Classifications

• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06F—ELECTRIC DIGITAL DATA PROCESSING
  • G06F40/00—Handling natural language data
  • G06F40/10—Text processing
  • G06F40/12—Use of codes for handling textual entities
  • G06F40/126—Character encoding
  • G06F40/129—Handling non-Latin characters, e.g. kana-to-kanji conversion
• • G—PHYSICS
  • G10—MUSICAL INSTRUMENTS; ACOUSTICS
  • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
  • G10L15/00—Speech recognition
  • G10L15/08—Speech classification or search
  • G10L15/18—Speech classification or search using natural language modelling
  • G10L15/183—Speech classification or search using natural language modelling using context dependencies, e.g. language models
  • G10L15/187—Phonemic context, e.g. pronunciation rules, phonotactical constraints or phoneme n-grams
• • G—PHYSICS
  • G10—MUSICAL INSTRUMENTS; ACOUSTICS
  • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
  • G10L15/00—Speech recognition
  • G10L15/06—Creation of reference templates; Training of speech recognition systems, e.g. adaptation to the characteristics of the speaker's voice
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06F—ELECTRIC DIGITAL DATA PROCESSING
  • G06F3/00—Input arrangements for transferring data to be processed into a form capable of being handled by the computer; Output arrangements for transferring data from processing unit to output unit, e.g. interface arrangements
  • G06F3/01—Input arrangements or combined input and output arrangements for interaction between user and computer
  • G06F3/02—Input arrangements using manually operated switches, e.g. using keyboards or dials
  • G06F3/023—Arrangements for converting discrete items of information into a coded form, e.g. arrangements for interpreting keyboard generated codes as alphanumeric codes, operand codes or instruction codes
  • G06F3/0233—Character input methods
  • G06F3/0237—Character input methods using prediction or retrieval techniques
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06F—ELECTRIC DIGITAL DATA PROCESSING
  • G06F40/00—Handling natural language data
  • G06F40/20—Natural language analysis
  • G06F40/205—Parsing
  • G06F40/216—Parsing using statistical methods
• • G—PHYSICS
  • G10—MUSICAL INSTRUMENTS; ACOUSTICS
  • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
  • G10L15/00—Speech recognition
  • G10L15/08—Speech classification or search
  • G10L15/14—Speech classification or search using statistical models, e.g. Hidden Markov Models [HMMs]
• • G—PHYSICS
  • G10—MUSICAL INSTRUMENTS; ACOUSTICS
  • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
  • G10L15/00—Speech recognition
  • G10L15/08—Speech classification or search
  • G10L15/18—Speech classification or search using natural language modelling
  • G10L15/183—Speech classification or search using natural language modelling using context dependencies, e.g. language models
• • G—PHYSICS
  • G10—MUSICAL INSTRUMENTS; ACOUSTICS
  • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
  • G10L15/00—Speech recognition
  • G10L15/08—Speech classification or search
  • G10L15/18—Speech classification or search using natural language modelling
  • G10L15/183—Speech classification or search using natural language modelling using context dependencies, e.g. language models
  • G10L15/19—Grammatical context, e.g. disambiguation of the recognition hypotheses based on word sequence rules
  • G10L15/197—Probabilistic grammars, e.g. word n-grams

Definitions

• Input methods can be employed to convert phonetic strings (reading) into display characters for East Asian languages such as Chinese, Korean, and Japanese, for example, and also process strokes such as in the Traditional Chinese characters.
• Ambiguity exists in conversion due to homonyms and various possible word segmentations.
• An input method tries to solve the ambiguity based on a general (e.g., baseline, default) language model and user input history.
• Adaptation to the user input history can be performed in several ways, for example, short-term memory and long-term memory. Short-term memory corresponds to the quickness of adaptation, and long-term memory corresponds to the stability of the adaptation. Conversion results are determined by adding information from the short-term and long-term memory to the general language model.
• Short-term memory can be implemented by boosting word scores or changing word rank based on a previous user choice of words (user input history). However, some words do not appear soon enough after being used and some words appear unexpectedly in unacceptable contexts after being used. Long-term memory can be implemented by accumulating user input history. However, some words still appear unexpectedly in unacceptable context in spite of the utilization of long-term memory.
• the architecture includes a history component for processing user input history for conversion of a phonetic string by a conversion process that output conversion results, and an adaptation component for adapting the conversion process to the user input history based on restriction(s) applied to short-term memory that impacts word appearances during the conversion process.
• the architecture performs probability boosting based on context-dependent probability differences (short-term memory), and dynamic linear- interpolation between long-term memory and baseline language model based on frequency of preceding context of word (long-term memory).
• FIG. 1 illustrates a computer-implemented phonetic system in accordance with the disclosed architecture.
• FIG. 2 illustrates a system that includes additional aspects of the phonetic system of FIG. 1.
• FIG. 3 illustrates a graph for the weights transition.
• FIG. 4 illustrates a graph for a cache weight transition.
• FIG. 5 illustrates a computer-implemented phonetic method.
• FIG. 6 illustrates additional aspects of the method of FIG. 5.
• FIG. 7 illustrates additional aspects of the method of FIG. 5.
• FIG. 8 illustrates a block diagram of a computing system operable to execute fast and stable adaptation for a statistical language model in accordance with the disclosed architecture.
• the disclosed architecture is a self-tuning technique where the user no longer needs to open a candidate list after using the product for a short period of time (e.g., 2-3 weeks). Moreover, the disclosed self-tuning technique improves a user's work performance.
• the architecture performs probability boosting based on context-dependent probability differences (short-term memory), and dynamic linear-interpolation between long-term memory and baseline language model based on frequency of preceding context of word (long-term memory).
• FIG. 1 illustrates a computer-implemented phonetic system 100 in accordance with the disclosed architecture.
• the system 100 includes a history component 102 for processing user input history 104 for conversion of a phonetic string 105 by a conversion process that output conversion results 106, and an adaptation component 108 for adapting the conversion process to the user input history 104 based on restriction(s) 110 applied to short-term memory 112 that impacts word appearances during the conversion process.
• the adaptation component 108 performs dynamic linear interpolation between long-term memory 114 and a baseline language model based on long-term memory 114.
• the restriction(s) 110 boost probability of a word when the word is other than a first candidate of a candidate list.
• the restriction(s) 110 applied to the short-term memory 112 employs a context-sensitive short-term memory bigram probability.
• the restriction(s) 110 applied to the short-term memory 112 boost a probability based on a word and a context of the word in a sentence.
• the context includes a preceding context and a succeeding context relative to the word in the sentence.
• the adaptation component 108 includes a learning algorithm that performs flag-learning based on a difference between a first candidate of a candidate list and a selected candidate of the candidate list and moves the selected candidate to a first conversion result position in a next conversion process.
• FIG. 2 illustrates a system 200 that includes additional aspects of the phonetic system 100 of FIG. 1.
• the system 200 includes the history component 102 for processing the user input history 104 for conversion of the phonetic string 105 by a conversion process 204, and the adaptation component 108 for adapting the conversion process 204 to the user input history 104 based on the restriction(s) 110 applied to the short-term memory 112 that impacts word appearances during the conversion process 204.
• the adaptation component 108 performs dynamic linear interpolation between the long-term memory 114 and a baseline language model 208 based on the long-term memory 114.
• the restriction(s) 110 boost probability of a word when the word is other than a first candidate of a candidate list.
• the restriction(s) 110 applied to the short-term memory 112 employs a context-sensitive short-term memory bigram probability.
• the restriction(s) 110 applied to the short-term memory 112 boosts a probability based on a word and a context of the word in a sentence.
• the context includes a preceding context and a succeeding context relative to the word in the sentence.
• the adaptation component 108 includes a learning algorithm that performs flag-learning based on a difference between a first candidate of a candidate list and a selected candidate of the candidate list and moves the selected candidate to a first conversion result position in a next conversion process.
• the system 200 further comprises a restriction component 206 for applying the restriction(s) 110 by boosting a probability based on context-dependent probability differences.
• the restriction component 206 can also apply one or more of the restriction(s) 110 to the long-term memory 114 by boosting a probability based on a context-dependent probability difference.
• the phonetic system 200 includes the history component 102 for processing the user input history 104 for conversion of the phonetic string 105 during the conversion process 204, the restriction component 206 for applying one or more of the restriction(s) 110 to the user input history 104 during the conversion process 204.
• the history 104 includes the short-term memory 112 and the long-term memory 114.
• the system 200 also includes the adaptation component 108 for adapting the conversion process 204 to the user input history 104 based on the restriction(s) 110.
• the restriction component 206 applies one or more of the restriction(s) 110 to the short-term memory 112.
• the applied restriction(s) 110 employ a context-sensitive short- term memory bigram probability, and one or more restrictions(s) 110 to the long-term memory 114 that boosts a probability based on a context-dependent probability difference.
• the adaptation component 108 performs dynamic linear interpolation between the long- term memory 114 and the baseline language model 208 based on the long-term memory 114.
• the restriction(s) 110 boost probability of a word when the word is other than a first candidate of a candidate list.
• the restriction(s) 110 applied to the short-term memory 112 boost a probability based on a word and a context of the word in a sentence.
• the context includes a preceding context and a succeeding context relative to the word in the sentence.
• the input method conversion result for an input phonetic string can be determined by the following probability: where W is a sentence that includes a word sequence and, ⁇ s > and ⁇ /s > are symbols for sentence-start and sentence-end, respectively.
• the equation is for the bigram model, but can be represented with trigram or higher-order n- gram models.
• the probability for each word can be defined as, where ⁇ , ⁇ , and ⁇ are linear interpolation coefficients that sum to one ( is a baseline bigram probability estimated from the training text database (when using the input method for the first time, only this probability has a value), is me bigram probability for the long-term memory, and is the bigram probability for the short-term memory.
• the bigram probability for the long- term memory can be calculated from the user input history, as follows. where C user (w n ) is the number of times the user used the word W n , and is the number of times the user uses the word sequence [0030]
• the bigram probability for short-term memory probability for words when the word is not the first candidate of the result, but user selects the word from the candidate list.
• the target weights ⁇ target an d ⁇ target are defined and used when is sufficiently large.
• Actual weights ⁇ and ⁇ for W n can be calculated as follows,
• FIG. 3 illustrates a graph 300 for the weights transition.
• the graph 300 shows the relative vertical range segments for short term memory ⁇ , long-term memory ⁇ , and baseline ⁇ , with the long-term memory ⁇ designated the ⁇ target , and the baseline designated the ⁇ target-
• the graph 300 indicates that as the number of times that the word is used increases, at a time t, the weighting for the long-term memory reaches the ⁇ target.
• W n-1 When C user (W n-1 ) is small, the long-term bigram probability tends to be high and yields an unexpected appearance of words. However, this weight-adjustment can suppress these kinds of side-effects.
• short-term memory Two approaches can be employed, either separately or combined: context-sensitive use of a short-term memory bigram probability, and probability boosting depending on the probability difference.
• context-sensitive use of short-term memory bigram probability the probability is regarded as zero when the selected-count is zero for the succeeding word sequence.
• FIG. 4 illustrates a graph 400 for a cache weight transition.
• a cache weight transition is provided using a linear function and the cache weight is only for the bigram cache (bicache).
• the bigram cache weight depends on a unigram cache (unicache) amount of the preceding word. This means that the weight for bigram cache probability P t> i cache ( w il w i-i) depends on
• the flag weight ⁇ + ⁇ is constant.
• the weight for the unigram cache is constant as well, but an offset value is added to the total unigram cache count to reduce the side- effects by the earlier cache.
• Flag-learning depends on the probability differences.
• the level of increase of a bigram flag changes depending on the amount of difference estimated between the first candidate and the selected candidate.
• the selected candidate becomes the first subsequent conversion result if the surrounding context is the same.
• the following cases can be considered and the algorithm below covers all cases. after con v ersion after editing a fter conversion a f ter editing after conversion after editing ⁇ after conversion after editing
• P(w b ⁇ w a ) is the word bigram probability before learning including baseline, cache, and flag probabilities.
• P L (w b ⁇ w ⁇ ) is the word bigram probability after learning. The change of cache probabilities is ignored here for simplification, and only the flag probabilities change after learning.
• the flag counts for candidate words which are the first candidates when a user selects an alternative candidate from the candidate list, are decremented by one after learning.
• the unigram flag counts for the corresponding candidate words, which candidate words are selected from the candidate list, are incremented by one.
• the bigram flag counts for the corresponding candidate words, which are selected from the candidate list, are incremented, the amount of increment to be determined.
• the amount of increment d can be calculated based on the differences of probabilities
• the flag-learning probability is prepared by corresponding to the flag-count.
• the range of the flag-count can be 8, 16 or 32, for example. The more the number of counts, the more precise this algorithm works.
• FIG. 1 The flag-learning probability is prepared by corresponding to the flag-count.
• the range of the flag-count can be 8, 16 or 32, for example. The more the number of counts, the more precise this algorithm works.
• FIG. 1 The flag-learning probability is prepared by corresponding to the flag-count.
• the range of the flag-count can be 8, 16 or 32, for example. The more the number of counts, the more precise this algorithm works.
• Included herein is a set of flow charts representative of exemplary methodologies for performing novel aspects of the disclosed architecture. While, for purposes of simplicity of explanation, the one or more methodologies shown herein, for example, in the form of a flow chart or flow diagram, are shown and described as a series of acts, it is to be understood and appreciated that the methodologies are not limited by the order of acts, as some acts may, in accord
• FIG. 5 illustrates a computer-implemented phonetic method.
• the user input history is processed for conversion of a phonetic string during a conversion process.
• restrictions are applied to the user input history during the conversion process, the history including short-term memory and long-term memory.
• the conversion process is adapted to the user input history based on the restrictions.
• FIG. 6 illustrates additional aspects of the method of FIG. 5.
• a restriction is applied that boosts a probability based on context-dependent probability differences.
• dynamic linear interpolation is performed between long-term memory and a baseline language model based on the long-term memory.
• probability of a word is boosted when the word is other than a first candidate of a candidate list.
• FIG. 7 illustrates additional aspects of the method of FIG. 5.
• a restriction is applied to the short-term memory that boosts a probability based on a word and a context of the word in a sentence.
• flag-learning is performed based on a difference between a first candidate of a candidate list and a selected candidate of the candidate list.
• the selected candidate is moved to a first conversion result position in a next conversion process.
• a component can be, but is not limited to being, a process running on a processor, a processor, a hard disk drive, multiple storage drives (of optical, solid state, and/or magnetic storage medium), an object, an executable, a thread of execution, a program, and/or a computer.
• a component can be, but is not limited to being, a process running on a processor, a processor, a hard disk drive, multiple storage drives (of optical, solid state, and/or magnetic storage medium), an object, an executable, a thread of execution, a program, and/or a computer.
• an application running on a server and the server can be a component.
• One or more components can reside within a process and/or thread of execution, and a component can be localized on one computer and/or distributed between two or more computers.
• FIG. 8 there is illustrated a block diagram of a computing system 800 operable to execute fast and stable adaptation for a statistical language model in accordance with the disclosed architecture.
• FIG. 8 and the following discussion are intended to provide a brief, general description of the suitable computing system 800 in which the various aspects can be implemented.
• the computing system 800 for implementing various aspects includes the computer 802 having processing unit(s) 804, a system memory 806, and a system bus 808.
• the processing unit(s) 804 can be any of various commercially available processors such as single-processor, multi-processor, single-core units and multi-core units.
• the system memory 806 can include volatile (VOL) memory 810 (e.g., random access memory (RAM)) and non-volatile memory (NON-VOL) 812 (e.g., ROM, EPROM, EEPROM, etc.).
• VOL volatile
• NON-VOL non-volatile memory
• a basic input/output system (BIOS) can be stored in the non-volatile memory 812, and includes the basic routines that facilitate the communication of data and signals between components within the computer 802, such as during startup.
• the volatile memory 810 can also include a high-speed RAM such as static RAM for caching data.
• the system bus 808 provides an interface for system components including, but not limited to, the memory subsystem 806 to the processing unit(s) 804.
• the system bus 808 can be any of several types of bus structure that can further interconnect to a memory bus (with or without a memory controller), and a peripheral bus (e.g., PCI, PCIe, AGP, LPC, etc.), using any of a variety of commercially available bus architectures.
• the computer 802 further includes storage subsystem(s) 814 and storage interface(s) 816 for interfacing the storage subsystem(s) 814 to the system bus 808 and other desired computer components.
• the storage subsystem(s) 814 can include one or more of a hard disk drive (HDD), a magnetic floppy disk drive (FDD), and/or optical disk storage drive (e.g., a CD-ROM drive DVD drive), for example.
• the storage interface(s) 816 can include interface technologies such as EIDE, ATA, SATA, and IEEE 1394, for example.
• One or more programs and data can be stored in the memory subsystem 806, a removable memory subsystem 818 (e.g., flash drive form factor technology), and/or the storage subsystem(s) 814 (e.g., optical, magnetic, solid state), including an operating system 820, one or more application programs 822, other program modules 824, and program data 826.
• a removable memory subsystem 818 e.g., flash drive form factor technology
• the storage subsystem(s) 814 e.g., optical, magnetic, solid state
• an operating system 820 e.g., one or more application programs 822, other program modules 824, and program data 826.
• the one or more application programs 822, other program modules 824, and program data 826 can include the system 100 an d components of FIG. 1, the system 200 and components of FIG. 2, the relationships represented by the graphs 300 and 400, and the methods represented by the flow charts of Figures 5-7, for example.
• programs include routines, methods, data structures, other software components, etc., that perform particular tasks or implement particular abstract data types. All or portions of the operating system 820, applications 822, modules 824, and/or data 826 can also be cached in memory such as the volatile memory 810, for example. It is to be appreciated that the disclosed architecture can be implemented with various commercially available operating systems or combinations of operating systems (e.g., as virtual machines).
• the storage subsystem(s) 814 and memory subsystems (806 and 818) serve as computer readable media for volatile and non- volatile storage of data, data structures, computer-executable instructions, and so forth.
• Computer readable media can be any available media that can be accessed by the computer 802 and includes volatile and nonvolatile media, removable and non-removable media.
• the media accommodate the storage of data in any suitable digital format. It should be appreciated by those skilled in the art that other types of computer readable media can be employed such as zip drives, magnetic tape, flash memory cards, cartridges, and the like, for storing computer executable instructions for performing the novel methods of the disclosed architecture.
• a user can interact with the computer 802, programs, and data using external user input devices 828 such as a keyboard and a mouse.
• Other external user input devices 828 can include a microphone, an IR (infrared) remote control, a joystick, a game pad, camera recognition systems, a stylus pen, touch screen, gesture systems (e.g., eye movement, head movement, etc.), and/or the like.
• the user can interact with the computer 802, programs, and data using onboard user input devices 830 such a touchpad, microphone, keyboard, etc., where the computer 802 is a portable computer, for example.
• I/O device interface(s) 832 are connected to the processing unit(s) 804 through input/output (I/O) device interface(s) 832 via the system bus 808, but can be connected by other interfaces such as a parallel port, IEEE 1394 serial port, a game port, a USB port, an IR interface, etc.
• the I/O device interface(s) 832 also facilitate the use of output peripherals 834 such as printers, audio devices, camera devices, and so on, such as a sound card and/or onboard audio processing capability.
• One or more graphics interface(s) 836 (also commonly referred to as a graphics processing unit (GPU)) provide graphics and video signals between the computer 802 and external display(s) 838 (e.g., LCD, plasma) and/or onboard displays 840 (e.g., for portable computer).
• graphics interface(s) 836 can also be manufactured as part of the computer system board.
• the computer 802 can operate in a networked environment (e.g., IP) using logical connections via a wired/wireless communications subsystem 842 to one or more networks and/or other computers.
• the other computers can include workstations, servers, routers, personal computers, microprocessor-based entertainment appliance, a peer device or other common network node, and typically include many or all of the elements described relative to the computer 802.
• the logical connections can include wired/wireless connectivity to a local area network (LAN), a wide area network (WAN), hotspot, and so on.
• LAN and WAN networking environments are commonplace in offices and companies and facilitate enterprise-wide computer networks, such as intranets, all of which may connect to a global communications network such as the Internet.
• the computer 802 When used in a networking environment the computer 802 connects to the network via a wired/wireless communication subsystem 842 (e.g., a network interface adapter, onboard transceiver subsystem, etc.) to communicate with wired/wireless networks, wired/wireless printers, wired/wireless input devices 844, and so on.
• the computer 802 can include a modem or has other means for establishing communications over the network.
• programs and data relative to the computer 802 can be stored in the remote memory/storage device, as is associated with a distributed system. It will be appreciated that the network connections shown are exemplary and other means of establishing a communications link between the computers can be used.
• the computer 802 is operable to communicate with wired/wireless devices or entities using the radio technologies such as the IEEE 8O2.xx family of standards, such as wireless devices operatively disposed in wireless communication (e.g., IEEE 802.11 over- the-air modulation techniques) with, for example, a printer, scanner, desktop and/or portable computer, personal digital assistant (PDA), communications satellite, any piece of equipment or location associated with a wirelessly detectable tag (e.g., a kiosk, news stand, restroom), and telephone.
• PDA personal digital assistant
• the communications can be a predefined structure as with a conventional network or simply an ad hoc communication between at least two devices.
• Wi-Fi networks use radio technologies called IEEE 802.1 Ix (a, b, g, etc.) to provide secure, reliable, fast wireless connectivity.
• IEEE 802.1 Ix a, b, g, etc.
• a Wi-Fi network can be used to connect computers to each other, to the Internet, and to wire networks (which use IEEE 802.3-related media and functions).

Landscapes

• Engineering & Computer Science (AREA)
• Physics & Mathematics (AREA)
• Theoretical Computer Science (AREA)
• Computational Linguistics (AREA)
• Health & Medical Sciences (AREA)
• Audiology, Speech & Language Pathology (AREA)
• Artificial Intelligence (AREA)
• General Engineering & Computer Science (AREA)
• Human Computer Interaction (AREA)
• General Physics & Mathematics (AREA)
• Multimedia (AREA)
• Acoustics & Sound (AREA)
• General Health & Medical Sciences (AREA)
• Probability & Statistics with Applications (AREA)
• Machine Translation (AREA)
• Document Processing Apparatus (AREA)

Priority Applications (3)

Applications Claiming Priority (2)

Publications (2)

Family

ID=42785345

Family Applications (1)

Country Status (6)

Cited By (2)

Families Citing this family (10)

Family Cites Families (35)

• 2009
  • 2009-03-30 US US12/413,606 patent/US8798983B2/en not_active Expired - Fee Related
• 2010
  • 2010-02-23 TW TW099105182A patent/TWI484476B/zh not_active IP Right Cessation
  • 2010-03-26 WO PCT/US2010/028932 patent/WO2010117688A2/en not_active Ceased
  • 2010-03-26 CN CN2010800158015A patent/CN102369567B/zh not_active Expired - Fee Related
  • 2010-03-26 JP JP2012503537A patent/JP2012522278A/ja active Pending
  • 2010-03-26 KR KR1020117022845A patent/KR101679445B1/ko not_active Expired - Fee Related

Also Published As

Similar Documents

Legal Events