WO2022131740A1 - Procédés et systèmes de génération d'abréviations pour un mot cible - Google Patents

Procédés et systèmes de génération d'abréviations pour un mot cible Download PDF

Info

Publication number
WO2022131740A1
WO2022131740A1 PCT/KR2021/018954 KR2021018954W WO2022131740A1 WO 2022131740 A1 WO2022131740 A1 WO 2022131740A1 KR 2021018954 W KR2021018954 W KR 2021018954W WO 2022131740 A1 WO2022131740 A1 WO 2022131740A1
Authority
WO
WIPO (PCT)
Prior art keywords
feature
phonetic
abbreviations
word
phonetic string
Prior art date
Application number
PCT/KR2021/018954
Other languages
English (en)
Inventor
Mayur AGGARWAL
Natasha MEENA
Original Assignee
Samsung Electronics Co., Ltd.
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
Application filed by Samsung Electronics Co., Ltd. filed Critical Samsung Electronics Co., Ltd.
Publication of WO2022131740A1 publication Critical patent/WO2022131740A1/fr

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR 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
    • G06F40/151—Transformation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR 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
    • G06F3/0236—Character input methods using selection techniques to select from displayed items
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR 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; CALCULATING OR 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
    • G06F3/048—Interaction techniques based on graphical user interfaces [GUI]
    • G06F3/0487—Interaction techniques based on graphical user interfaces [GUI] using specific features provided by the input device, e.g. functions controlled by the rotation of a mouse with dual sensing arrangements, or of the nature of the input device, e.g. tap gestures based on pressure sensed by a digitiser
    • G06F3/0488—Interaction techniques based on graphical user interfaces [GUI] using specific features provided by the input device, e.g. functions controlled by the rotation of a mouse with dual sensing arrangements, or of the nature of the input device, e.g. tap gestures based on pressure sensed by a digitiser using a touch-screen or digitiser, e.g. input of commands through traced gestures
    • G06F3/04886—Interaction techniques based on graphical user interfaces [GUI] using specific features provided by the input device, e.g. functions controlled by the rotation of a mouse with dual sensing arrangements, or of the nature of the input device, e.g. tap gestures based on pressure sensed by a digitiser using a touch-screen or digitiser, e.g. input of commands through traced gestures by partitioning the display area of the touch-screen or the surface of the digitising tablet into independently controllable areas, e.g. virtual keyboards or menus
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR 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
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00—Handling natural language data
    • G06F40/20—Natural language analysis
    • G06F40/274—Converting codes to words; Guess-ahead of partial word inputs
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00—Handling natural language data
    • G06F40/20—Natural language analysis
    • G06F40/279—Recognition of textual entities
    • G06F40/284—Lexical analysis, e.g. tokenisation or collocates
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods

Definitions

  • the present disclosure generally relates to generating one or more abbreviations corresponding to a target word, and particularly relates to generating the one or more abbreviations corresponding to the target word through machine learning/artificial intelligence criteria.
  • a context analysis may be performed for providing possible word suggestions to the user.
  • a context analysis may be performed for providing next word suggestions.
  • the user may write a word and based upon a contextual analysis, an emoji or a GIF suggestion may be provided to the user.
  • the user may misspell a word. In such a case, a contextual or statistical analysis may be performed and accordingly, a suggestion for auto-correction may be provided to the user
  • suggestions that offer abbreviated forms of an input word may be provided to the users.
  • most of these techniques use a pre-defined knowledge base for assessing the input words and providing the abbreviation suggestions.
  • these solutions may not be able to accurately understand the context/phonetical information when providing abbreviation suggestions.
  • inaccurate abbreviation suggestions may be provided to the user.
  • a method of generating abbreviations corresponding to at least one target word includes, identifying the at least one target word pertaining to the input word.
  • the method includes determining a phonetic associated with the at least one target word.
  • the method includes generating at least one feature vector based on at least one of phoneme type, articulation information and sequential information corresponding to one or more phonemes associated with the phonetic.
  • the method includes determining one or more valid phonetic string sequences based on the at least one feature vector.
  • the method includes determining one or more sets of abbreviations corresponding to the one or more valid phonetic string sequences.
  • the method further includes generating one or more abbreviations from the one or more set of abbreviations for replacement of the at least one target word.
  • a system of generating abbreviations corresponding to at least one target word includes a word generator configured to identify the at least one target word pertaining to the input word.
  • the system includes a phonetic generator configured to determine a phonetic associated with the at least one target word.
  • the system includes a phonetic string sequence processor including a feature generator configured to generating at least one feature vector based on at least one of phoneme type, articulation information and a sequential information corresponding to one or more phonemes associated with the phonetic.
  • the phonetic string sequence processor further includes a dynamic sequence selector configured to determine one or more valid phonetic string sequences based on the at least one feature vector.
  • the system further includes an abbreviation processor including a phonetic string determining engine configured to determine one or more sets of abbreviations corresponding to the one or more valid phonetic string sequences. Further, the phonetic string determining engine is configured to generate one or more abbreviations from the one or more set of abbreviations for replacement of the at least one target word.
  • an abbreviation processor including a phonetic string determining engine configured to determine one or more sets of abbreviations corresponding to the one or more valid phonetic string sequences. Further, the phonetic string determining engine is configured to generate one or more abbreviations from the one or more set of abbreviations for replacement of the at least one target word.
  • a computer-readable storage medium having a computer program stored thereon that performs, when executed by a processor, the method above.
  • FIG. 1 illustrates a flow diagram depicting a method of generating abbreviations corresponding to at least one target word, in accordance with an embodiment of the present subject matter
  • FIG. 2 illustrates a schematic block diagram of a system for generating the abbreviations corresponding to at least one target word, in accordance with an embodiment of the present subject matter
  • FIG. 3 illustrates an operational flow diagram depicting a process of generating one or more abbreviations corresponding to at least one target word, in accordance with an embodiment of the present subject matter
  • FIG. 4 illustrates a flow diagram depicting a process to analyze a context of an input word and generating a phonetic corresponding to at least one target word, in accordance with an embodiment of the present subject matter
  • FIG. 5a illustrates an operational flow diagram depicting a process for a phonetic string sequence processing, in accordance with an embodiment of the present subject matter
  • FIG. 5b illustrates an operational flow diagram depicting a process for generating articulation feature, in accordance with an embodiment of the present subject matter
  • FIG. 5c illustrates an operational flow diagram depicting role of a dynamic feature selector, in accordance with an embodiment of the present subject matter
  • FIG. 5d illustrates an operational flow diagram depicting a process for feeding features to a dynamic sequence selector, in accordance with an embodiment of the present subject matter
  • FIG. 5e illustrates an operational flow diagram depicting a process for dynamic sequence validation, in accordance with an embodiment of the present subject matter
  • FIG. 6a illustrates an operational flow diagram depicting a method for processing one or more abbreviations, in accordance with an embodiment of the present subject matter
  • FIG. 6b illustrates an operational flow diagram for determining a phonetically closest string, in accordance with an embodiment of the present subject matter
  • FIG. 6c illustrates a diagram depicting a word database including a vocabulary space, in accordance with an embodiment of the present subject matter
  • FIG. 6d illustrates a diagram depicting an identification of an N dimensional sphere, in accordance with an embodiment of the present subject matter
  • FIG. 6e illustrates a diagram depicting formation of the one or more abbreviations, in accordance with an embodiment of the present subject matter
  • FIG.6f illustrates a diagram depicting probability of usage of one or more abbreviation of a phonetic, in accordance with an embodiment of the present subject matter
  • FIG. 7a illustrates a detailed architecture of a method to generate one or more abbreviations corresponding to at least one target word, in accordance with an embodiment of the present subject matter
  • FIG. 7b illustrates another detailed architecture of the method to generate the one or more abbreviations corresponding to at least one target word, in accordance with an embodiment of the present subject matter
  • FIG. 7c illustrates yet another detailed architecture of the method to generate the one or more abbreviations corresponding to at least one target word, in accordance with an embodiment of the present subject matter
  • FIG. 8a illustrates an application use case diagram depicting a scenario of a device understanding context associated with one or more words, according to an existing technique
  • FIG. 8b illustrates an application use case diagram depicting a scenario of the device configured to understand a phonetic, in accordance with an embodiment of the present subject matter
  • FIG. 9 illustrates a process for generating one or more abbreviations corresponding to a target word, in accordance with an embodiment of the present subject matter
  • FIG. 10a illustrates a use case, according to an existing technique
  • FIG. 10b illustrates the use case of FIG. 10a, in accordance with an embodiment of the present subject matter
  • FIGS. 11-19 illustrate a number of use cases, in accordance with an embodiment of the present subject matter
  • FIG. 20 illustrates a process for suggesting and/or expanding an abbreviation, in accordance with an embodiment of the present subject matter
  • FIG. 21 illustrates a representative architecture to provide tools and development environment described herein for a technical-realization of the implementation in FIG. 1 and FIG. 20 through an AI model-based computing device, in accordance with an embodiment of the present subject matter.
  • FIG. 1 illustrates a flow diagram of a method 100 of generating abbreviations corresponding to at least one target word, in accordance with an embodiment of the present subject matter.
  • the method 100 may be configured to generate the abbreviations corresponding to the at least one target word upon performing any of an Artificial Intelligence (AI) technique, and a Machine Learning (ML) technique on the at least one target word.
  • AI Artificial Intelligence
  • ML Machine Learning
  • the method 100 includes identifying (step 102) the at least one target word pertaining to an input word.
  • the method 100 includes determining (step 104) a phonetic associated with the at least one target word.
  • the method 100 includes generating (step 106) at least one feature vector based on at least one of phoneme type, articulation information and sequential information corresponding to one or more phonemes associated with the phonetic.
  • the method includes, determining (step 108) one or more valid phonetic string sequences based on the at least one feature vector.
  • the method further includes, determining (step 110) one or more sets of abbreviations corresponding to the one or more valid phonetic string sequences.
  • the method also includes, generating (step 112) one or more abbreviations from the one or more set of abbreviations for replacement of the at least one target word.
  • FIG. 2 illustrates a schematic block diagram 200 of a system 202 to generate abbreviations corresponding to at least one target word.
  • the system 202 may be incorporated in a device. Examples of the device may include, but are not limited to a laptop, a tab, a smart phone, a Personal Computer (PC).
  • the at least one target word may be identified in response to receiving an input word.
  • the system 202 may be configured to receive the input word in the form of a text, or a speech from a user.
  • the system 202 may be configured to determine a context associated with the input word for identifying the at least one target word pertaining to the input word.
  • the system 202 may be configured to generate at least one feature vector corresponding to the at least one target word.
  • the generated least one feature vector may represent a feature matrix that comprises the at least one feature vector.
  • the system 202 may be configured to determine one or more sets of abbreviations corresponding to one or more valid phonetic string sequences.
  • the one or more valid phonetic string sequences may be based on the at least one feature vector and the articulation information corresponding to the at least one feature vector.
  • the system 202 may be configured to generating one or more abbreviations from the one or more set of abbreviations for replacement of the at least one target word. Details of the above aspects performed by the system 202 shall be explained below.
  • the system 202 includes a processor 204, a memory 206, data 208, module (s) 210, resource (s) 212, a display unit 214, a context analysis engine 216, a word generator 218, a phonetic generator 220, a phonetic string sequence processor 222, a feature generator 224, a dynamic sequence selector 226, an abbreviation processor 228, and a phonetic string determining engine 230.
  • the processor 204, the memory 206, the data 208, the module (s) 210, the resource (s) 212, the display unit 214, context analysis engine 216, the word generator 218, the phonetic generator 220, the phonetic string sequence processor 222, the feature generator 224, the dynamic sequence selector 226, the abbreviation processor 228, and the phonetic string determining engine 230 may be communicatively coupled to one another.
  • At least one of the plurality of modules may be implemented through an AI model.
  • a function associated with AI may be performed through the non-volatile memory or the volatile memory, and/or the processor.
  • the processor 204 may include one or a plurality of processors.
  • one or a plurality of processors may be a general purpose processor, such as a central processing unit (CPU), an application processor (AP), or the like, a graphics-only processing unit such as a graphics processing unit (GPU), a visual processing unit (VPU), and/or an AI-dedicated processor such as a neural processing unit (NPU).
  • CPU central processing unit
  • AP application processor
  • GPU graphics-only processing unit
  • VPU visual processing unit
  • NPU neural processing unit
  • a plurality of processors control the processing of the input data in accordance with a predefined operating rule or artificial intelligence (AI) model stored in the non-volatile memory or the volatile memory.
  • the predefined operating rule or artificial intelligence model is provided through training or learning.
  • being provided through learning means that, by applying a learning technique to a plurality of learning data, a predefined operating rule or AI model of a desired characteristic is made.
  • the learning may be performed on a device itself in which AI according to an embodiment is performed, and/or may be implemented through a separate server/system.
  • the AI model may consist of a plurality of neural network layers. Each layer has a plurality of weight values, and performs a layer operation through calculation of a previous layer and an operation of a plurality of weights.
  • neural networks include, but are not limited to, convolutional neural network (CNN), deep neural network (DNN), recurrent neural network (RNN), restricted Boltzmann Machine (RBM), deep belief network (DBN), bidirectional recurrent deep neural network (BRDNN), generative adversarial networks (GAN), and deep Q-networks.
  • CNN convolutional neural network
  • DNN deep neural network
  • RNN recurrent neural network
  • RBM restricted Boltzmann Machine
  • DNN deep belief network
  • BNN bidirectional recurrent deep neural network
  • GAN generative adversarial networks
  • the learning technique is a method for training a predetermined target device (for example, a robot) using a plurality of learning data to cause, allow, or control the target device to make a determination or prediction.
  • Examples of learning techniques include, but are not limited to, supervised learning, unsupervised learning, semi-supervised learning, or reinforcement learning.
  • a method of generating abbreviations corresponding to an input word which is an analog signal, via (e.g., a microphone) and convert the speech part into computer readable text using an automatic speech recognition (ASR) model.
  • ASR automatic speech recognition
  • NLU natural language understanding
  • the ASR model or NLU model may be an artificial intelligence model.
  • the artificial intelligence model may be processed by an artificial intelligence-dedicated processor designed in a hardware structure specified for artificial intelligence model processing.
  • the artificial intelligence model may be obtained by training.
  • the artificial intelligence model may include a plurality of neural network layers. Each of the plurality of neural network layers includes a plurality of weight values and performs neural network computation by computation between a result of computation by a previous layer and the plurality of weight values.
  • Language understanding is a technique for recognizing and applying/processing human language/text and includes, e.g., natural language processing, machine translation, dialog system, question answering, or speech recognition/synthesis.
  • the system 202 may be understood as one or more of a hardware, a software, a logic-based program, a configurable hardware, and the like.
  • the processor 204 may be a single processing unit or a number of units, all of which could include multiple computing units.
  • the processor 204 may be implemented as one or more microprocessors, microcomputers, microcontrollers, digital signal processors, central processing units, processor cores, multi-core processors, multiprocessors, state machines, logic circuitries, application-specific integrated circuits, field-programmable gate arrays and/or any devices that manipulate signals based on operational instructions.
  • the processor 204 may be configured to fetch and/or execute computer-readable instructions and/or data stored in the memory 206.
  • the memory 206 may include any non-transitory computer-readable medium known in the art including, for example, volatile memory, such as static random access memory (SRAM) and/or dynamic random access memory (DRAM), and/or non-volatile memory, such as read-only memory (ROM), erasable programmable ROM (EPROM), flash memory, hard disks, optical disks, and/or magnetic tapes.
  • volatile memory such as static random access memory (SRAM) and/or dynamic random access memory (DRAM)
  • non-volatile memory such as read-only memory (ROM), erasable programmable ROM (EPROM), flash memory, hard disks, optical disks, and/or magnetic tapes.
  • ROM read-only memory
  • EPROM erasable programmable ROM
  • the data 208 serves, amongst other things, as a repository for storing data processed, received, and generated by one or more of the processor 204, the module (s) 210, the resource (s) 212, the display unit 214, the context analysis engine 216, the word generator 218, the phonetic generator 220, the phonetic string sequence processor 222, the feature generator 224, the dynamic sequence selector 226, the abbreviation processor 228, and the phonetic string determining engine 230.
  • the module(s) 210 may include routines, programs, objects, components, data structures, etc., which perform particular tasks or implement data types.
  • the module(s) 210 may also be implemented as, signal processor(s), state machine(s), logic circuitries, and/or any other device or component that manipulate signals based on operational instructions.
  • the module(s) 210 may be implemented in hardware, as instructions executed by at least one processing unit, e.g., processor 204, or by a combination thereof.
  • the processing unit may be a general-purpose processor that executes instructions to cause the general-purpose processor to perform operations or, the processing unit may be dedicated to performing the required functions.
  • the module(s) 210 may be machine-readable instructions (software) which, when executed by a processor/processing unit, may perform any of the described functionalities.
  • the module(s) 210 may be machine-readable instructions (software) which, when executed by a processor 204/processing unit, perform any of the described functionalities.
  • the resource(s) 212 may be physical and/or virtual components of the system 202 that provide inherent capabilities and/or contribute towards the performance of the system 202.
  • Examples of the resource(s) 212 may include, but are not limited to, a memory (e.g., the memory 206), a power unit (e.g. a battery), a display unit (e.g., the display unit 214) etc.
  • the resource(s) 212 may include a power unit/battery unit, a network unit, etc., in addition to the processor 204, and the memory 206.
  • the display unit 214 may display various types of information (for example, media contents, multimedia data, text data, etc.) to the system 202.
  • the display unit 214 may include, but is not limited to, a liquid crystal display (LCD), a light-emitting diode (LED) display, an organic LED (OLED) display, a plasma cell display, an electronic ink array display, an electronic paper display, a flexible LCD, a flexible electrochromic display, and/or a flexible electrowetting display.
  • the context analysis engine 216, the word generator 218, the phonetic generator 220, the phonetic string sequence processor 222, the feature generator 224, the dynamic sequence selector 226, the abbreviation processor 228, and the phonetic string determining engine 230 include routines, programs, objects, components, data structures, etc., which perform particular tasks or implement data types.
  • the context analysis engine 216, the word generator 218, the phonetic generator 220, the phonetic string sequence processor 222, the feature generator 224, the dynamic sequence selector 226, the abbreviation processor 228, and the phonetic string determining engine 230 may also be implemented as, signal processor(s), state machine(s), logic circuitries, and/or any other device or component that manipulate signals based on operational instructions.
  • the context analysis engine 216, the word generator 218, the phonetic generator 220, the phonetic string sequence processor 222, the feature generator 224, the dynamic sequence selector 226, the abbreviation processor 228, and the phonetic string determining engine 230 can be implemented in hardware, instructions executed by a processing unit, or by a combination thereof.
  • the processing unit can comprise a computer, a processor, such as the processor 204, a state machine, a logic array or any other suitable devices capable of processing instructions.
  • the processing unit can be a general-purpose processor which executes instructions to cause the general-purpose processor to perform the required tasks or, the processing unit can be dedicated to perform the required functions.
  • the context analysis engine 216 may be configured to determine a context from the input word in response to receiving the input word from a user.
  • the input word may be one of a complete word and an incomplete word.
  • the context may be based on a statement from which the input word is taken out to process.
  • the context analysis engine 216 may be configured to determine the context based on one or more previous user inputs and historical user input data. Such data, in an example, may be stored in the data 208.
  • the word generator 218 may be configured to identify the at least one target word in response to receiving the input word from the user.
  • the target word may be understood as a standard dictionary word that is either inferred from the input word or is same as the input word.
  • the target word may be the predicted complete input word.
  • the target word may also be a next predicted word based on the context analysis of the input word.
  • the target word may be the next predicted word based on the context analysis of the input word.
  • the target word may also be same as the input word.
  • the at least one target word may be determined based on the context of the complete or incomplete input word, determined by the context analysis engine 216.
  • the phonetic generator 220 may be configured to determine a phonetic associated with the at least one target word.
  • the phonetic may include pronunciation information associated with the at least one target word.
  • the phonetic may include one or more phonemes related to the pronunciation information.
  • the phonetic string sequence processor 222 may be configured to determine one or more valid phonetic string sequences based on at least one feature vector related to the phonetic and the at least one target word.
  • the feature generator 224 incorporated in the phonetic string sequence processor 222 may be configured to generate the at least one feature vector associated with the phonetic.
  • the at least one feature vector may be based on the one or more phonemes such that the at least one feature vector may include at least one of phoneme type, articulation information and sequential information corresponding to one or more phonemes.
  • the feature generator 224 may be configured to generate the at least one feature vector categorizing the one or more phonemes into one or more vowels and one or more consonants. Further, the feature generator 224 may be configured to perform a classification of the one or more vowels and the one or more consonants categorized from the one or more phonemes. In an embodiment, the classification may be based on the articulation information related to the one or more phonemes.
  • the feature generator 224 may be configured to categorize the one or more phonemes associated with the phonetic into one or more vowels and one or more consonants. Further, the feature generator 224 may be configured to determine the phoneme type the articulation information and the sequential information corresponding to the one or more phonemes categorized as the one or more vowels and the one or more consonants. Further, the feature generator 224 may be configured to generate the at least one feature vector based on the at least one of the phoneme type, the determined articulation information and the sequential information corresponding to the one or more phonemes associated with the phonetic. Further, the feature generator 224 may be configured to encode the generated at least one feature vector to generate at least one articulatory feature vector. Specifically, the feature generator 224 may be configured to encode categorical data such as the phoneme type the articulation information in at least one feature vector to generate at least one articulatory feature vector.
  • the articulation information associated with the one or more phonemes categorized as the one or more vowels may include at least one of a height of a tongue, a tenseness, and a backness associated with the lips while speaking the one or more vowels.
  • the articulation information associated with the one or more phonemes categorized as the one or more consonants may include at least one of an articulation place, a manner, and a voice while speaking the one or more consonants.
  • the phoneme type associated with the one or more phonemes comprises information whether each of the one or more phonemes are vowel or not.
  • the sequential information associated with the one or more phonemes comprises order information associated with the phonetic.
  • the feature generator 224 may be configured to perform an encoding operation on the one or more vowels and the one or more consonants.
  • the encoding operation may be based on the classification.
  • the encoding operation may include encoding the one or more vowels and the one or more consonants with a set of numerical values for generating the at least one feature vector corresponding to the phonetic.
  • the at least one feature vector may be generated in a matrix form that comprises at least one feature vector.
  • the dynamic sequence selector 226 may be configured to determine the one or more valid phonetic string sequences based on the at least one feature vector.
  • the dynamic sequence selector 226 may be configured to obtain the at least one feature vector from the at least one feature vector generator 224.
  • the dynamic sequence selector 226 may further be configured to obtain previous iteration output information.
  • the previous output information may be related to previous one or more valid phonetic string sequences determined by the dynamic sequence selector 226.
  • the dynamic sequence selector 226 may be configured to iteratively generate a left feature and a right feature from the feature vector by dividing the feature vector.
  • the left feature and the right feature may be based on a previous iteration information related to the feature vector.
  • the previous iteration information may include information related to a previous left feature and a previous right feature generated from the feature vector.
  • the dynamic sequence selector 226 may be configured to initiate an iteration from a first feature to a penultimate feature of the feature vector.
  • the dynamic sequence selector 226 may be configured to form a phonetic string sequence based on the left feature and the right feature.
  • the phonetic string sequence may include the right feature and the left feature in a position such that the right feature is positioned after the left feature.
  • the dynamic sequence selector 226 may be configured to validate the phonetic string sequence formed for each iteration.
  • the dynamic sequence selector 226 may be configured to validate the phonetic string sequence by ascertaining whether a muscle change is caused or not upon pronouncing the right feature after the left feature. Further, the dynamic sequence selector 226 may be configured to determine the phonetic string sequence to be the one or more valid phone string sequences when pronouncing the right feature after the left feature related to the phonetic string sequence causes the muscle change.
  • the abbreviation processor 228 may be configured to determine one or more sets of abbreviations corresponding to the one or more valid phonetic string sequences and one or more abbreviations.
  • the phonetic string determining engine 230 incorporated in the abbreviation processor 228 may be configured to determine the one or more sets of abbreviations and the one or more abbreviations from the one or more sets of abbreviations for replacement of the at least one target word.
  • the phonetic string determining engine 230 may be configured to determine the one or more sets of abbreviations by determining one or more substrings corresponding to each of the one or more valid phonetic string sequences. Further, the phonetic string determining engine 230 may be configured to identify one or more phonetically matching strings related to each of the one or more substrings from a word database.
  • the word database may include a number of strings. Further, the number of strings may include one or more strings whose corresponding phonemes are similar are positioned closely in vocabulary space associated with the word database.
  • the phonetic string determining engine 230 may be configured to determine the one or more sets of abbreviations based on the phonetically matching strings identified from the word database corresponding to the one or more substrings.
  • the phonetically matching strings may further be related to each of the one or more valid phonetic string sequences.
  • the phonetic string determining engine 230 may further be configured to generate the one or more abbreviations from the one or more sets of abbreviations for replacement of the at least one target word based on at least one of a number of parameters.
  • the number of parameters may include a length of the one or more abbreviation, a usage of the one or more abbreviations, a context associated with the at least one target word, and a historic abbreviation feedback database comprising of a number of previously user-selected abbreviations and a corresponding confidence score of each of the number of previously user-selected abbreviations.
  • the display unit 214 may be configured to display the one or more abbreviations generated from the one or more sets of abbreviations. Further, the display unit 214 may be configured to receive a selection of an abbreviation from the one or more abbreviations.
  • the abbreviation processor 228 may be configured to replace the at least one target word with the selected abbreviation from the one or more abbreviations.
  • the concepts of the present subject matter may be implemented for condensing text of a paragraph to obtain a reduced number of characters.
  • the at least one word may be selected from a user selected input text.
  • the user selected input text may be the paragraph which is intended to be condense and said user selected input text may comprise a number of words.
  • the input word(s) may be taken from the user selected input text, i.e., the paragraph.
  • the target word(s) may be identified accordingly, and as explained above, abbreviations may be generated and used to replace the words in the paragraph, thereby resulting in a paragraph having reduced characters.
  • FIG. 3 illustrates an operational flow diagram 300 depicting a method to generate one or more abbreviations corresponding to at least one target word, in accordance with an embodiment of the present subject matter.
  • generating the one or more abbreviations may be based on one of an Artificial Intelligence (AI) technique, and a Machine Learning (ML) technique.
  • the one or more abbreviations may be based on one or more phonemes generated from a phonetic related to the least one target word.
  • the method includes receiving a word.
  • the word may be the input word as referred in the FIG. 1. Further, in an embodiment, the word may be typed or written on a user device by a user. In an embodiment, the input word may be one of a complete word, and an incomplete word. Further, the input word may be a single word. In an embodiment, the input word may be a part of a statement.
  • the method includes performing a context analysis on the input word.
  • the context analysis may be performed to determine a most probable word on the input word by the context analysis engine 216 referred in the FIG. 2.
  • the most probable word may be referred as the at least one target word as referred in the FIG. 1.
  • the at least one target word may be determined based on the context of the statement determined through the context analysis engine 216.
  • the method further includes determining whether the at least one target word may be one of a valid dictionary word, and a personal abbreviation.
  • the personal abbreviation may be the one of one or more abbreviations referred in the FIG. 1.
  • the at least one target word may be same as the input word.
  • the method includes extracting a phonetic of the at least one target word.
  • the phonetic may include pronunciation information related to the at least one target word.
  • the pronunciation information may be referred as articulation information as referred as in the FIG. 1.
  • the phonetic related to the at least one target word may be extracted by the phonetic generator 220 as referred in the FIG. 2.
  • the method includes classifying one or more phonemes of the phonetic based on articulatory information related to the one or more phonemes.
  • the articulatory information may include movement of a human muscle upon speaking the one or more phonemes.
  • the classification may be performed by the feature generator 224 as referred in the FIG. 2.
  • the method includes encoding the articulatory information related to the one or more phonemes and sequential information related to the one or more phonemes.
  • the encoding is performed to form at least one articulatory feature vector.
  • the encoding may be performed by the feature generator 224.
  • the at least one articulatory feature vector may be referred as the at least one feature vector as referred in the FIG. 1.
  • the method includes a dynamic selection of a left and a right feature for validation of a phonetic string sequence based on previous output of the validation.
  • the left feature and the right feature may be formed from the at least one feature vector in an iterative process.
  • the phonetic string sequence upon validation, may be determined as one or more valid phonetic string sequences as referred in the FIG. 1.
  • the method includes, performing intelligent validation of the right feature and the left feature such that the validation determines the possibility of the right feature appearing after the left feature.
  • the validation may include determining whether placing the right feature after the left feature causes a change in the human muscle movement upon speaking.
  • the one or more valid phonetic string sequences may include the phonetic string sequence corresponding to the right feature and the left feature causing the change in the human muscle movement upon placing the right feature after the left feature.
  • the validation may be performed by the dynamic sequence selector 226 referred in the FIG. 2. In an embodiment, where it is determined that the right feature upon being placed after the left feature does not cause a change in human muscle moment, the phonetic string sequence corresponding to the left feature and the right feature may be discarded.
  • the method includes finding a phonetically closest string for one or more substrings related to the valid sequence.
  • the phonetically closest string may be one or more phonetically matching strings.
  • the method includes replacing the one or more substrings related to the one or more valid phonetic string sequences with the phonetically closest string.
  • the method includes generating an abbreviation list based on the one or more valid phonetic string sequences and the phonetically closest string.
  • the abbreviation list may also be referred as one or more sets of abbreviations including one or more abbreviations as referred in the FIG. 1.
  • the abbreviation list may be generated by the phonetic string determining engine 230 incorporated in the abbreviation processor 228 referred in the FIG. 2.
  • the method includes performing a prioritization of the abbreviation list based on length, plurality of usage, and context, or the like.
  • the prioritization may be performed by the phonetic string determining engine 230.
  • the method includes suggesting a final output to the user.
  • the final output may include an abbreviation from the abbreviation list.
  • the abbreviation may be among the one or more abbreviations referred in the FIG. 1.
  • the one or more abbreviations may be suggested to the user by the display unit 214 referred in the FIG. 2.
  • FIG. 4 illustrates a flow diagram 400 depicting a process to analyze a context of an input word and generating a phonetic corresponding to at least one target word, in accordance with an embodiment of the present subject matter.
  • the at least one target word may be identified based on the input word.
  • the at least one target word may be same as the input word.
  • the context may be analyzed based on the input word, to determine the at least one target word for generating one or more abbreviations.
  • the phonetic may be generated.
  • the process may be executed by the context analysis engine 216, and the word generator 218, and the phonetic generator 220 referred in the FIG. 2.
  • the method includes receiving an input from the user.
  • the input may be the input word.
  • the input may be one of oral, hand-written, and typed.
  • the input may be received in any of a number of languages.
  • the input may be one of a complete word, and an incomplete word of a statement intended by the user.
  • the method includes determining the context of the statement.
  • the context may be determined through a number of techniques. Examples of the number of techniques may include, but are not limited to, GloVe (Global Vectors), GPT, word2vec or the like.
  • the input may be "He was wearing a whi".
  • a co-occurrence matrix may be calculated.
  • “wearing” shall be related to a men's apparel name.
  • the apparel may be accompanied by an adjective such as "white”.
  • a probabilistic model may be used.
  • a formulation from the Glove may be used as mentioned below:
  • P ik may denote a probability of watching words "w i " and "u k " with one another.
  • a word determiner module may be configured to determine a target word based on an input word.
  • the target word as explained herein, may be either the same as the input word or a next predicted word determined based on the input word.
  • the input word may be a complete word or a partial word.
  • the target word may be a complete input word, as determined based on a partial input word.
  • the word determiner module may be configured to analyze a context of the statement/sentence, in which the input word is being fed. In another example, the word determiner module may be configured to determine the target without analyzing the context of the statement. For instance, the word determiner module may determine a complete input word as the target word based on a partial input word or in another example, the target word may be considered same as the input word.
  • the word determiner module may be the word generator 218 and the next target possible word may be the at least one target word as referred in the FIG. 1.
  • the method includes generating the phonetic related to the at least one word through the phonetic generator 220.
  • the phonetic related to the at least one word may be extracted from a dictionary of words, an audible database related to the user to understand the phonetic based on a stress placement detection in audible data.
  • the phonetic may include one or more phonemes.
  • FIG. 5a illustrates an operational flow diagram 500a depicting a process for a phonetic string sequence processing, in accordance with an embodiment of the present subject matter.
  • the phonetic string sequence processing may be based on generating at least one feature vector and determining one or more valid phonetic string sequences. Further, the process may be performed by the phonetic string sequence processor 222 as referred in the FIG. 2.
  • the phonetic string sequence processor 222 may be configured to divide a phonetic related at least one target word into one or more valid phonetic string sequences.
  • the one or more valid phonetic string sequences may be defined as a sequence where each substring division takes place based on articulation such as a change in muscle of the user upon speaking.
  • the phonetic may be received as an input at the phonetic string sequence processor 222.
  • the phonetic may be of length "n" such that "n-1" places are identified for dividing the phonetic.
  • phonetic string sequences may be possible.
  • the process includes an articulation based classification of one or more phonemes related to the phonetic into one or more vowels and one or more non-vowels. Further, the process includes classifying the one or more vowels based on a height of a tongue (low/non-low), a tenseness (tense/lax) and a backness (round/non-round - roundness of lips). Furthermore, the process includes classifying the one or more consonants based on a place of articulation (coronal/non-coronal), a manner (obstruent/sonorant) and a voice (voice/voiceless). Further, the process includes sequential information related to sequence of one or more phonemes that comprises order of string.
  • the process includes observing a number of modalities/features for the validating AI module.
  • the number of modalities/features may include:
  • - articulation information 3 for vowels: roundness of lips, for consonants - voiced/voiceless, and
  • x 0 [x 01 , x 02 , x 03 , x 04 , x 05 ].
  • x 1 -x n may be similarly defined.
  • the process may include additional articulatory information or related information.
  • the present articulation information may contain most relevant information such that the feature matrix may be generalized as X - of n x m dimension, mathematically be defined as:
  • FIG. 5b illustrates an operational flow diagram 500b depicting a process for generating the articulatory feature generation of white may be "waIt", and length may be 4.
  • the process includes a dynamic sequence selection for determining the one or more valid phonetic string sequences.
  • the dynamic sequence selection may be performed by the dynamic sequence selector 226 as referred in the FIG. 2.
  • the process includes receiving the matrix x as an input.
  • the matrix x may comprise at least one feature vector as referred in the FIG. 1.
  • the complete articulatory feature matrix (also referred as a set of articulatory feature vector in Figure 5b.) may represent output of encoding based on feature matrix X(also referred as a set of feature vector in Figure 5b.).
  • articulation information of phoneme w, a, I, t may be mapped to numerical value.
  • phoneme type of one or more phoneme may be mapped to numerical value.
  • the process further includes receiving a previous iteration output as a feedback.
  • the dynamic feature selection may take the complete articulatory feature matrix and generates two feature vectors left(L) and right(R).
  • the complete articulatory feature matrix may be the at least one feature vector in the form of a matrix.
  • the left feature and the right feature may be generated from the feature vector. Further, based on the previous iteration output, the next iteration feature vectors are selected. If the previous feature vectors (L(k-1) and R(k-1)) are invalid sequence, the further combinations related to the previous feature vectors may not be considered to be the one or more valid phonetic string sequences. In an embodiment, the dynamic feature selection may further reduce a processing complexity of the phonetic string sequences validation by removing invalid combinations in each step.
  • FIG. 5c illustrates an operational flow diagram 500c depicting a role of dynamic feature selector may be to select the left feature (L[k]) and the right feature (R[k]) for the AI model.
  • the dynamic feature selector may be the dynamic sequence selector 226 and the AI model may be the phonetic string sequence processor 222.
  • the left feature (L[k]) and the right feature (R[k]) may be selected based on an output related to a previous iteration of the AI model (y[k-1]), where k represents the iteration number.
  • the AI phonetic string sequence validation may be configured to process 16-32 possible sequences such that there may be more than one possible target contextual words, leading to even more complexity.
  • one possible target contextual words may be the at least one target word.
  • the process includes feeding the features based on the output.
  • FIG. 5d illustrates an operational flow diagram 500d depicting a process for feeding the features to the dynamic sequence selector.
  • the features may be the left feature and the right feature.
  • L(k) / R(k) may represent whether an R(k) pronunciation feature coming after an L(k) pronunciation feature may cause change in speaking muscles or not.
  • the change in speaking muscles may be similar to change in muscle movement upon placing the right feature after the left feature.
  • the dynamic sequence selector 226 may be configured to store the left feature and the right feature in a left feature stack and a right feature stack based on the output of the previous iteration.
  • the left feature and the right feature may be a part of the complete feature vector X.
  • the dynamic sequence selector 226 may be configured to dynamically select the features probable to be valid. Based on the previous input, feature combinations classified as invalid may not be used again in future further iterations to save processing energy.
  • the process includes performing an AI phonetic string sequence validation for generating the one or more valid phonetic string sequences.
  • the AI phonetic string sequence validation may be performed by the dynamic sequence selector 226 also referred in the FIG. 2.
  • FIG. 5e illustrates an operational flow diagram 500e depicting a process for the dynamic sequence validation.
  • the process includes receiving the left feature and the right feature as an input. Based on the input, the process includes generating an output determining a probability of change in articulatory muscles when R feature appears after L.
  • the process for determining the probability of change may be based on a machine learning model such as a binary logistic regression model.
  • a cost function is defined as:
  • x is the input feature vector including articulatory information.
  • the output of each iteration, y[k], may be used to generate a cumulative output for the next model-abbreviation processor.
  • the abbreviation processor may be the abbreviation processor 228 as referred in the FIG. 2.
  • a matrix Y may be stored for maintaining the one or more valid phonetic string sequences.
  • the matrix Y may be expressed as:
  • v number of valid sequences, is a (1 x n-1) vector of 1 or 0 specifying "i th " particular position as valid or invalid division point.
  • Y may be (v x n-1) matrix.
  • valid phonetic string sequences of phonetic of phonetic "waIt” may comprises (1 0 0), (0 0 1) and (1 0 1).
  • Valid phonetic string sequences (1 0 0) may correspond with "w-aIt”
  • valid phonetic string sequences (0 0 1) may correspond with "waI-t”
  • valid phonetic string sequences (1 0 1) may correspond with "w-aI-t”.
  • FIG. 6a illustrates an operational flow diagram 600a depicting a method for processing one or more abbreviations, in accordance with an embodiment of the present subject matter.
  • the processing of the one or more abbreviations may be performed by the abbreviation processor 228 as referred in the FIG. 2.
  • the abbreviation processor includes the phonetic string determining engine 230.
  • the abbreviation processor 228 may be configured to use an input of the AI phonetic string sequence validation and generate one or more contextual and phonetically similar abbreviations.
  • the one or more contextual and phonetically similar abbreviations may be referred as the one or more abbreviations from one or more sets of abbreviations as referred in the FIG. 1.
  • the abbreviation processor 228 may be configured to receive the one or more valid phonetic string sequences as the input.
  • the method includes determining a phonetically closest string.
  • FIG. 6b illustrates an operational flow diagram 600b for determining the phonetically closest string.
  • the phonetic string determining engine 230 may be configured to determine the phonetically matching string.
  • the method includes obtaining articulation feature information (P) of a substring.
  • the articulatory information of the substring may be derived using International Phonetic Alphabet (IPA) standards.
  • the articulation feature information (P) may be encoded to a numerical vector as articulatory information are categorical.
  • the numerical vector may correspond to a set of numerical values.
  • a substring input may be "waI" and articulatory information of w, articulatory information of a and articulatory information of I may be encoded to numerical vector as articulation information are categorical.
  • P(waI) may comprises numerical vector of articulatory information of w, articulatory information of a and articulatory information of I.
  • Articulatory information of a consonant contains a consonant, a place of articulation, a manner of articulation or the like.
  • articulatory information of vowel may include a vowel, a position, a height of tongue or the like.
  • the method includes locating the articulation feature information in an N dimensional vocabulary space V.
  • the N dimensional vocabulary space V may be present in a word database.
  • the FIG. 6c illustrates a diagram 600c depicting the word database including the vocabulary space V.
  • the vocabulary space may include dictionary words and a user's personal words may further be added.
  • the articulatory feature vector P is a point of the vocabulary space. In an embodiment, the point may be located by using the feature information as coordinates of the vocabulary space.
  • the feature information of the substring war may lie at a place surrounded by similar sounding strings such as the phonetically matching strings.
  • the articulatory feature vector P representing feature information for the substring may be made as an N dimensional by padding with 0s.
  • the method includes identifying an N dimensional sphere with center at the P in the vocabulary space.
  • the feature vector P may be taken as a center point for a n-sphere, as follows:
  • R represents any point of vocabulary space and pi represents i-th coordinate of P.
  • the value of R may be selected based on the one or more phonetically matching strings required.
  • the value of R can be chosen based on the number of phonetically similar substrings required. Keeping it too large will generate erroneous abbreviations, which may not be phonetically matching. So, this has to be tuned, but may be kept small.
  • FIG. 6d illustrates a diagram 600d depicting an identification of the N dimensional sphere.
  • the method Upon identifying the N dimensional sphere, the method includes obtaining a cosine similarity of P with all the points within the sphere. The method includes obtaining the cosine similarity with P for all vectors lying within the n-sphere (Q), to find a match score or a cosine distance in the vocabulary space. The score may be utilized to prioritize the abbreviation while suggestion.
  • an array "valid" including one or more phonetically matching strings for each of the substrings of each valid phonetic string sequence is obtained.
  • the method includes forming the abbreviation based on the array "valid"
  • the abbreviation may be one or more abbreviation generated from one or more sets of abbreviations, as described above in the FIG. 1.
  • formation of the abbreviation may include determining possible combinations of the matching strings
  • matching strings are determined. For instance, for substring "waI”, “y” and “why" may be determined as phonetically closest or matching strings and for substring "t" determined matching string may be "t”.
  • one or more abbreviation of a valid phonetic string sequence "waI-t" may be comprises at least one of "yt” and "whyt".
  • an abbreviation may be determined by taking one or more combination of matching strings determined above.
  • FIG. 6e illustrates a diagram 600e depicting formation of the abbreviation, according to an embodiment of the present disclosure.
  • an array "valid” as described above may be provided as input.
  • valid is represented by an 'i, j" representation as presented in the diagram.
  • W represents an abbreviation list that may include a number of concatenated strings having phonetically similar strings to each substrings of the valid phonetic string sequence "waI-t".
  • the method includes prioritization of the one or more abbreviations.
  • the prioritization may be based on computing a length of each string and sorting based on the length.
  • the length may be corresponding to each of the one or more strings and sorting may be based on length of strings.
  • the prioritization may include grouping based on same length of strings.
  • N groups may be formed for further checking a plurality of usage.
  • the method includes generating a selection based on the group in a sequential manner such as a group may be explored first and random selection of string from the group may be suggested to the user.
  • FIG.6f illustrates a diagram depicting probability of usage of one or more abbreviation of a phonetic, according to an embodiment of the present disclosure.
  • the generated abbreviations may be prioritized.
  • a probability distribution may be formed that includes p(v) and v.
  • p(v) represents the probability of usage of a given abbreviation "v”.
  • v(a) may represent abbreviation "w8" of phonetic "waIt”
  • v(b) may represent abbreviation "yt" of phonetic "waIt”.
  • p(v) may be determined based on history of usage of abbreviations by the user.
  • the prioritization may be based on the formula:
  • u t is the confidence score
  • p(v t ) is the probability of choosing abbreviation v at t th trial
  • a t represents action of user on t th trial, which is decided based on maximum of p(v t ) (above likelihood) and u t (confidence score).
  • the confidence score is used for determining the action using the maximum p(v t ) + u t .
  • the confidence score may be determined based on the following formula:
  • t is the total number of trials and n t is the total number of times action a is selected.
  • p(v t ) + u t For each iteration we compute p(v t ) + u t .
  • the UCB formula which is based on reinforcement learning techniques may be used in an exemplary implementation of the present disclosure, for determining the system confidence.
  • other techniques for suggesting the abbreviations based on usage history may be implemented.
  • FIG. 7a illustrates a detailed architecture 700a of a method to generate one or more abbreviations corresponding to at least one target word, in accordance with an embodiment of the present subject matter.
  • the one or more abbreviations may be generated upon applying one or more techniques on at the least one target word in response to receiving an input word.
  • Examples of the one or more technique may include, but are not limited to, an AI technique, and a ML technique.
  • the detailed architecture may include a system for generating the one or more abbreviations.
  • the system may be the system 202 as referred in the FIG. 2.
  • the context analysis AI engine may be configured to receive an input provided by a user.
  • the input may be the input word as referred in the FIG. 1.
  • the input word may be one of a complete word, and incomplete word.
  • the input may be in the form of a speech or a voice. Further, the input may be converted into a text through text generator.
  • the context analysis AI engine may include a context analysis module and a word generator module.
  • the context analysis module may be the context analysis engine 216 and the word generator module may be the word generator 218 as referred in the FIG. 2.
  • the context analysis module may be configured to understand a linguistic context related to the word.
  • the linguistic context may be a context related to the input word as referred in the FIG. 1.
  • a context analysis may be performed to predict a possible nearest matching word.
  • the word generator module may be configured to generate an appropriate word based on the context.
  • an output may be at least one possible complete target word.
  • the possible nearest matching word, the appropriate word, and the at least one possible complete target word may be at the least one target word as referred in the FIG. 1. Further, the method may proceed towards step 704a.
  • the phonetic generator may be configured to generate a phonetic related to the at least one target word received from a generative AI engine.
  • the generative AI engine may be the context analysis AI engine.
  • the phonetic generator may be the phonetic generator 220 as referred in the FIG. 2. Upon generating the phonetic, the method may proceed towards step 706a.
  • a phonetic string sequence processor may be configured to generate at least one feature vector and one or more valid phonetic string sequences related to the at least one target word.
  • the phonetic string sequence processor may include a feature generator and a dynamic sequence validator.
  • the phonetic string sequence processor may be the phonetic string sequence processor 222, the feature generator may be the feature generator 224.
  • the feature generator may be configured to process the incoming phonetic of target word from the phonetic generator.
  • the incoming phonetic may be the phonetic as referred in the FIG. 1.
  • the feature generator may be configured to classify the at least one target word.
  • the processing and the classification may be performed by an articulation classification unit incorporated in the feature generator.
  • a phoneme sequence encoder incorporated in the feature generator may be configured to encode categorical data to a number of numeral values.
  • the categorical data may include one or more vowels and one or more consonants related to one or more phonemes of the phonetic as referred in the FIG. 1.
  • the categorical data may include phoneme type and articulation information related to one or more phonemes of the phonetic as referred in the FIG. 1.
  • the method may proceed towards step 708a at the dynamic sequence validator.
  • the dynamic sequence validator may be configured to determine one or more valid phonetic sequence strings.
  • the dynamic sequence validator may include a dynamic feature selector and an AI phonetic string sequence validator module.
  • the dynamic feature selector may be the dynamic sequence selector 226 as referred in the FIG. 2.
  • the dynamic sequence selector may be configured to select a number of required features from an entire feature matrix through a dynamic feature selector feedback module incorporated in the dynamic sequence selector.
  • the entire feature matrix may be referred as the feature vector and the number of required features may be a left feature and a right feature.
  • the number of required features may be forwarded to an intelligent phonetic string sequence module.
  • the intelligent phonetic string sequence module may be the AI phonetic string sequence validator module.
  • the intelligent phonetic string sequence validation module may be configured to discard invalid one or more phonetic string sequences.
  • at least one phonetic string sequence not discarded may be amongst one or more valid phonetic string sequences as referred in the FIG. 1.
  • the method may proceed towards an abbreviation processor at the step 710a.
  • the abbreviation processor may be configured to determine one or more abbreviation related to the at least one target word.
  • the abbreviation processor may be the abbreviation processor 228 as referred in the FIG. 2.
  • the abbreviation processor 228 may include a phonetic closest string determiner module, an abbreviation formation module, an abbreviation prioritization module, and an abbreviation suggestive module.
  • the phonetic closest string determiner module may be referred as the phonetic string determining engine 230.
  • the phonetic closest string determiner module may be configured to find one or more closest phonetically matching strings using cosine similarity.
  • the one or more closet phonetically matching strings may be referred as one or more phonetically matching strings.
  • the abbreviation forming module may be configured to combine the one or more phonetically matching strings to form the one or more abbreviations.
  • the abbreviation prioritization module may be configured to prioritize the one or more abbreviations based on a match score, a length of the abbreviation, the context analysis and a number of desired features.
  • a feedback reinforcement AI module may be configured to update a confidence score of an abbreviation suggestion based on a user feedback. For any action performed by user on suggested abbreviation, the feedback reinforcement AI module may be configured to increase or decrease the confidence score of the abbreviation. In an embodiment, the feedback reinforcement AI module may be in a two-way communication with a media device for displaying the one or more abbreviations and receive the user feedback.
  • the method for generating the one or more abbreviations may be processed at a higher speed upon utilizing one of a Graphical Processing Unit (GPU), a Neural Processing unit (NPU), and an AI engine while performing the steps 702a through 712a.
  • GPU Graphical Processing Unit
  • NPU Neural Processing unit
  • AI engine an AI engine while performing the steps 702a through 712a.
  • data related to each component of the detailed architecture 700a may be stored in a knowledgebase/database in the media device.
  • the knowledgebase/database may be the database present in the memory 206 as referred in the FIG. 2.
  • the data may be the data 208 referred in the FIG. 2.
  • FIG. 7b illustrates another detailed architecture 700b of the method to generate one or more abbreviations corresponding to the at least one target word, in accordance with an embodiment of the present subject matter.
  • the other detailed architecture 700b may be an embodiment of the detailed architecture 700a.
  • the other detailed architecture 700b may include the number of components incorporated in the detailed architecture 700a, a standard dictionary word validator, and ordered substring formation module.
  • the method upon receiving an input as an input text in the form of one of the text and the speech, at step 702a, the method includes validating whether the input text is a dictionary word or a non-standard word.
  • the input text may be the input word as referred in the FIG. 7a.
  • the method includes performing steps 702a through 712a as referred in the FIG. 7a.
  • the method includes forming a number of substrings corresponding to the word. Further, upon formation of the number of substrings, the method includes validating the number of substrings. Continuing with the above embodiment, the method includes performing steps 704a through 712a for each of the number of substrings.
  • FIG. 7c illustrates another detailed architecture 700c of the method to generate the one or more abbreviations corresponding to the at least one target word, in accordance with an embodiment of the present subject matter.
  • the other detailed architecture 700c may be an embodiment of the detailed architecture 700a.
  • the other detailed architecture 700c may include the number of components incorporated in the detailed architecture 700a, 700b, an Automatic Speech Recognizer (ASR), a language detector, and a Voice Activity Detector (VAD).
  • ASR Automatic Speech Recognizer
  • VAD Voice Activity Detector
  • the method includes the ASR for converting a speech or a voice input.
  • the language detector may be configured to detect the language of the input.
  • the ASR may be configured to forward a converted voice input to the word generator module for further processing.
  • the articulation classification unit, the dynamic sequence validator and the phonetic closest string determiner may be incorporated on a server.
  • FIG. 8a illustrates a use case 800a depicting an existing scenario of a device understanding context associated with one or more words.
  • a keyboard of the device may suggest a contextual word, auto-complete the word, or suggest GIFs/Emoji.
  • FIG. 8b illustrates a use case 800b depicting a scenario of the device configured to understand a phonetic, in accordance with an embodiment of the present subject matter.
  • the use case 800b may be based on one or more of AI based techniques.
  • the present scenario depicts that the device is configured to suggest new abbreviations to concise a chat and also preserve the phonetics associated with the one or more words.
  • FIG. 9 illustrates a process 900 for generating one or more abbreviations corresponding to at least one target word, in accordance with an embodiment of the present subject matter.
  • the at least one target word may be same as the input word.
  • the process includes determining a context of an input word received as an input.
  • the context may be determined for generating one or more contextual target words.
  • the one or more contextual target words may be referred as the at least one target word as referred in the FIG. 1.
  • the process includes generating a phonetic associated with the one or more contextual target words.
  • the process includes performing phonetic string sequence processing.
  • at least one feature vector and one or more phonetic string sequences may be generated related to the at least one target word.
  • the process includes dynamic sequence validation upon the one or more phonetic string sequences for generating one or more valid phonetic strings sequences.
  • the process includes performing abbreviation processing on the one or more valid phonetic strings sequences.
  • the abbreviation processing may include determining a phonetically closest string, forming of one or more abbreviations, and prioritization of the one or more abbreviations based on a user feedback.
  • the process includes generating the one or more abbreviations as an output.
  • FIG. 10a illustrates another use case 1000a depicting an existing scenario of a user exceeding out of character limit.
  • the user may seek to post an exact statement of an eye-witness as a report.
  • the report may exceed out of the character limit of an application using which the user is attempting to post the report.
  • FIG. 10b illustrates a use case 1000b depicting a scenario of the user avoiding to exceed the character limit by using a one or more abbreviations, in accordance with an embodiment of the present subject matter.
  • the user may post the exact eye witness report with a system capable of suggesting the one or more abbreviations incorporated in a device to help the user post within the character limit.
  • the system may be the system 202 referred in the FIG. 2.
  • FIG. 11 illustrates another use case 1100 depicting a user using one or more abbreviations for typing during an online lecture, in accordance with an embodiment of the present subject matter.
  • a user device of the user may implement the system 202 and accordingly generate the one or more abbreviations, as described herein.
  • FIG. 12 illustrates another use case 1200 depicting a user taking a computer based language test, in accordance with an embodiment of the present subject matter.
  • the user may be listening to a speech and accordingly, may be typing the speech on a user device, for example, a smartphone.
  • the device may be configured to generate one or more abbreviations related to one or more words in the speech. The one or more abbreviations may be generated for assisting the user in typing the speech fast.
  • the device may generate the one or more abbreviations through the system 202 incorporated in the device.
  • FIG. 13 illustrates another use case 1300 depicting a user using one or more abbreviations generated at a device for short-hand writing, in accordance with an embodiment of the present subject matter.
  • the device may be configured to suggest phonetically short form of each word user intends to type such that the user need not remember short-hand writing norms.
  • the device may generate the one or more abbreviations through the process depicted by the operation flow diagram 300.
  • FIG. 14 illustrates another use case diagram 1400 depicting a user typing notes during a meeting at a device, in accordance with an embodiment of the present subject matter.
  • the device may suggest the one or more abbreviations through the process depicted by the operation flow diagram 300 based on phonetics of one or more words typed by the user.
  • FIG. 15 illustrates another use case diagram 1500 depicting sharing of one or more abbreviations generated at a device in response to receiving one or more words, in accordance with an embodiment of the present subject matter.
  • a community receiving the one or more abbreviations may be unaware of certain abbreviations usage which may create confusion.
  • the one or more abbreviations may be used across community.
  • the user's device may put the one or more abbreviations in brackets instead of replacing the original word.
  • the other side device may be capable of using the one or more abbreviations to include in a vocabulary space of the user at a receiving end.
  • the device may generate the one or more abbreviations through the process depicted by the operation flow diagram 300.
  • FIG. 16 illustrates another use case 1600 depicting a device suggesting one or more words based one or more abbreviations, in accordance with an embodiment of the present subject matter.
  • a user uses the one or more abbreviations such as E.g. sum1, in2, r, yesterde or the like.
  • the device may be capable of decoding/expanding the one or more abbreviations.
  • FIG. 17 illustrates another use case 1700 depicting a voice assistant utilizing one or more abbreviations, in accordance with an embodiment of the present subject matter.
  • a number of voice assistants may be used to type a message for a user.
  • the voice assistants may use abbreviated words instead of typing whole words.
  • the one or more abbreviations may be generated through the system 202.
  • FIG. 18 illustrates another use case 1800 depicting a user sending a message to another user, in accordance with an embodiment of the present subject matter.
  • the user is texting sending the message to the other user.
  • the user is using system suggestive abbreviations which the user may understand.
  • the other user uses a device that does not use a smarter system and the other user does not know that neighbor can be written as "nbr” and white can be written as "yt".
  • the message from the user to the other user is sent like "neighbor(nbr)... Looking (lukin) ... white (yt) ".
  • the one or more abbreviations may be generated through the system 202 incorporated in the device used by the user.
  • FIG. 19 illustrates another use case 1900 depicting a device configured to translate a message in one or more languages, in accordance with an embodiment of the present subject matter.
  • FIG. 19 illustrates another use case 1900 depicting a device configured to translate a message in Germany and English.
  • FIG. 20 illustrates a process 2000 for suggesting and/or expanding an abbreviation, in accordance with an embodiment of the present subject matter.
  • the abbreviation may be received as in input text.
  • the abbreviation may be one or more abbreviations as referred in the FIG. 1.
  • the abbreviation upon expansion, may be converted into a word.
  • the word may be an input word as referred in the FIG. 1.
  • the process includes receiving the input text from a user.
  • the input text may be one of a complete word, and an incomplete word.
  • the process includes performing a context analysis on the input text for determining a context related to the input text.
  • the context may be determined for further generating at least one target word related to the input text.
  • the at least one target word may be same as the input text.
  • the process includes determining whether the at least one target word is a standard dictionary word or not. In an embodiment, where it is determined that the at least one target word is the standard dictionary word, the process includes performing steps 306 through 322 as referred in the FIG. 3 for further generating one or more abbreviations related to the at least one target word. In an embodiment, where it is determined that the at least one word is not the standard dictionary word, the process may proceed towards step 2008.
  • the process includes determining whether the input text is a standard abbreviation or not. In an embodiment, where it is determined that the input text is the standard abbreviation, the process terminates. In an embodiment, where it is determined that the input text is not the standard abbreviation, the process may proceed towards step 2010.
  • the process includes validating whether each substring sequences of the one or more ordered substring sequences is the dictionary word or not. In an embodiment, where it is determined that the one or more ordered substring sequences are not the dictionary word, the one or more ordered substring sequences may be discarded. In an embodiment, where it is determined that the one or more ordered substring sequences are the dictionary word, the process may proceed towards step 2014.
  • the process includes generating a phonetic associated with the one or more ordered substring sequences. Furthermore, at step 2016, the process includes processing the phonetic for further expanding the abbreviation.
  • FIG. 21 illustrates a representative architecture 2100 to provide tools and development environment described herein for a technical-realization of the implementation in preceding figures through a virtual personal assistance (VPA) based computing device.
  • FIG. 21 is merely a non-limiting example, and it will be appreciated that many other architectures may be implemented to facilitate the functionality described herein.
  • the architecture may be executing on hardware such as a computing machine 200 of FIG. 2 that includes, among other things, processors, memory, and various application-specific hardware components.
  • the architecture 2100 may include an operating-system, libraries, frameworks or middleware.
  • the operating system may manage hardware resources and provide common services.
  • the operating system may include, for example, a kernel, services, and drivers defining a hardware interface layer.
  • the drivers may be responsible for controlling or interfacing with the underlying hardware.
  • the drivers may include display drivers, camera drivers, Bluetooth drivers, flash memory drivers, serial communication drivers (e.g., Universal Serial Bus (USB) drivers), Wi-Fi drivers, audio drivers, power management drivers, and so forth depending on the hardware configuration.
  • USB Universal Serial Bus
  • a hardware interface layer includes libraries which may include system libraries such as file-system (e.g., C standard library) that may provide functions such as memory allocation functions, string manipulation functions, mathematic functions, and the like.
  • the libraries may include API libraries such as audio-visual media libraries (e.g., multimedia data libraries to support presentation and manipulation of various media format such as MPEG4, H.264, MP3, AAC, AMR, JPG, PNG), database libraries (e.g., SQLite that may provide various relational database functions), web libraries (e.g. WebKit that may provide web browsing functionality), and the like.
  • system libraries such as file-system (e.g., C standard library) that may provide functions such as memory allocation functions, string manipulation functions, mathematic functions, and the like.
  • API libraries such as audio-visual media libraries (e.g., multimedia data libraries to support presentation and manipulation of various media format such as MPEG4, H.264, MP3, AAC, AMR, JPG, PNG), database libraries (e.g., SQLite that may provide
  • a middleware may provide a higher-level common infrastructure such as various graphic user interface (GUI) functions, high-level resource management, high-level location services, and so forth.
  • GUI graphic user interface
  • the middleware may provide a broad spectrum of other APIs that may be utilized by the applications or other software components/modules, some of which may be specific to a particular operating system or platform.
  • module used in this disclosure may refer to a certain unit that includes one of hardware, software and firmware or any combination thereof.
  • the module may be interchangeably used with unit, logic, logical block, component, or circuit, for example.
  • the module may be the minimum unit, or part thereof, which performs one or more particular functions.
  • the module may be formed mechanically or electronically.
  • the module disclosed herein may include at least one of ASIC (Application-Specific Integrated Circuit) chip, FPGAs (Field-Programmable Gate Arrays), and programmable-logic device, which have been known or are to be developed.
  • ASIC Application-Specific Integrated Circuit
  • FPGAs Field-Programmable Gate Arrays
  • programmable-logic device which have been known or are to be developed.
  • the architecture 2100 depicts an aggregation of VPA based mechanisms and ML/NLP based mechanism in accordance with an embodiment of the present subject matter.
  • a user-interface defined as input and interaction 2101 refers to overall input. It can include one or more of the following -touch screen, microphone, camera etc.
  • a first hardware module 2102 depicts specialized hardware for ML/NLP based mechanisms. In an example, the first hardware module 2102 comprises one or more of neural processors, FPGA, DSP, GPU etc.
  • a second hardware module 2112 depicts specialized hardware for executing the VPA device-related audio and video simulations.
  • ML/NLP based frameworks and APIs 2104 correspond to the hardware interface layer for executing the ML/NLP logic based on the underlying hardware.
  • the frameworks may be one or more or the following - Tensorflow, cafe, NLTK, GenSim, ARM Compute etc.
  • VPA simulation frameworks and APIs 2114 may include one or more of - VPA Core, VPA Kit, Unity, Unreal etc.
  • a database 2108 depicts a pre-trained voice feature database.
  • the database 2108 may be remotely accessible through cloud.
  • the database 2108 may partly reside on cloud and partly on-device based on usage statistics.
  • Another database 2118 refers the speaker enrollment DB or the voice feature DB that will be used to authenticate and respond to the user.
  • the database 2118 may be remotely accessible through cloud. In other example, the database 2118 may partly reside on the cloud and partly on-device based on usage statistics.
  • a rendering module 2105 is provided for rendering audio output and trigger further utility operations as a result of user authentication.
  • the rendering module 2105 may be manifested as a display cum touch screen, monitor, speaker, projection screen, etc.
  • a general-purpose hardware and driver module 2103 corresponds to the computing device 200 as referred in FIG. 2 and instantiates drivers for the general purpose hardware units as well as the application-specific units (2102, 2112).
  • the NLP/ML mechanism and VPA simulations underlying the present architecture 2100 may be remotely accessible and cloud-based, thereby being remotely accessible through a network connection.
  • a computing device such as a VPA device may be configured for remotely accessing the NLP/ML modules and simulation modules may comprise skeleton elements such as a microphone, a camera a screen/monitor, a speaker etc.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • General Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Health & Medical Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Computational Linguistics (AREA)
  • General Health & Medical Sciences (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • Human Computer Interaction (AREA)
  • Biomedical Technology (AREA)
  • Computing Systems (AREA)
  • Molecular Biology (AREA)
  • Evolutionary Computation (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Data Mining & Analysis (AREA)
  • Biophysics (AREA)
  • Probability & Statistics with Applications (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Machine Translation (AREA)

Abstract

L'invention concerne un procédé de génération d'abréviations correspondant à au moins un mot cible. Le procédé comprend l'identification du ou des mots cibles se rapportant à un mot d'entrée. Le procédé comprend la détermination d'une phonétique associé au(x) mot(s) cible(s). Le procédé comprend la génération d'au moins un vecteur d'attributs d'après au moins un élément parmi le type de phonème, des informations d'articulation et une information séquentielle correspondant à un ou plusieurs phonèmes associés à la phonétique. En outre, le procédé comprend la détermination d'une ou de plusieurs séquences valides de chaînes phonétiques d'après le ou les vecteurs d'attributs. Le procédé comprend la détermination d'un ou de plusieurs ensembles d'abréviations correspondant à la ou aux séquences valides de chaînes phonétiques. Le procédé comprend en outre la génération d'une ou de plusieurs abréviations à partir de l'ensemble ou des ensembles d'abréviations en vue du remplacement du ou des mots cibles.
PCT/KR2021/018954 2020-12-14 2021-12-14 Procédés et systèmes de génération d'abréviations pour un mot cible WO2022131740A1 (fr)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
IN202011054259 2020-12-14
IN202011054259 2020-12-14

Publications (1)

Publication Number Publication Date
WO2022131740A1 true WO2022131740A1 (fr) 2022-06-23

Family

ID=82059273

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/KR2021/018954 WO2022131740A1 (fr) 2020-12-14 2021-12-14 Procédés et systèmes de génération d'abréviations pour un mot cible

Country Status (1)

Country Link
WO (1) WO2022131740A1 (fr)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20230297332A1 (en) * 2022-03-21 2023-09-21 Microsoft Technology Licensing, Llc Input table normalization systems

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2009041220A1 (fr) * 2007-09-26 2009-04-02 Nec Corporation Dispositif et programme de génération d'abréviation, et procédé de génération d'abréviation
US20180335313A1 (en) * 2016-01-11 2018-11-22 Alibaba Group Holding Limited Method and apparatus for obtaining abbreviated name of point of interest on map
CN110263123A (zh) * 2019-06-05 2019-09-20 腾讯科技（深圳）有限公司 机构名简称的预测方法、装置和计算机设备

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2009041220A1 (fr) * 2007-09-26 2009-04-02 Nec Corporation Dispositif et programme de génération d'abréviation, et procédé de génération d'abréviation
US20180335313A1 (en) * 2016-01-11 2018-11-22 Alibaba Group Holding Limited Method and apparatus for obtaining abbreviated name of point of interest on map
CN110263123A (zh) * 2019-06-05 2019-09-20 腾讯科技（深圳）有限公司 机构名简称的预测方法、装置和计算机设备

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
GRZEGORZ KONDRAK: "A new algorithm for the alignment of phonetic sequences", ASSOCIATION FOR COMPUTATIONAL LINGUISTICS, N. EIGHT STREET, STROUDSBURG, PA, 18360 07960-1961 USA, 29 April 2000 (2000-04-29) - 4 May 2000 (2000-05-04), N. Eight Street, Stroudsburg, PA, 18360 07960-1961 USA , pages 288 - 295, XP058251016 *
RAMA TARAKA: "Automatic cognate identification with gap-weighted string subsequences.", PROCEEDINGS OF THE 2015 CONFERENCE OF THE NORTH AMERICAN CHAPTER OF THE ASSOCIATION FOR COMPUTATIONAL LINGUISTICS: HUMAN LANGUAGE TECHNOLOGIES, ASSOCIATION FOR COMPUTATIONAL LINGUISTICS, STROUDSBURG, PA, USA, 5 June 2015 (2015-06-05), Stroudsburg, PA, USA, pages 1227 - 1231, XP055942754, DOI: 10.3115/v1/N15-1130 *

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20230297332A1 (en) * 2022-03-21 2023-09-21 Microsoft Technology Licensing, Llc Input table normalization systems

Similar Documents

Publication Publication Date Title
WO2021132927A1 (fr) Dispositif informatique et procédé de classification de catégorie de données
CN106598939B (zh) 一种文本纠错方法及装置、服务器、存储介质
WO2019117466A1 (fr) Dispositif électronique pour analyser la signification de la parole, et son procédé de fonctionnement
WO2022102937A1 (fr) Procédés et systèmes pour prédire des actions qui ne sont pas par défaut à l'égard d'énoncés non structurés
WO2022203167A1 (fr) Procédé de reconnaissance vocale, appareil, dispositif électronique et support de stockage lisible par ordinateur
WO2018034426A1 (fr) Procédé de correction automatique d'erreurs dans un corpus balisé à l'aide de règles pdr de noyau
WO2020105856A1 (fr) Appareil électronique pour traitement d'énoncé utilisateur et son procédé de commande
EP3908961A1 (fr) Dispositif électronique et procédé de commande de dispositif électronique associé
JP4930379B2 (ja) 類似文検索方法、類似文検索システム及び類似文検索用プログラム
WO2020263034A1 (fr) Dispositif de reconnaissance d'entrée vocale d'un utilisateur et procédé de fonctionnement associé
WO2020213842A1 (fr) Structures multi-modèles pour la classification et la détermination d'intention
WO2021225256A1 (fr) Dispositif électronique et son procédé de commande
WO2021029643A1 (fr) Système et procédé de modification d'un résultat de reconnaissance vocale
WO2020130447A1 (fr) Procédé de fourniture de phrases basé sur un personnage et dispositif électronique de prise en charge de ce dernier
WO2022131740A1 (fr) Procédés et systèmes de génération d'abréviations pour un mot cible
WO2022131450A1 (fr) Procédé de production de résumés de documents utilisant une couverture basée sur l'injection de bruit et une association de mots, et support d'enregistrement et dispositif pour le réaliser
WO2021040490A1 (fr) Procédé et appareil de synthèse de la parole
WO2012063981A1 (fr) Procédé et dispositif pour entrer rapidement un texte à l'aide d'un écran tactile
WO2022169172A1 (fr) Dispositif de traitement d'interrogation d'utilisateur et son procédé d'utilisation
Li et al. Applications of recurrent neural network language model in offline handwriting recognition and word spotting
WO2023101377A1 (fr) Procédé et appareil pour effectuer une diarisation de locuteur sur la base d'une identification de langue
WO2022139122A1 (fr) Dispositif électronique et son procédé de commande
WO2021230470A1 (fr) Dispositif électronique et son procédé de commande
WO2021054613A1 (fr) Dispositif électronique et procédé de commande de dispositif électronique associé
KR100910302B1 (ko) 멀티모달 기반의 정보 검색 장치 및 방법

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 21907044

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 21907044

Country of ref document: EP

Kind code of ref document: A1