the plurality of word components contains a character; and the word includes a sentence 5. The method described in item 1 of the scope of patent application, Wherein the pattern contains speech recognition; and each of the plurality of candidate word components is 0. 6. The method according to item 1 of the scope of patent application, wherein one of the plurality of recognition results of a word includes an indication, It indicates that any one of the candidate component sets has the same possibility of matching one of the uses of the word; and the alphabetic characters contained in the candidate word component set. 7. The method as described in item 1 of the scope of patent application, wherein the information indicating the possibility of using the table includes at least one of the following: the frequency of using words in the language; the frequency of using words by users; and How often words are used in the document. 8. The method as described in item 1 of the scope of patent application, wherein the data indicating the possibility of using the table includes at least one of the following: phrases in the language; word pairs in the language; and candidate words . Recognize selected words that contain a single tone. The user enters the language. Word string Word string 36 200538969 The triplet in the language. 9. The method according to item 1 of the scope of patent application, wherein the data indicating the possibility of using the word list includes at least one of the following: data indicating the form of the language; and data indicating the grammar rules of the language.
    10. The method as described in item 1 of the scope of patent application, wherein the data indicating the possibility of using the word list at least includes: a contextual data indicating that the user who received the word has input. 1 1 The method as described in item 1 of the scope of patent application, wherein the user input specifies only a part of a complete set of word components of the word.
    1 2. The method according to item 1 of the scope of patent application, wherein the one or more candidate words include a part of words formed by a combination of candidate word components in the plurality of recognition results and the plurality of candidate words Part of a combination of candidate word components in the recognition result. 1 3 · The method according to item 1 of the scope of patent application, wherein the one or more candidate words include a plurality of candidate words; and the method further includes the following steps: presenting the plurality of candidate words for selection; And 37 200538969 receives a user input to select one of the plurality of candidate words. 1 4 · The method described in item 13 of the scope of patent application, further comprising the following steps: Predicting one or more candidate words based on one of the next word input expected by a user.
    1 5. The method as described in item 13 of the scope of patent application, wherein the plurality of candidate words are presented in the order of the possibility of user input consistent with the word 0 16. as described in item 1 of the scope of patent application The method further includes the following steps: automatically selecting a most likely one from among one or more candidate words as a recognition word input by a user of the word; according to the expected next word input by a user Most likely to predict one or more candidate terms. 1 7 · The method according to item 1 of the scope of patent application, further comprising the following steps: automatically accenting one or more characters; automatically capitalizing one or more characters; 38 200538969 automatically adding one or more characters Punctuation marks; and automatically add one or more definition symbols. 1 8 · The method as described in item 1 of the scope of patent application, wherein each of the plurality of recognition results includes a plurality of likelihood indicators related to a plurality of candidate word components, respectively, so as to indicate that it meets a part of the user input Relative possibilities.
    19. A machine-readable medium having instruction data, when the machine-readable medium is executed on a data processing system, it will cause the system to execute a method for recognizing language input, the method including at least the following step:
    Processing a user input of a word in a language by performing pattern recognition to generate a plurality of recognition results for a plurality of word components, respectively, at least one of the plurality of recognition results including a plurality of candidate word components and a plurality of Likelihood indicators, the plurality of likelihood indicators indicating the degree of likelihood that the plurality of word components and a portion of the user input are consistent with each other; and from the plurality of recognition results and a list of words that can be pointed out Useability data to determine one or more candidate words for the word entered by the user. 2 0. The media as described in item 19 of the scope of patent application, wherein the one or more 39
    200538969 The candidate word includes a plurality of candidate words; and the method further includes the steps of: presenting the plurality of candidate words for selection; receiving a user input to select one of the plurality of candidate words; and a basis A user is expected to input one of the next word selection to predict one or more candidate words. 21. The medium according to item 19 of the scope of patent application, wherein the method includes the following steps: automatically selecting a recognition word from one or more candidate words that is most likely to be entered by a user of the word; One or more candidate words are predicted based on the most probable next-word input expected by a user. 22.-A data processing system for identifying language input, including at least a processing component, which is used to process a user input for processing a word through pattern recognition to generate a plurality of words respectively. A recognition result, at least one of the plurality of recognition results including a plurality of candidate word components and a plurality of likelihood indicators, and the plurality of likelihood indicators indicate a possibility that the plurality of word components and a part of the use are consistent with each other Degree of sexuality; and a judging component, which is used to select from the plurality of identification results and the next step, which includes the more capable ones, including: Language I One of the productions.
    200538969 Use the data of a word list to determine the candidate word of the word or one entered by the supplier. 23. The data processing system as described in item 22 of the scope of patent application, wherein one or more candidate words include a plurality of candidate words; and the system includes: a presentation component for presenting the plurality of candidate word selections A receiving component, which is used for receiving a user input to select one of the candidate words; and among them, the plurality of candidate words are present with the possibility of user input matching the word. 24. The data processing system described in item 22 of the scope of the patent application, each of the plurality of identification results includes a plurality of likelihood indicators that are respectively related to a plurality of word components to indicate a relative possibility that conforms to a part of the use. 25. The data processing system according to item 22 of the scope of patent application, comprising means for any of the following: automatically accenting one or more characters; automatically capitalizing one or more characters; automatically increasing One or more punctuation marks; and multiple ones in which the system can be selected for selection
    200538969 Automatically add one or more definition symbols. 26. The data processing system described in item 22 of the scope of the patent application, wherein the selection of the plurality of candidate words causes the pattern recognition to adjust the subsequent index of the selected one or more word components of the candidate words. 27. A method for processing language input in a data processing system, which includes at least the following steps: receiving a plurality of recognition results of a plurality of word components, and processing a user input of a word in a language, At least one of the plurality of identifications includes a plurality of candidate word components and a plurality of likelihood indicators, the plurality of likelihood indicators indicating a degree of sexuality in which the plurality of word points and a portion of the user input are consistent with each other; And from the plurality of recognition results and data that can indicate the possibility of using a word list, determine one or more candidate words that can be input by the user. 28. The method as described in item 27 of the scope of patent application, wherein the candidate component includes at least any of the following: a stroke derived from handwriting recognition, speech recognition, or keyboard input 1 a stroke derived from handwriting recognition, speech recognition, or keyboard input The reversible party in Zitian made it possible to process several words to make the word 42 200538969 a phoneme derived from handwriting recognition, speech recognition, or keyboard input; and from handwriting recognition, speech recognition, or keyboard One syllable for input or one syllable for other speech expressions. 2 9. The method according to item 27 of the scope of patent application, wherein the above-mentioned language is any one of alphabetic or semantic.
    30. The method as described in item 27 of the scope of patent application, wherein the step of determining one or more candidate words further comprises the following steps: eliminating the plurality of candidate word component combinations of the plurality of recognition results. 3 1. The method as described in item 30 of the scope of patent application, wherein the step of determining one or more candidate words further includes the following steps: selecting a plurality of candidate words from a word list in the language, and the plural number Each candidate word contains a combination of candidate word components in the plurality of recognition results. 32. The method according to item 31 of the scope of patent application, further comprising the following steps: determining one or more candidate words from the plurality of recognition results and data indicating the possibility of using a word list; Multiple likelihood indicators to indicate the likelihood of user input matching the term. 43 200538969 3 3 · The method described in item 32 of the scope of patent application, further comprising the following steps: Sort the one or more candidate words according to the one or more possibility indicators. 34. The method according to item 33 of the patent application scope, further comprising the following steps:
    Automatically select a word from the one or more candidate words. 35. The method of claim 34, wherein the step of automatically selecting is performed on any of the following: a phrase in the language; a word pair in the language; and three in the language Hyphenation. 36. The method described in item 34 of the scope of patent application, wherein the step of automatically selecting is performed according to any of the following: the form of the language; and the grammatical rules of the language. 3 7. The method as described in item 34 of the scope of patent application, wherein the step of automatically selecting is performed based on the context input by the user who received the word 44 200538969. 38. The method according to item 34 of the scope of patent application, further comprising the following steps: predicting a plurality of candidate words according to a word automatically selected by a user who is expected to input the next word. 39. The method according to item 33 of the scope of patent application, further comprising the following
    Steps: presenting the one or more candidate words for a user to select; and receiving a user input to select a word from the plurality of candidate words. 40. The method of claim 39, wherein the plurality of candidate words are presented according to the order of the one or more likelihood indicators. 4 1 · The method as described in item 39 of the scope of patent application, further comprising the following steps: predicting a plurality of candidate words according to the expected next-word input by a user. 42. The method as described in item 27 of the scope of patent application, wherein one of the plurality of recognition results of a word component includes a prediction indicating that any one of the candidate words in the set of 45 200538969 points has a matching word The use of words is part of the same possibility. 43. The method as described in item 27 of the scope of patent application, wherein the data indicating the possibility of using the list includes any of the following: the frequency of using words in the language; the frequency of using words by users; and How often words are used in the document. Enter the word
    44. The method according to item 27 of the scope of patent application, further comprising any step of: automatically accenting one or more characters; automatically capitalizing one or more characters; automatically adding one or more punctuation marks ; And automatically add one or more definition symbols. 45. — A machine-readable medium with instruction data. When the machine medium is executed on a data processing system, it will make the system a method for identifying language input. The method includes at least steps: Multiple recognition results of word components, a user input of a word in a language, at least one of the plurality of recognitions includes multiple candidate word components, and the following
    May read a number of performance indicators that perform one or more of the following steps to process the results, the plurality of likelihood indicators indicating the degree of likelihood of conformity of the plurality of word components and a portion of the user input relative to each other; and One or more candidate words of the word that can be input by the user are determined from the plurality of recognition results and data indicating the use possibility of the word list.
    46. The medium of claim 45, wherein the step of determining one or more candidate words includes the following steps: eliminating a plurality of candidate word component combinations from a plurality of recognition results; and one of the languages A plurality of candidate words are selected from the word list, and the plurality of candidate words contain a combination of candidate word components in the plurality of recognition results. 47. The media as described in item 46 of the scope of patent application; the method further includes the following steps: judging one or more candidate words from the plurality of recognition results and information indicating the possibility of using a word list One or more likelihood indicators to indicate the likelihood of user input matching the word; sort the one or more candidate words according to the one or more likelihood indicators; automatically from the one or more candidate words Choose one of them; and 47
TW94103440A 2004-02-11 2005-02-03 Handwriting and voice input with automatic correction TW200538969A (en)

Priority Applications (3)

Application Number Priority Date Filing Date Title
US54417004P true 2004-02-11 2004-02-11
US11/043,506 US7319957B2 (en) 2004-02-11 2005-01-25 Handwriting and voice input with automatic correction
US11/043,525 US20050192802A1 (en) 2004-02-11 2005-01-25 Handwriting and voice input with automatic correction

Publications (1)

Publication Number Publication Date
TW200538969A true TW200538969A (en) 2005-12-01

Family

ID=34865026

Family Applications (1)

Application Number Title Priority Date Filing Date
TW94103440A TW200538969A (en) 2004-02-11 2005-02-03 Handwriting and voice input with automatic correction

Country Status (9)

Country Link
EP (1) EP1714234A4 (en)
JP (1) JP2007524949A (en)
KR (1) KR100912753B1 (en)
CN (1) CN1918578B (en)
AU (1) AU2005211782B2 (en)
BR (1) BRPI0507577A (en)
CA (1) CA2556065C (en)
TW (1) TW200538969A (en)
WO (1) WO2005077098A2 (en)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
TWI474227B (en) * 2008-03-11 2015-02-21 Microsoft Corp Interpreting ambiguous inputs on a touch-screen
TWI587281B (en) * 2014-11-07 2017-06-11 Papago Inc Voice control system and its method
TWI616868B (en) * 2014-12-30 2018-03-01 鴻海精密工業股份有限公司 Meeting minutes device and method thereof for automatically creating meeting minutes
TWI619115B (en) * 2014-12-30 2018-03-21 鴻海精密工業股份有限公司 Meeting minutes device and method thereof for automatically creating meeting minutes

Families Citing this family (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2008076812A (en) * 2006-09-22 2008-04-03 Honda Motor Co Ltd Voice recognition device, voice recognition method and voice recognition program
US8032374B2 (en) 2006-12-05 2011-10-04 Electronics And Telecommunications Research Institute Method and apparatus for recognizing continuous speech using search space restriction based on phoneme recognition
KR100908444B1 (en) * 2006-12-05 2009-07-21 한국전자통신연구원 Continuous with the search space limitations on the phoneme recognized based on the speech recognition apparatus and method
WO2010084973A1 (en) 2009-01-20 2010-07-29 日本電気株式会社 Input device, information processing device, input method, and program
JP2011065322A (en) * 2009-09-16 2011-03-31 Konica Minolta Holdings Inc Character recognition system and character recognition program, and voice recognition system and voice recognition program
US8543382B2 (en) * 2010-10-27 2013-09-24 King Abdulaziz City for Science and Technology (KACST) Method and system for diacritizing arabic language text
CN103631802B (en) * 2012-08-24 2015-05-20 腾讯科技(深圳)有限公司 Song information searching method, device and corresponding server
DE102013009375A1 (en) * 2012-12-28 2014-07-03 Volkswagen Aktiengesellschaft Method for entering and recognizing a character string
CN105810197B (en) * 2014-12-30 2019-07-26 联想(北京)有限公司 Method of speech processing, voice processing apparatus and electronic equipment
WO2017013719A1 (en) * 2015-07-17 2017-01-26 楽天株式会社 Character recognition device, character recognition method, and character recognition program
KR101636823B1 (en) * 2015-11-27 2016-07-07 (주)인키움 Server and Method for Providing Personal History Automatically
CN106406807A (en) * 2016-09-19 2017-02-15 北京云知声信息技术有限公司 A method and a device for voice correction of characters

Family Cites Families (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US4003025A (en) * 1975-12-24 1977-01-11 International Business Machines Corporation Alphabetic character word upper/lower case print convention apparatus and method
US5244802A (en) * 1987-11-18 1993-09-14 Phytogen Regeneration of cotton
US5828991A (en) * 1995-06-30 1998-10-27 The Research Foundation Of The State University Of New York Sentence reconstruction using word ambiguity resolution
US5917941A (en) * 1995-08-08 1999-06-29 Apple Computer, Inc. Character segmentation technique with integrated word search for handwriting recognition
US5950160A (en) * 1996-10-31 1999-09-07 Microsoft Corporation Method and system for displaying a variable number of alternative words during speech recognition
US5926566A (en) * 1996-11-15 1999-07-20 Synaptics, Inc. Incremental ideographic character input method
US5896321A (en) * 1997-11-14 1999-04-20 Microsoft Corporation Text completion system for a miniature computer
US20020152075A1 (en) * 2001-04-16 2002-10-17 Shao-Tsu Kung Composite input method

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
TWI474227B (en) * 2008-03-11 2015-02-21 Microsoft Corp Interpreting ambiguous inputs on a touch-screen
TWI609302B (en) * 2008-03-11 2017-12-21 微軟技術授權有限責任公司 Interpreting ambiguous inputs on a touch-screen
TWI587281B (en) * 2014-11-07 2017-06-11 Papago Inc Voice control system and its method
TWI616868B (en) * 2014-12-30 2018-03-01 鴻海精密工業股份有限公司 Meeting minutes device and method thereof for automatically creating meeting minutes
TWI619115B (en) * 2014-12-30 2018-03-21 鴻海精密工業股份有限公司 Meeting minutes device and method thereof for automatically creating meeting minutes

Also Published As

Publication number Publication date
AU2005211782A1 (en) 2005-08-25
AU2005211782B2 (en) 2009-01-22
WO2005077098A3 (en) 2005-11-03
JP2007524949A (en) 2007-08-30
WO2005077098B1 (en) 2005-12-08
KR100912753B1 (en) 2009-08-18
CA2556065C (en) 2012-07-03
CN1918578B (en) 2012-05-02
CN1918578A (en) 2007-02-21
KR20070090075A (en) 2007-09-05
WO2005077098A8 (en) 2007-05-10
CA2556065A1 (en) 2005-08-25
WO2005077098A2 (en) 2005-08-25
EP1714234A4 (en) 2012-03-21
BRPI0507577A (en) 2007-07-03
EP1714234A2 (en) 2006-10-25

Similar Documents

Publication Publication Date Title
Kristensson et al. SHARK2: a large vocabulary shorthand writing system for pen-based computers
US7679534B2 (en) Contextual prediction of user words and user actions
US8074184B2 (en) Modifying electronic documents with recognized content or other associated data
US6307549B1 (en) Reduced keyboard disambiguating system
JP4416643B2 (en) Multimodal input method
US8311796B2 (en) System and method for improving text input in a shorthand-on-keyboard interface
US8441454B2 (en) Virtual keyboard system with automatic correction
US10156981B2 (en) User-centric soft keyboard predictive technologies
TWI394065B (en) Multiple predictions in a reduced keyboard disambiguating system
US6795579B2 (en) Method and apparatus for recognizing handwritten chinese characters
US8583440B2 (en) Apparatus and method for providing visual indication of character ambiguity during text entry
CN1296806C (en) Reduced keyboard disambiguating system
CN105573503B (en) For receiving the method and system of the text input on touch-sensitive display device
US6970599B2 (en) Chinese character handwriting recognition system
US6173253B1 (en) Sentence processing apparatus and method thereof,utilizing dictionaries to interpolate elliptic characters or symbols
US6401065B1 (en) Intelligent keyboard interface with use of human language processing
JP5819924B2 (en) Recognition architecture for generating Asian characters
CN1324436C (en) System and method for improved user input on personal computing devices
EP2133772B1 (en) Device and method incorporating an improved text input mechanism
US9606634B2 (en) Device incorporating improved text input mechanism
US20050262442A1 (en) Correction widget
US20090119582A1 (en) Handheld Electronic Device and Method for Dual-Mode Disambiguation of Text Input
Beaufort et al. A hybrid rule/model-based finite-state framework for normalizing SMS messages
JP4864712B2 (en) Intelligent speech recognition with user interface
US8200865B2 (en) Efficient method and apparatus for text entry based on trigger sequences