Speech recognition method, device and equipment and computer storage medium

Through the analysis and matching screening of voice commands, the problems of single statements and typos in voice control are solved, and more efficient speech recognition and user experience improvement are achieved.

CN119028351APending Publication Date: 2024-11-26QINGDAO HAIER WASHING MASCH CO LTD +1
View PDF 0 Cites 3 Cited by

Patent Information

Application Number
CN202310612356.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2023-05-26
Publication Date
2024-11-26

Smart Images

  • Figure CN119028351A_ABST
    Figure CN119028351A_ABST
Patent Text Reader

Abstract

The invention belongs to the technical field of electric appliances, and particularly relates to a speech recognition method, which comprises the following steps of: acquiring a speech instruction of a user, analyzing the speech instruction into a text and a text length, comparing the text length with a preset threshold value, vectorizing the text to obtain word vectors when the text length is greater than the preset threshold value, and sequentially performing cosine similarity matching on the word vectors to obtain a speech recognition result. Performing semantic similarity matching and confidence selection to obtain a matching result; when the text length is not greater than a preset threshold value, performing strong matching strategy screening on the text to obtain a matching result, and determining a target control instruction according to the matching result; according to the method, the phenomena that the user voice instruction interaction time is long, the user cannot understand after speaking for many times and wrongly written characters occur in voice-to-text conversion are avoided, the voice recognition efficiency is improved, and the user experience is improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application belongs to the field of electrical technology, and specifically relates to a speech recognition method, device, equipment and computer storage medium. Background Art

[0002] With the continuous development and progress of clothing processing programs, voice interaction functions have received much attention. Therefore, whether the voice interaction mechanism can efficiently recognize instructions has become a major factor in users' selection of clothing processing equipment.

[0003] The current caching mechanism of clothing processing programs usually converts user query statements into a specific code and then stores it in memory for fast access. When the user makes the same query again, the system will first check whether there is already a cache in the memory. If so, the result will be returned directly. Otherwise, real-time processing and calculation will be performed.

[0004] However, the above method has problems when performing voice control, such as single expression, stiff and not colloquial, and does not support intention reasoning ability. The user voice command interaction time is long, the user cannot understand after speaking multiple times, and typos occur in the voice-to-text conversion, resulting in the inability to recognize commands. Summary of the invention

[0005] The present application provides a speech recognition method, apparatus, device and storage medium to solve the problems in the prior art of single and stiff speech during voice control, lack of colloquialism, lack of support for intention reasoning, long user voice command interaction time, inability to understand after repeated speech, and typos in speech-to-text conversion leading to command failure to be recognized.

[0006] In a first aspect, the present application provides a speech recognition method, comprising:

[0007] Get the natural language input by the user;

[0008] Parsing the natural language to obtain a text instruction, wherein the text instruction includes: text length and text;

[0009] According to the text length and a preset threshold, the text and the cache database are matched and screened to obtain a matching result;

[0010] According to the matching result, a target control instruction is determined.

[0011] Optionally, before obtaining the natural language input by the user, the method further includes:

[0012] Obtain historical voice interaction data from historical data;

[0013] A cache database is established according to the historical voice interaction data, and the cache database includes: a correspondence between historical voice instructions and control instructions.

[0014] Optionally, performing matching screening on the text and a cache database according to the text length and a preset threshold to obtain a matching result includes:

[0015] Determine whether the length of the text is greater than a preset threshold;

[0016] If so, performing similarity matching screening on the text and the cache database to obtain a matching result;

[0017] If not, a strong matching strategy is performed on the text and the cache database to obtain a matching result.

[0018] Optionally, performing similarity matching screening on the text and the cache database to obtain a matching result includes:

[0019] Performing word segmentation on the text to obtain at least one word vector;

[0020] A similarity match is performed based on the at least one word vector and the cache database to obtain a target historical voice instruction.

[0021] Optionally, performing similarity matching with the cache database to obtain a target historical voice instruction includes:

[0022] Processing the at least one word vector and the cache database by cosine similarity matching to obtain at least one first screening result;

[0023] Inputting the at least one first screening result into a convolutional neural network to obtain at least one second screening result, wherein the convolutional neural network is used to perform semantic similarity matching on the at least one first screening result;

[0024] The second screening result whose confidence level meets a preset condition among the at least one second screening result is used as the target historical voice instruction.

[0025] Optionally, the performing a strong matching strategy screening on the text and the cache database to obtain a matching result includes:

[0026] A target matching program is obtained, and matching processing is performed on the text and the cache database according to the target matching program to obtain a matching result.

[0027] Optionally, the target matching program includes: a one-to-one matching program, a rule template matching program, and a statistical method matching program. According to the target matching program, the text and the cache database are matched to obtain a matching result, including:

[0028] Perform one-to-one matching, rule template matching and statistical method matching on the text and the cache database at the same time to obtain sub-matching results output by each matching program;

[0029] The matching result is determined according to the sub-matching result.

[0030] In a second aspect, the present application provides a speech recognition device, comprising:

[0031] An acquisition module is used to acquire the natural language input by the user;

[0032] A processing module, used for parsing the natural language to obtain a text instruction, wherein the text instruction includes: text length and text;

[0033] The processing module is further used to perform matching screening on the text and the cache database according to the text length and a preset threshold to obtain a matching result;

[0034] A determination module is used to determine a target control instruction according to the matching result.

[0035] Optionally, the acquisition module is further used to acquire historical voice interaction data in the historical data;

[0036] The processing module is further used to establish a cache database according to the historical voice interaction data, and the cache database includes: a correspondence between historical voice instructions and control instructions.

[0037] Optionally, the speech recognition device further includes: a judgment module;

[0038] The judging module is used to judge whether the length of the text is greater than a preset threshold;

[0039] The processing module is specifically used to perform similarity matching screening on the text and the cache database to obtain a matching result;

[0040] The processing module is specifically used to perform strong matching strategy screening on the text and the cache database to obtain a matching result.

[0041] Optionally, the processing module is further used to perform word segmentation processing on the text to obtain at least one word vector;

[0042] The processing module is specifically used to perform similarity matching based on the at least one word vector and the cache database to obtain the target historical voice instruction.

[0043] Optionally, the processing module is specifically used to process the at least one word vector and the cache database by cosine similarity matching to obtain at least one first screening result;

[0044] The processing module is specifically used to input the at least one first screening result into a convolutional neural network to obtain at least one second screening result, and the convolutional neural network is used to perform semantic similarity matching on the at least one first screening result;

[0045] The processing module is specifically configured to use, among the at least one second screening result, a second screening result whose confidence level meets a preset condition as a target historical voice instruction.

[0046] Optionally, the acquisition module is further used to acquire a target matching program, and perform matching processing on the text and the cache database according to the target matching program to obtain a matching result.

[0047] Optionally, the processing module is specifically used to perform one-to-one matching, rule template matching and statistical method matching on the text and the cache database at the same time, and obtain sub-matching results output by each matching program;

[0048] The determination module is further configured to determine the matching result according to the sub-matching result.

[0049] In a third aspect, the present application provides a speech recognition device, comprising:

[0050] Memory;

[0051] processor;

[0052] Wherein, the memory stores computer-executable instructions;

[0053] The processor executes the computer-executable instructions stored in the memory to implement the speech recognition method as described in the first aspect and various possible implementations of the first aspect.

[0054] In a fourth aspect, the present application provides a computer storage medium, characterized in that the computer storage medium stores computer execution instructions, and when the computer execution instructions are executed by a processor, they are used to implement the nursing machine control method as described in the first aspect and various possible implementations of the first aspect.

[0055] The speech recognition method provided by the present application obtains the user's voice command, parses the voice command into text and text length, compares the text length with a preset threshold, and when the text length is greater than the preset threshold, vectorizes the text to obtain a word vector, and performs cosine similarity matching, semantic similarity matching, and confidence selection on the word vector in turn to obtain a matching result; when the text length is not greater than the preset threshold, the text is screened by a strong matching strategy to obtain a matching result, and the target control command is determined based on the matching result; the method avoids the long interaction time of the user's voice command, the phenomenon of not being able to understand after speaking multiple times, and typos in speech-to-text conversion, thereby improving the efficiency of speech recognition and enhancing the user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0057] Figure 1 This is the process of the speech recognition method provided by this application Figure 1 ;

[0058] Figure 2 This is the process of the speech recognition method provided by this application Figure 2 ;

[0059] Figure 3 A schematic diagram of the structure of the speech recognition device provided in this application;

[0060] Figure 4 A schematic diagram of the structure of the speech recognition device provided in this application.

[0061] The above drawings have shown clear embodiments of the present application, which will be described in more detail later. These drawings and text descriptions are not intended to limit the scope of the present application in any way, but to illustrate the concept of the present application to those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION

[0062] In order to make the purpose, technical solutions and advantages of this application clearer, the technical solutions in this application will be clearly and completely described below in conjunction with the drawings in this application. Obviously, the described embodiments are part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0063] The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in sequences other than those illustrated or described herein, for example.

[0064] In the embodiments of the present application, the words "exemplary" or "for example" are used to indicate examples, illustrations or descriptions. Any embodiment or design described as "exemplary" or "for example" in the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "exemplary" or "for example" is intended to present related concepts in a specific way.

[0065] The current caching mechanism of clothing processing programs usually converts user queries into a specific code and then stores it in memory for fast access. When the user makes the same query again, the system will first check whether there is already a cache in the memory. If so, the result will be returned directly. Otherwise, real-time processing and calculation will be performed.

[0066] However, the above method has problems when performing voice control, such as single expression, stiff and not colloquial, and does not support intention reasoning ability. The user voice command interaction time is long, the user cannot understand after speaking multiple times, and typos occur in the voice-to-text conversion, resulting in the inability to recognize commands.

[0067] The speech recognition method provided by the present application obtains the user's voice command, parses the voice command into text and text length, compares the text length with a preset threshold, and when the text length is greater than the preset threshold, vectorizes the text to obtain a word vector, and performs cosine similarity matching, semantic similarity matching, and confidence selection on the word vector in turn to obtain a matching result; when the text length is not greater than the preset threshold, the text is screened by a strong matching strategy to obtain a matching result, and the target control command is determined based on the matching result; the method avoids the long interaction time of the user's voice command, the phenomenon of not being able to understand after speaking multiple times, and typos in speech-to-text conversion, thereby improving the efficiency of speech recognition and enhancing the user experience.

[0068] The technical solution of the present application and how the technical solution of the present application solves the above-mentioned technical problems are described in detail below with specific embodiments. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of the present application will be described below in conjunction with the accompanying drawings.

[0069] Figure 1 This is the process of the speech recognition method provided by this application Figure 1 ;like Figure 1As shown, the speech recognition method provided in this embodiment includes:

[0070] S101: Acquire natural language input by a user.

[0071] Among them, natural language refers to the instructions issued by the user through spoken language. For example, the user can input natural language such as "It rained today, I stepped into a puddle on the road and got my clothes dirty".

[0072] When a user needs to interact with a clothing treatment device by voice, the user can express the need in a variety of ways, such as using a local language or a language other than the clothing treatment device key words. The most intuitive way is to use the clothing treatment device key words. After the user expresses the need, the clothing treatment device obtains the natural language sent by the user.

[0073] S102: parsing the natural language to obtain a text instruction, wherein the text instruction includes: text length and text.

[0074] After the natural language is acquired, the natural language is parsed to obtain a text instruction that can be "understood" by the clothing processing device. The text instruction includes text and length information corresponding to the text.

[0075] The purpose of this step of processing natural language by Jie Xin is to facilitate the subsequent clothing processing equipment to adopt different matching methods for different types of texts. At the same time, the text is conducive to conversion into vectors, which can be consistent with the language mode of the cache database, and can improve the efficiency and accuracy of speech recognition.

[0076] S103: According to the text length and a preset threshold, the text and the cache database are matched and screened to obtain a matching result.

[0077] The preset threshold is used to distinguish the text length corresponding to the text, and the preset threshold is a fixed value, and the preset threshold can be a relatively small threshold, for example. The matching result is used to determine the result that the clothing processing device recognizes the executable instruction after the natural language is processed.

[0078] The cache database is a pre-set database, which may store, for example, the correspondence between voice commands and control commands.

[0079] Matching screening may include, for example, similarity matching screening, random matching screening, and strong matching screening. In this step, a corresponding matching screening method may be selected for matching based on the relationship between the text length and a preset threshold, thereby obtaining a matching result. The present application does not impose any special restrictions on the specific implementation method of matching and screening the text and the cache database based on the text length and the preset threshold to obtain a matching result.

[0080] Optionally, a specific implementation method for obtaining a matching result is provided below, as follows:

[0081] Determine whether the text length is greater than a preset threshold; if so, perform similarity matching screening on the text and the cache database to obtain a matching result; if not, perform strong matching strategy screening on the text and the cache database to obtain a matching result.

[0082] Among them, similarity matching is used to indicate the similarity between the word vector obtained after the text is converted into a vector and the cache database. This can be the similarity of the word vector, the similarity of the text semantics, or the similarity of the shape. Strong matching strategy screening is used to indicate direct associative matching of the text. In order to improve the matching accuracy, the text length here is relatively short.

[0083] The purpose of this step of judging whether the text length is greater than a preset threshold is to clarify the method by which the corresponding text should be matched with the cache database; when the text length is greater than the preset threshold, the text needs to be firstly matched by cosine similarity to obtain a first matching result, and then the first matching result needs to be matched by semantic similarity to obtain a second matching result, and then the second matching result needs to be compared by confidence to obtain a final matching result; when the text length is not greater than the preset threshold, the text needs to be screened by at least one strong matching strategy at the same time to obtain a matching result.

[0084] S104: Determine a target control instruction according to the matching result.

[0085] Among them, the target control instruction is a control instruction executable by the clothing processing device, wherein the matching result is associated with the target control instruction. After determining the target control instruction according to the matching result, the clothing processing device controls the opening of the corresponding control according to the target control instruction to execute the clothing processing program.

[0086] For example, the matching result is "clothes are splashed with mud" and the target control instruction is "washing". At this time, the corresponding clothing processing program will be automatically started to start the laundry process.

[0087] The speech recognition method provided in this embodiment obtains natural language input by the user, parses the natural language, and obtains text instructions. The text instructions include: text length and text. According to the text length and a preset threshold, the text and a cache database are matched and screened to obtain a matching result. According to the matching result, a target control instruction is determined. The method avoids the long interaction time of user voice commands, the phenomenon of not being understood after speaking multiple times, and typos in voice-to-text conversion, thereby improving the efficiency of speech recognition and enhancing the user experience.

[0088] Figure 2 This is the process of the speech recognition method provided by this application Figure 2 .like Figure 2 As shown, in this embodiment Figure 1 Based on the embodiment, the speech recognition method is described in detail. The speech recognition method shown in this embodiment includes:

[0089] S201: Acquire historical voice interaction data in historical data.

[0090] Among them, the historical voice interaction data includes: natural language, text instructions corresponding to the natural language, control instructions, etc. The clothing processing device obtains the historical voice interaction data by acquiring the natural language input by the user in the historical data and the control instructions executed according to the natural language.

[0091] S202: Establishing a cache database according to the historical voice interaction data, wherein the cache database includes: a correspondence between historical voice instructions and control instructions.

[0092] The cache database is a database composed of historical voice interaction data that is manually stored in batches in the early stage, and the interactive data is automatically stored after going online. The cache database will associate historical voice commands with control commands to obtain the corresponding relationship between historical voice commands and control commands. When the input natural language is a historical voice command, the corresponding control command can be directly obtained and the corresponding operation can be performed.

[0093] For example, the historical voice command stored in the cache database is: "The clothes are splashed with mud and need to be washed", and the control instruction corresponding to the historical voice command is "execute the washing program".

[0094] S203: Acquire the natural language input by the user.

[0095] Among them, step S203 is similar to the above step S101 and will not be repeated here.

[0096] S204: parsing the natural language to obtain a text instruction, wherein the text instruction includes: text length and text.

[0097] Among them, step S204 is similar to the above-mentioned step S102 and will not be repeated here.

[0098] S205: Determine whether the text length is greater than a preset threshold; if so, execute step S206, if not, execute step S208.

[0099] The preset threshold value may be set to 8-10 characters, for example. The reason for setting the preset threshold value to a smaller value is that when the text length is shorter, the matching result can be obtained more quickly, thereby improving the matching efficiency.

[0100] S206: Perform word segmentation processing on the text to obtain at least one word vector.

[0101] Among them, word vector is the data in vector form obtained after word segmentation of text.

[0102] The process of word segmentation can be as follows: Sentence A: This dress is a little dirty, I want to change one; Sentence B: This dress is not very clean, I want another one. Step 1: Word segmentation, Sentence A: This / dress / is / a little / dirty, I want / change / one; Sentence B: This / dress / is / not / clean, I want / another / one. Step 2: List all the words, this / dress, a little / dirty, I want, change, one, not / clean, another. Step 3: Calculate the word frequency, Sentence A: This 1, dress 1, a little 1, dirty 1, I want 1, change 1, one 1, not / clean 0, another 0; Sentence B: This 1, dress 1, a little 0, dirty 0, I want 1, change 0, one 1, not / clean 1, another 1. Step 4: Write out the word vector, Sentence A: (1, 1, 1, 1, 1, 1, 0, 0, 0), Sentence B: (1, 1, 0, 0, 1, 0, 1, 1, 1, 1).

[0103] S207: Perform similarity matching based on the at least one word vector and the cache database to obtain the target historical voice instruction.

[0104] The target historical voice command is used to indicate that the clothes processing device has recognized the command that can start the clothes processing device, and the similarity matching includes: cosine similarity and semantic similarity.

[0105] For example, word vector A is matched with the cache database for similarity, and finally B has a relatively high match. Then the target historical voice command is obtained based on the command recognizable by the clothing processing device corresponding to B.

[0106] Optionally, the at least one word vector and the cache database are processed by cosine similarity matching to obtain at least one first filtering result; the at least one first filtering result is input into a convolutional neural network to obtain at least one second filtering result, and the convolutional neural network is used to perform semantic similarity matching on the at least one first filtering result; and the second filtering result whose confidence meets a preset condition in the at least one second filtering result is used as the target historical voice instruction.

[0107] Among them, cosine similarity is the two corresponding vectors obtained after two texts are vectorized. By calculating the cosine value of the two vectors, the larger the cosine value, the higher the similarity of the two texts, and the smaller the cosine value, the lower the similarity of the two texts; the convolutional neural network determines the similarity of two sentences through text recognition. For example, the cosine similarity of the two sentences "I want to wash clothes" and "I don't wash clothes" is relatively high, while the convolutional neural network can be used to obtain the low semantic similarity of the two sentences; the confidence is to determine the interval value with a confidence greater than 95% for at least one of the cosine value and semantic similarity obtained after the text is vectorized, and find the corresponding text instructions in the interval as the target historical voice instructions.

[0108] Cosine similarity matching compares the text corresponding to the natural language with the data in the cache database by word segmentation to obtain the word vectors of the text and the cache database, and obtains the cosine value of the word vector of the text through cosine calculation. For example, the screening results with cosine values ​​greater than 0.5 can be retained, which is the first screening result; the text corresponding to the first screening result is then input into the convolutional neural network model for semantic similarity matching. For example, the screening results with semantic similarity greater than 0.5 can be retained, which is the second screening result. By determining the confidence interval of the second screening result, a second screening result that meets the confidence level is obtained, and the screening result is used as the target historical voice instruction.

[0109] S208: Obtain a target matching program, and perform matching processing on the text and the cache database according to the target matching program to obtain a matching result.

[0110] The target matching program is used to further parse the input instruction, and the matching result is the unique result obtained by matching the input instruction with the data in the cache database. The target matching program may include, for example: a one-to-one matching program, a rule template matching program, and a statistical method matching program. The matching result exists in the cache database, and the matching result can control the clothing processing program through the corresponding associated instruction.

[0111] Optionally, one-to-one matching, rule template matching and statistical method matching are simultaneously performed on the text and the cache database to obtain sub-matching results output by each matching program; and the matching result is determined based on the sub-matching results.

[0112] Among them, the one-to-one matching is to make a one-to-one correspondence between the input instruction and the data in the cache database between characters or vectors. When the input instruction corresponds one-to-one with a type of data in the cache database, it is considered that the natural language matching is successful.

[0113] Rule template matching is an association relationship established through semantic connection, so that the input instructions can obtain matching results through semantic association. For example, if the input instructions contain clear semantic connections between execution actions and clothing types such as "wash, start, cancel, pause" and "shirt", matching results will be obtained based on this association relationship in the cache database.

[0114] The statistical method is to determine the matching results based on the user's frequency of use. For example, if "weather" and "Qingdao" have a weak semantic correlation with the data in the cache database, you can force a match based on the user's historical frequency of use to obtain a matching result.

[0115] The above three target matching procedures can be performed simultaneously, or the first target matching procedure can be performed first, and after the first target matching procedure successfully matches, the other two target matching procedures can be performed. This application does not impose any limitation on this.

[0116] After obtaining three sub-matching results respectively through the above three target matching program methods, a corresponding matching result can be jointly determined based on the three sub-matching results.

[0117] For example, any one of the three sub-matching results may be used as the matching result, or the accuracy of the three sub-matching results may be calculated first, and then the sub-matching result with the highest accuracy may be used as the matching result.

[0118] The speech recognition method provided in this embodiment obtains historical speech interaction data in historical data and establishes a cache database, obtains natural language input by the user, parses the natural language to obtain text instructions, and determines whether the text length is greater than a preset threshold. If the text length is greater than the preset threshold, the text is segmented to obtain at least one word vector, and the word vector is matched with the cache database for similarity to obtain the target historical voice instruction; if the text length is not greater than the preset threshold, the target matching program is obtained, the text and the cache database are matched to obtain a matching result, and the target control instruction is determined; this method avoids the long time of user voice instruction interaction, the phenomenon of not being able to understand after speaking multiple times, and typos in speech-to-text conversion, thereby improving the efficiency of speech recognition and enhancing user experience.

[0119] Figure 3 This is a schematic diagram of the structure of the speech recognition device provided in this application. Figure 3 As shown, the present application provides a speech recognition device, the speech recognition device 300 includes:

[0120] An acquisition module 301 is used to acquire a natural language input by a user;

[0121] The processing module 302 is used to parse the natural language to obtain a text instruction, wherein the text instruction includes: text length and text;

[0122] The processing module 302 is further used to perform matching screening on the text and the cache database according to the text length and a preset threshold to obtain a matching result;

[0123] The determination module 303 is used to determine the target control instruction according to the matching result.

[0124] Optionally, the acquisition module 301 is further used to acquire historical voice interaction data in the historical data;

[0125] The processing module 302 is further configured to establish a cache database according to the historical voice interaction data, wherein the cache database includes: a correspondence between historical voice instructions and control instructions.

[0126] Optionally, the speech recognition device further includes: a judgment module 304;

[0127] The judging module 304 is used to judge whether the length of the text is greater than a preset threshold;

[0128] The processing module 302 is specifically used to perform similarity matching screening on the text and the cache database to obtain a matching result;

[0129] The processing module 302 is specifically used to perform strong matching strategy screening on the text and the cache database to obtain a matching result.

[0130] Optionally, the processing module 302 is further configured to perform word segmentation processing on the text to obtain at least one word vector;

[0131] The processing module 302 is specifically configured to perform similarity matching based on the at least one word vector and the cache database to obtain a target historical voice instruction.

[0132] Optionally, the processing module 302 is specifically configured to process the at least one word vector and the cache database by using cosine similarity matching to obtain at least one first screening result;

[0133] The processing module 302 is specifically used to input the at least one first screening result into a convolutional neural network to obtain at least one second screening result, and the convolutional neural network is used to perform semantic similarity matching on the at least one first screening result;

[0134] The processing module 302 is specifically configured to use the second screening result whose confidence level meets a preset condition in the at least one second screening result as the target historical voice instruction.

[0135] Optionally, the acquisition module 301 is further used to acquire a target matching program, and perform matching processing on the text and the cache database according to the target matching program to obtain a matching result.

[0136] Optionally, the processing module 302 is specifically used to perform one-to-one matching, rule template matching and statistical method matching on the text and the cache database at the same time to obtain sub-matching results output by each matching program;

[0137] The determination module 303 is further configured to determine the matching result according to the sub-matching result.

[0138] Figure 4 This is a schematic diagram of the structure of the speech recognition device provided in this application. Figure 4 As shown, the present application provides a speech recognition device, the speech recognition device 400 includes: a receiver 401, a transmitter 402, a processor 403 and a memory 404.

[0139] Receiver 401, for receiving instructions and data;

[0140] A transmitter 402, used for sending instructions and data;

[0141] Memory 404, for storing computer-executable instructions;

[0142] The processor 403 is used to execute the computer-executable instructions stored in the memory 404 to implement the various steps performed by the speech recognition method in the above embodiment. For details, please refer to the relevant description in the above speech recognition method embodiment.

[0143] Optionally, the memory 404 may be independent or integrated with the processor 403 .

[0144] When the memory 404 is independently provided, the electronic device further includes a bus for connecting the memory 404 and the processor 403 .

[0145] The present application also provides a computer-readable storage medium, in which computer-executable instructions are stored. When a processor executes the computer-executable instructions, the speech recognition method performed by the above-mentioned speech recognition device is implemented.

[0146] It will be appreciated by those skilled in the art that all or some of the steps, systems, and functional modules / units in the methods disclosed above may be implemented as software, firmware, hardware, and appropriate combinations thereof. In hardware implementations, the division between the functional modules / units mentioned in the above description does not necessarily correspond to the division of physical components; for example, a physical component may have multiple functions, or a function or step may be performed by several physical components in cooperation. Some or all physical components may be implemented as software executed by a processor, such as a central processing unit, a digital signal processor, or a microprocessor, or implemented as hardware, or implemented as an integrated circuit, such as an application-specific integrated circuit. Such software may be distributed on a computer-readable medium, which may include a computer storage medium (or non-transitory medium) and a communication medium (or temporary medium). As known to those skilled in the art, the term computer storage medium includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules, or other data). Computer storage media include, but are not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disks (DVD) or other optical disk storage, magnetic cassettes, magnetic tapes, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to store the desired information and can be accessed by a computer. In addition, it is well known to those of ordinary skill in the art that communication media typically contain computer-readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transport mechanism, and may include any information delivery media.

[0147] So far, the technical solution of the present application has been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present application is obviously not limited to these specific embodiments, and the above embodiments are only used to illustrate the technical solution of the present application rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein by equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A speech recognition method, characterized in that: The method comprises: Get the natural language input by the user; Parsing the natural language to obtain a text instruction, wherein the text instruction includes: text length and text; According to the text length and a preset threshold, the text and the cache database are matched and screened to obtain a matching result; According to the matching result, a target control instruction is determined.

2. The method according to claim 1, characterized in that Before obtaining the natural language input by the user, the method further includes: Obtain historical voice interaction data from historical data; A cache database is established according to the historical voice interaction data, and the cache database includes: a correspondence between historical voice instructions and control instructions.

3. The method according to claim 2, characterized in that The step of matching and screening the text and the cache database according to the text length and the preset threshold to obtain a matching result includes: Determine whether the length of the text is greater than a preset threshold; If so, performing similarity matching screening on the text and the cache database to obtain a matching result; If not, a strong matching strategy is performed on the text and the cache database to obtain a matching result.

4. The method according to claim 3, characterized in that The performing similarity matching screening on the text and the cache database to obtain a matching result includes: Performing word segmentation on the text to obtain at least one word vector; A similarity match is performed based on the at least one word vector and the cache database to obtain a target historical voice instruction.

5. The method according to claim 4, characterized in that The performing similarity matching based on the at least one word vector and the cache database to obtain the target historical voice instruction includes: Processing the at least one word vector and the cache database by cosine similarity matching to obtain at least one first screening result; Inputting the at least one first screening result into a convolutional neural network to obtain at least one second screening result, wherein the convolutional neural network is used to perform semantic similarity matching on the at least one first screening result; The second screening result whose confidence level meets a preset condition among the at least one second screening result is used as the target historical voice instruction.

6. The method according to claim 3, characterized in that The text and the cache database are screened by a strong matching strategy to obtain a matching result, including A target matching program is obtained, and matching processing is performed on the text and the cache database according to the target matching program to obtain a matching result.

7. The method according to claim 6, characterized in that The target matching program includes: a one-to-one matching program, a rule template matching program, and a statistical method matching program. According to the target matching program, the text and the cache database are matched to obtain a matching result, including: Perform one-to-one matching, rule template matching and statistical method matching on the text and the cache database at the same time to obtain sub-matching results output by each matching program; The matching result is determined according to the sub-matching result.

8. A speech recognition device, characterized in that: include: An acquisition module is used to acquire the natural language input by the user; A processing module, used for parsing the natural language to obtain a text instruction, wherein the text instruction includes: text length and text; The processing module is further used to perform matching screening on the text and the cache database according to the text length and a preset threshold to obtain a matching result; A determination module is used to determine a target control instruction according to the matching result.

9. A speech recognition device, characterized in that: include: Memory; processor; Wherein, the memory stores computer-executable instructions; The processor executes the computer-executable instructions stored in the memory to implement the speech recognition method according to any one of claims 1 to 7.

10. A computer storage medium, characterized in that: The computer storage medium stores computer execution instructions, which, when executed by a processor, are used to implement the speech recognition method according to any one of claims 1 to 7.

Citation Information

Cited By

  • Business financial reconciliation method and system

    CN120147042A

  • Business and financial reconciliation method and system

    CN120147042B

  • Image searching method and device, electronic equipment and storage medium

    CN120950721A