A method, device and storage medium for acquiring text of voice information

Abstract

The disclosed embodiment discloses a method, device, electronic device and storage medium for acquiring text of speech information, the method comprising: acquiring speech information to be recognized, and acquiring initial recognition text through speech recognition technology; inputting the initial recognition text into a trained inverse text standardization model to perform inverse text standardization processing on the initial recognition text, and using the processing result as the standard recognition text of the speech information. The technical solution of the disclosed embodiment performs inverse text standardization processing on the initial text after speech recognition through a trained inverse text standardization model, and then acquires the standard recognition text, realizing the full-text inverse text mapping from the input end to the output end, and the inverse text method combined with the specific context also improves the accuracy of the inverse text processing, and at the same time, avoids the tedious process of updating and maintenance, saving manpower and time costs.

Description

Technical Field

[0001] The embodiments of the present disclosure relate to speech recognition technology, and more particularly to a method, device, electronic device and storage medium for acquiring text from speech information. Background Art

[0002] With the continuous advancement of science and technology, speech recognition technology has developed rapidly and has been widely used in many fields such as intelligent control, providing great convenience for people's social life.

[0003] Through speech recognition technology, after obtaining the recognized text, in order to facilitate user reading, inverse text normalization (ITN) processing will be performed, that is, the Chinese character information is converted into a form that is easier for users to interpret. For example, "three hundred and twenty-seven" is converted into "327". Usually, after obtaining the speech recognition text, the Chinese character information is converted into corresponding digital information through a mapping rule comparison table.

[0004] However, this inverse text processing method has poor maintainability. Each addition, deletion, modification, and other operations on the mapping rule comparison table require a list update, and inverse text mapping can only be performed on the information existing in the mapping rule comparison table. The inverse text function has great limitations. Summary of the invention

[0005] The present disclosure provides a method, device, equipment and storage medium for acquiring text of speech information, so as to implement inverse text normalization processing on text information after speech recognition.

[0006] In a first aspect, an embodiment of the present disclosure provides a method for acquiring text of voice information, comprising:

[0007] Acquire the voice information to be recognized, and obtain the initial recognition text through voice recognition technology;

[0008] The initial recognition text is input into a trained inverse text standardization model to perform inverse text standardization processing on the initial recognition text, and the processing result is used as the standard recognition text of the voice information.

[0009] In a second aspect, an embodiment of the present disclosure provides a device for acquiring text of voice information, including:

[0010] An initial recognition text acquisition module is used to acquire the voice information to be recognized and obtain the initial recognition text through voice recognition technology;

[0011] The inverse text standardization execution module is used to input the initial recognition text into the trained inverse text standardization model to perform inverse text standardization processing on the initial recognition text, and use the processing result as the standard recognition text of the voice information.

[0012] In a third aspect, an embodiment of the present disclosure provides an electronic device, including a memory, a processing device, and a computer program stored in the memory and executable on the processing device, wherein the processing device implements the text acquisition method of voice information of any embodiment of the present disclosure when executing the program.

[0013] In a fourth aspect, an embodiment of the present disclosure provides a storage medium containing computer executable instructions, which, when executed by a computer processor, are used to execute the text acquisition method of voice information of any embodiment of the present disclosure.

[0014] The technical solution of the disclosed embodiment performs inverse text normalization processing on the initial text after speech recognition through a trained inverse text normalization model, and then obtains the standard recognition text, realizing the full-text inverse text mapping from the input end to the output end, and the inverse text method combined with the specific context also improves the accuracy of the inverse text processing. At the same time, the learning ability of the inverse text normalization model improves the scope of application of inverse text, avoids the tedious process of updating and maintenance, and saves manpower and time costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] The above and other features, advantages and aspects of the embodiments of the present disclosure will become more apparent with reference to the following detailed description in conjunction with the accompanying drawings. Throughout the accompanying drawings, the same or similar reference numerals represent the same or similar elements. It should be understood that the drawings are schematic and the originals and elements are not necessarily drawn to scale.

[0016] Figure 1 is a flow chart of an embodiment of a method for obtaining text of voice information disclosed in the present invention;

[0017] Figure 2 is a flow chart of another embodiment of a method for obtaining text of voice information disclosed in the present invention;

[0018] Figure 3 is a flow chart of another embodiment of a method for obtaining text of voice information disclosed in the present invention;

[0019] Figure 4 It is a structural block diagram of an embodiment of a device for acquiring text of voice information disclosed in the present invention;

[0020] Figure 5 It is a structural block diagram of an electronic device suitable for implementing the embodiments of the present disclosure. DETAILED DESCRIPTION

[0021] Embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as being limited to the embodiments described herein, which are instead provided for a more thorough and complete understanding of the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are only for exemplary purposes and are not intended to limit the scope of protection of the present disclosure.

[0022] It should be understood that the various steps described in the method embodiments of the present disclosure may be performed in different orders and / or in parallel. In addition, the method embodiments may include additional steps and / or omit the steps shown. The scope of the present disclosure is not limited in this respect.

[0023] The term "including" and its variations used herein are open inclusions, i.e., "including but not limited to". The term "based on" means "based at least in part on". The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments". The relevant definitions of other terms will be given in the following description.

[0024] It should be noted that the concepts such as "first" and "second" mentioned in the present disclosure are only used to distinguish different devices, modules or units, and are not used to limit the order or interdependence of the functions performed by these devices, modules or units.

[0025] It should be noted that the modifications of "one" and "plurality" mentioned in the present disclosure are illustrative rather than restrictive, and those skilled in the art should understand that unless otherwise clearly indicated in the context, it should be understood as "one or more".

[0026] The names of the messages or information exchanged between multiple devices in the embodiments of the present disclosure are only used for illustrative purposes and are not used to limit the scope of these messages or information.

[0027] Embodiment 1

[0028] Figure 1 This is a flowchart of a method for acquiring text from speech information provided in the first embodiment of the present disclosure. This embodiment is applicable to the case where inverse text normalization is performed on text information after speech recognition. The method can be executed by a device for acquiring text from speech information in the embodiment of the present disclosure. The device can be implemented by software and / or hardware and integrated in an electronic device. The method specifically includes the following steps:

[0029] S110: Acquire voice information to be recognized, and obtain initial recognition text through voice recognition technology.

[0030] Speech recognition is the process of converting speech signals into text through machine recognition and understanding; the recognition process of speech information can be achieved through Automatic Speech Recognition (ASR) technology and / or Natural Language Understanding (NLU) technology; ASR is the conversion of vocabulary content in human speech into computer-readable input, such as keystrokes, binary codes or character sequences; NLU focuses on the semantics of text and accurately expresses the semantic information conveyed by the text. Specifically, based on a large amount of speech corpus information, after speech signal processing and knowledge mining, a corresponding speech recognition model is established. When the speech information to be recognized is obtained, the speech recognition model that has been trained obtains the recognized text of the speech information through speech feature extraction and feature matching, that is, the initial recognized text. In the embodiment of the present invention, optionally, the technical method adopted for speech recognition and the type of speech recognition model are not specifically limited.

[0031] S210, inputting the initial recognition text into a trained inverse text standardization model to perform inverse text standardization processing on the initial recognition text, and using the processing result as the standard recognition text of the voice information.

[0032] The inverse text standardization model is used to perform inverse text standardization processing on text information to obtain standard recognition text that conforms to reading habits; the inverse text standardization model performs inverse text standardization processing by taking each sentence in the initial text as the object of inverse text processing, that is, the target vocabulary to be inverse textized and the semantic information of the context in which it is located are all taken as processing objects. Compared with the inverse text processing method that only takes a certain word or phrase as the replacement object, the above technical solution comprehensively considers the specific language environment of each sentence, and the trained inverse text standardization model has the ability to map the full text from the input end to the output end.

[0033] Optionally, in the embodiments of the present disclosure, before inputting the initial recognition text into the trained inverse text normalization model, it further includes: constructing a first sequence-to-sequence model, and performing inverse text normalization training on the first sequence-to-sequence model by using a set of inverse text corpus pairs to obtain a trained inverse text normalization model; wherein, the set of inverse text corpus pairs includes at least one inverse text corpus pair. The sequence-to-sequence (seq2seq) model is a neural network with an Encoder-Decoder structure. The input is a sequence, and the output is also a sequence. In the Encoder, a variable-length sequence is transformed into a fixed-length vector representation, and the Decoder converts this fixed-length vector into a variable-length target signal sequence, thereby realizing the conversion from an indefinite-length input to an indefinite-length output. For example, when Chinese characters are translated into Arabic numerals, the length of the Arabic numerals (output) may be shorter than that of the Chinese characters (input), or may be longer than that of the Chinese characters, and the length of the output is uncertain. The sequence-to-sequence model can include various types. For example, a seq2seq model based on a Recurrent Neural Network (RNN) and a seq2seq model based on a Convolution (CONV) neural network, etc. Optionally, in the embodiments of the present disclosure, the type of neural network used for the sequence-to-sequence model is not specifically limited.

[0034] Optionally, in the embodiments of the present disclosure, the inverse text corpus pair includes a natural number corpus pair, a number information corpus pair, a symbol information corpus pair, a time scenario corpus pair, and / or a conventional collocation corpus pair. Natural numbers include zero and positive integers, and the Chinese character descriptions corresponding to natural numbers include corresponding "digit" descriptions (such as "ten", "hundred", "thousand", etc.). For example, in the natural number corpus pair, the inverse text corresponding to "three hundred and twenty-seven" is mapped to "327". Specifically, the natural number corpus pair includes a ten-thousand-digit divisible corpus pair and a non-ten-thousand-digit divisible corpus pair. A ten-thousand-digit divisible natural number is a natural number that can be divided evenly by the number "10,000". For a natural number with a large number of digits that can be divided evenly by "10,000", it ends with "ten thousand" or "hundred million" to avoid too many digits and inconvenient reading. For example, in the ten-thousand-digit divisible corpus pair, the inverse text corresponding to "one million three hundred thousand" is mapped to "1.3 million"; the inverse text corresponding to "twenty-nine billion" is mapped to "29 billion". A non-ten-thousand-digit divisible natural number is a natural number that cannot be divided evenly by the number "10,000", and it is expanded in the form of Arabic numerals in sequence, and every three digits are separated by a comma for easy reading by users. For example, in the non-ten-thousand-digit divisible corpus pair, the inverse text corresponding to "seven hundred and eighty-one" is mapped to "781", and the inverse text corresponding to "one million one hundred and thirteen thousand six hundred and seventy-one" is mapped to "1,133,671".

[0035] The Chinese character description corresponding to the number information does not include the corresponding "digit" description. Therefore, for the number information corpus pair, the Chinese character description can be directly translated. For example, the inverse text mapping corresponding to "two zero one eight" in the number information corpus pair is "2018"; optionally, in an embodiment of the present disclosure, the number information corpus pair includes an ID number corpus pair, a telephone number corpus pair, a license plate number corpus pair, an account number corpus pair, a password corpus pair, a date corpus pair, a number corpus pair and / or an address information corpus pair; in particular, for number information, there are usually specific digits, for example, an ID card number is 18 digits and a mobile phone number is 11 digits. Therefore, when performing inverse text normalization training on the inverse text normalization model, intensive training can be performed on the inverse text corpus pairs containing numbers with specific digits.

[0036] The symbol information corpus pair includes not only numbers but also some special characters. The numbers are associated with the semantic environment of the context in which they are located. Therefore, in addition to the need to reverse textualize the numbers, the associated Chinese characters also need to be reverse textualized into corresponding symbols. The symbol information corpus pair may specifically include currency symbol corpus pairs (for example, the reverse text corresponding to "RMB 50 yuan" is "￥50 yuan"), mathematical symbol corpus pairs (for example, the reverse text corresponding to "five plus seven equals twelve" is "5+7=12") and / or unit symbol corpus pairs (for example, the reverse text corresponding to "degrees Celsius" is "℃"). In particular, for uncommon special symbols, if symbols are used instead of characters, it is not conducive to user understanding. Therefore, when constructing the symbol information corpus pair, only specific symbols are used. For example, the currency symbol corpus pair only includes the mapping relationships between "RMB" and "US dollar" and the currency symbols "￥" and "$" respectively, and does not include other types of currency symbol corpus pairs.

[0037] The moment-scene corpus pair describes the content information related to the "time point", which is different from the digital inverse text mapping relationship and the symbolic inverse text mapping relationship. In the moment-scene corpus pair, the inverse text mapping format corresponding to the part of the Chinese characters representing numbers is in the form of ":" separating "hour" and "minute". For example, the inverse text mapping of "get up at 8:30 in the morning" is "get up at 8:30 in the morning", and the inverse text mapping of "have a meeting at 2 o'clock in the afternoon" is "have a meeting at 2:00 in the afternoon".

[0038] Conventional collocation corpus pairs are some fixed expression habits that describe situations where Chinese characters contain numbers but are not suitable to be expressed in Arabic numerals. In such cases, reverse text normalization is not required, and the original Chinese character expression can be retained. Conventional collocation corpus pairs can include common language corpus pairs, event corpus pairs, festival corpus pairs, and / or name corpus pairs; among them, common language corpus pairs include idiom corpus pairs (for example, the reverse text mapping corresponding to "five colors and ten hues" is still "five colors and ten hues"), poem corpus pairs, ancient text corpus pairs, overview corpus pairs (for example, the reverse text mapping of "the first" is "the first"), and stereotyped words, etc.; event corpus pairs, such as the "May 4th Movement"; festival corpus pairs, such as the "National Day long holiday on October 1st"; name corpus pairs include personal names, place names, organization names, and thing names, etc.

[0039] Specifically, both the source corpus and the target corpus in the reverse text corpus pair are sentences; the source corpus is the original corpus before reverse text normalization, and the target corpus is the regularized corpus after reverse text normalization. In order to enable the trained reverse text normalization model to have the full-text reverse text mapping ability from the input end to the output end, when constructing the reverse text corpus pair, both the source corpus and the target corpus are in the form of sentences, rather than a single word form. For example, if the source corpus is "I didn't get up until 7:45 this morning", the corresponding target corpus is "I didn't get up until 7:45 this morning", rather than just mapping "seven forty-five" to "7:40".

[0040] The technical solution of the embodiment of the present disclosure performs reverse text normalization processing on the initial text of speech recognition through the trained reverse text normalization model, and then obtains the standard recognition text, realizing the full-text reverse text mapping ability from the input end to the output end. The reverse text normalization method combined with the specific context also improves the accuracy of reverse text normalization processing. At the same time, the learning ability of the reverse text normalization model improves the applicable range of reverse text normalization and also avoids the cumbersome process of updating and maintenance, saving labor costs and time costs.

[0041] Embodiment Two

[0042] Figure 2 It is a flowchart of a method for obtaining text of voice information provided by the second embodiment of the present disclosure. This embodiment is specific based on the above embodiment. In this embodiment, according to the obtained standard text information and the first positive text normalization model, a set of reverse text corpus pairs is constructed. The method of this embodiment specifically includes the following steps:

[0043] S210. According to the obtained reverse text mapping table, construct a set of positive text corpus pairs and obtain standard text information.

[0044] The inverse text mapping table can be an existing list containing the mapping relationship between Chinese characters and digital symbols, or it can be an exemplary table constructed according to the construction rules of the above technical solution, and the inverse text mapping table contains a limited number of mapping relationship items. By exchanging the identity of the source corpus and the target corpus in each mapping relationship item in the inverse text mapping table, a set of positive text corpus pairs can be obtained, wherein the source corpus in each positive text corpus pair is information expressed in digital symbols, and the target corpus is information expressed in Chinese characters. Standard text information is formal text information obtained through electronic newspapers, electronic books, web pages, etc. that conforms to the user's reading habits, and serves as the source corpus information in the constructed inverse text corpus pair set.

[0045] S220: construct a second sequence-to-sequence model, and perform positive text standardization training on the second sequence-to-sequence model using the positive text corpus pair set to obtain a trained first positive text standardization model.

[0046] The second sequence-to-sequence model is trained on normalized text, that is, the information expressed by digital symbols is taken as input and the information expressed by Chinese characters is taken as output, so that it has the ability to normalize text.

[0047] S230: Input the standard text information into the first positive text standardization model to obtain corresponding first standard text mapping information.

[0048] Through the training of the above technical solution, the first main text standardization model has the ability to map digital symbol expression information into Chinese character expression information; therefore, the obtained standard text information is used as input information and input into the first main text standardization model, and the matching Chinese character expression information as output information can be obtained, that is, the first standard text mapping information.

[0049] S240: Construct the inverse text corpus pair set according to the standard text information and the first standard text mapping information.

[0050] The sources of standard text information are numerous and easy to obtain. Therefore, the inverse text corpus pair set constructed by the above technical solution includes a large number of training samples, which is much larger than the number of inverse text corpus pairs directly obtained through the inverse text mapping table. Therefore, the above technical solution greatly expands the number of training samples in the inverse text corpus pair set, providing sample guarantee for the training of the inverse text standardization model.

[0051] The technical solution of the disclosed embodiment obtains the training of the first positive text standardization model through the inverse text mapping table, and constructs an inverse text corpus pair set containing a large number of training samples based on the obtained standard text information, which provides sample support for the training of the inverse text standardization model and improves the processing capability of the trained inverse text standardization model.

[0052] Embodiment 3

[0053] Figure 3 : is a flowchart of a method for obtaining text of speech information provided by Embodiment 3 of the present disclosure. This embodiment is specific based on the above embodiment. In this embodiment, a set of inverse text corpus pairs is obtained according to the obtained standard text information and multiple second positive text standardization models. The method of this embodiment specifically includes the following steps:

[0054] S310: construct a positive text corpus pair set according to the obtained inverse text mapping table, and obtain standard text information.

[0055] S320, constructing a plurality of third sequence-to-sequence models, and performing positive text standardization training on each of the third sequence-to-sequence models through the positive text corpus pair set to obtain a plurality of trained second positive text standardization models; wherein each of the third sequence-to-sequence models is constructed based on a different neural network.

[0056] Each third sequence-to-sequence model is constructed based on a different neural network. For example, two third sequence-to-sequence models are constructed based on a recurrent neural network and a convolutional operation neural network respectively. Even if they are trained with the same training sample set, the two second positive text normalization models have different positive text processing capabilities.

[0057] S330: Input the standard text information into each of the second normal text standardization models respectively to obtain corresponding multiple second standard text mapping information.

[0058] Since the processing capabilities of each second normal text standardization model are different, different mapping information may be obtained for the same standard text information, or the same mapping information may be obtained.

[0059] S340: Select target second standard text mapping information with the same text content from each of the second standard text mapping information, and determine matching target standard text information according to the target second standard text mapping information.

[0060] Taking the above technical solution as an example, two third sequence-to-sequence models are constructed based on the recurrent neural network and the convolutional operation neural network respectively. The standard text information A is "the total amount of purchased materials is one hundred and three thousand four hundred and fifteen". Through the above two second text standardization models, the second standard text mapping information A1 and the second standard text mapping information A2 are obtained as "the total amount of purchased materials is 103415" and "the total amount of purchased materials is 103,415" respectively. Obviously, the output text content is not the same. Therefore, the standard text information A, the second standard text mapping information A1 and the second standard text mapping information A2 cannot be used as valid information; the standard text information B is "the meeting ends at 5:30 pm". Through the above two second text standardization models, the second standard text mapping information B1 and the second standard text mapping information B2 are obtained as "the meeting ends at 5:30 pm" and "the meeting ends at 5:30 pm" respectively. Obviously, the output text content is the same. Therefore, the second standard text mapping information B1 or the second standard text mapping information B2 is used as the target second standard text mapping information, and the corresponding target standard text information is further determined to be the standard text information B.

[0061] S350: Construct the inverse text corpus pair set according to the target standard text information and the target second standard text mapping information.

[0062] The technical solution of the disclosed embodiment obtains the training of multiple second positive text standardization models through the inverse text mapping table, and after obtaining multiple second standard text mapping information, further constructs an inverse text corpus pair set based on the target second standard text mapping information with the same text content, thereby ensuring the accuracy of the training samples in the inverse text corpus pair set, avoiding the error in the obtained output information due to the training error of the second positive text standardization model, and thus failing to obtain an accurate inverse text corpus pair set, affecting the training effect of the inverse text standardization model.

[0063] Embodiment 4

[0064] Figure 4 It is a structural block diagram of a text acquisition device for voice information provided by Embodiment 4 of the present disclosure, which specifically includes: an initial recognition text acquisition module 410 and an inverse text normalization execution module 420.

[0065] The initial recognition text acquisition module 410 is used to acquire the voice information to be recognized and obtain the initial recognition text through the voice recognition technology;

[0066] The inverse text standardization execution module 420 is used to input the initial recognition text into the trained inverse text standardization model to perform inverse text standardization processing on the initial recognition text, and use the processing result as the standard recognition text of the voice information.

[0067] The technical solution of the disclosed embodiment performs inverse text normalization processing on the initial text of speech recognition through a trained inverse text normalization model, and then obtains the standard recognition text, thereby realizing the full-text inverse text mapping capability from the input end to the output end, and the inverse text method combined with the specific context also improves the accuracy of the inverse text processing. At the same time, the learning ability of the inverse text normalization model improves the scope of application of inverse text, avoids the tedious process of updating and maintenance, and saves manpower and time costs.

[0068] Optionally, based on the above technical solution, the text acquisition device of voice information further includes:

[0069] An inverse text standardization model acquisition module is used to construct a first sequence-to-sequence model, and perform inverse text standardization training on the first sequence-to-sequence model through an inverse text corpus pair set to obtain a trained inverse text standardization model; wherein the inverse text corpus pair set includes at least one inverse text corpus pair.

[0070] Optionally, based on the above technical solution, the inverse text corpus pair includes a natural number corpus pair, a number information corpus pair, a symbol information corpus pair, a moment scene corpus pair and / or an idiomatic collocation corpus pair.

[0071] Optionally, based on the above technical solution, the natural number corpus pairs include corpus pairs that are divisible by ten thousand digits and corpus pairs that are not divisible by ten thousand digits.

[0072] Optionally, based on the above technical solution, the source corpus and the target corpus in the inverse text corpus pair are both sentences.

[0073] Optionally, based on the above technical solution, the text acquisition device of voice information further includes:

[0074] A standard text information acquisition module is used to construct a positive text corpus pair set according to the acquired inverse text mapping table and obtain standard text information;

[0075] A first normalized text model acquisition module is used to construct a second sequence-to-sequence model, and perform normalized text training on the second sequence-to-sequence model using the normal text corpus pair set to obtain a trained first normalized text model;

[0076] A first standard text mapping information acquisition module, used for inputting the standard text information into the first positive text standardization model to acquire corresponding first standard text mapping information;

[0077] The first inverse text corpus pair set is used to construct the inverse text corpus pair set according to the standard text information and the first standard text mapping information.

[0078] Optionally, based on the above technical solution, the text acquisition device of voice information further includes:

[0079] A second main text standardization model acquisition module is used to construct a plurality of third sequence-to-sequence models, and perform main text standardization training on each of the third sequence-to-sequence models through the main text corpus pair set to obtain a plurality of trained second main text standardization models; wherein each of the third sequence-to-sequence models is constructed based on a different neural network;

[0080] A second standard text mapping information acquisition module, used for inputting the standard text information into each of the second normal text standardization models respectively to obtain corresponding multiple second standard text mapping information;

[0081] A target standard text information acquisition module, used to select target second standard text mapping information with the same text content from each of the second standard text mapping information, and determine matching target standard text information according to the target second standard text mapping information;

[0082] The second inverse text corpus pair set is used to construct the inverse text corpus pair set according to the target standard text information and the target second standard text mapping information.

[0083] The above device can execute the text acquisition method of voice information provided by any embodiment of the present disclosure, and has the corresponding functional modules and beneficial effects of the execution method. For technical details not described in detail in this embodiment, please refer to the method provided by any embodiment of the present disclosure.

[0084] Embodiment 5

[0085] Figure 5 The schematic diagram of the structure of an electronic device 500 suitable for implementing the embodiment of the present disclosure is shown. The terminal device in the embodiment of the present disclosure may include but is not limited to mobile terminals such as mobile phones, notebook computers, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), vehicle terminals (such as vehicle navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 5 The electronic device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0086] like Figure 5 As shown, the electronic device 500 may include a processing device (e.g., a central processing unit, a graphics processing unit, etc.) 501, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 502 or a program loaded from a storage device 508 into a random access memory (RAM) 503. In the RAM 503, various programs and data required for the operation of the electronic device 500 are also stored. The processing device 501, the ROM 502, and the RAM 503 are connected to each other via a bus 504. An input / output (I / O) interface 505 is also connected to the bus 504.

[0087] Typically, the following devices may be connected to the I / O interface 505: an input device 506 including, for example, a touch screen, a touch pad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, etc.; an output device 507 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; a storage device 508 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 509. The communication device 509 may allow the electronic device 500 to communicate with other devices wirelessly or by wire to exchange data. Although Figure 5 The electronic device 500 is shown with various devices, but it should be understood that it is not required to implement or possess all the devices shown. More or fewer devices may be implemented or possessed instead.

[0088] In particular, according to an embodiment of the present disclosure, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present disclosure includes a computer program product, which includes a computer program carried on a non-transitory computer-readable medium, and the computer program contains program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through a communication device 509, or installed from a storage device 508, or installed from a ROM 502. When the computer program is executed by the processing device 501, the above-mentioned functions defined in the method of the embodiment of the present disclosure are executed.

[0089] It should be noted that the computer-readable medium disclosed above may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that may be used by or in combination with an instruction execution system, device or device. In the present disclosure, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, in which a computer-readable program code is carried. This propagated data signal may take a variety of forms, including but not limited to an electromagnetic signal, an optical signal, or any suitable combination of the above. The computer readable signal medium may also be any computer readable medium other than a computer readable storage medium, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer readable medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (radio frequency), etc., or any suitable combination of the above.

[0090] In some embodiments, the client and the server may communicate using any currently known or future developed network protocol such as HTTP (HyperText Transfer Protocol), and may be interconnected with any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network ("LAN"), a wide area network ("WAN"), an internet (e.g., the Internet), and a peer-to-peer network (e.g., an ad hoc peer-to-peer network), as well as any currently known or future developed network.

[0091] The computer-readable medium may be included in the electronic device, or may exist independently without being installed in the electronic device.

[0092] The above-mentioned computer-readable medium carries one or more programs. When the above-mentioned one or more programs are executed by the electronic device, the electronic device: obtains the voice information to be recognized, and obtains the initial recognition text through voice recognition technology; inputs the initial recognition text into the trained inverse text standardization model to perform inverse text standardization processing on the initial recognition text, and uses the processing result as the standard recognition text of the voice information.

[0093] Computer program code for performing the operations of the present disclosure may be written in one or more programming languages ​​or a combination thereof, including, but not limited to, object-oriented programming languages, such as Java, Smalltalk, C++, and conventional procedural programming languages, such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on the remote computer. In cases involving a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0094] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present disclosure. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some implementations as replacements, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0095] The modules involved in the embodiments described in the present disclosure can be implemented by software or by hardware. The name of the module does not constitute a limitation on the module itself in some cases. For example, the inverse text standardization execution module can be described as "for inputting the initial recognition text into the trained inverse text standardization model to perform inverse text standardization on the initial recognition text, and using the processing result as the standard recognition text of the voice information". The functions described above in this article can be at least partially performed by one or more hardware logic components. For example, without limitation, exemplary types of hardware logic components that can be used include: field programmable gate array (FPGA), application specific integrated circuit (ASIC), application specific standard product (ASSP), system on chip (SOC), complex programmable logic device (CPLD), etc.

[0096] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0097] According to one or more embodiments of the present disclosure, [Example 1] provides a method for obtaining text of voice information, including:

[0098] Acquire the voice information to be recognized, and obtain the initial recognition text through voice recognition technology;

[0099] The initial recognition text is input into a trained inverse text standardization model to perform inverse text standardization processing on the initial recognition text, and the processing result is used as the standard recognition text of the voice information.

[0100] According to one or more embodiments of the present disclosure, [Example 2] provides the method of Example 1, further comprising:

[0101] A first sequence-to-sequence model is constructed, and the first sequence-to-sequence model is trained for inverse text standardization through an inverse text corpus pair set to obtain a trained inverse text standardization model; wherein the inverse text corpus pair set includes at least one inverse text corpus pair.

[0102] According to one or more embodiments of the present disclosure, [Example 3] provides the method of Example 2, further comprising:

[0103] The reverse text corpus pairs include natural number corpus pairs, number information corpus pairs, symbol information corpus pairs, moment scene corpus pairs and / or idiomatic collocation corpus pairs.

[0104] According to one or more embodiments of the present disclosure, [Example 4] provides the method of Example 3, further comprising:

[0105] The natural number corpus pairs include corpus pairs that are divisible by ten thousand digits and corpus pairs that are not divisible by ten thousand digits.

[0106] According to one or more embodiments of the present disclosure, [Example 5] provides the method of Example 2, further comprising:

[0107] The source corpus and the target corpus in the inverse text corpus pair are both sentences.

[0108] According to one or more embodiments of the present disclosure, [Example 6] provides the method of Example 2, further comprising:

[0109] According to the obtained inverse text mapping table, a positive text corpus pair set is constructed and standard text information is obtained;

[0110] Constructing a second sequence-to-sequence model, and performing text standardization training on the second sequence-to-sequence model using the text corpus pair set to obtain a trained first text standardization model;

[0111] Inputting the standard text information into the first normal text standardization model to obtain corresponding first standard text mapping information;

[0112] The inverse text corpus pair set is constructed according to the standard text information and the first standard text mapping information.

[0113] According to one or more embodiments of the present disclosure, [Example 7] provides the method of Example 6, further comprising:

[0114] Constructing a plurality of third sequence-to-sequence models, and performing text standardization training on each of the third sequence-to-sequence models through the text corpus pair set to obtain a plurality of trained second text standardization models; wherein each of the third sequence-to-sequence models is constructed based on a different neural network;

[0115] Inputting the standard text information into each of the second normal text standardization models respectively to obtain corresponding multiple second standard text mapping information;

[0116] Selecting target second standard text mapping information with the same text content from each of the second standard text mapping information, and determining matching target standard text information according to the target second standard text mapping information;

[0117] The inverse text corpus pair set is constructed according to the target standard text information and the target second standard text mapping information.

[0118] According to one or more embodiments of the present disclosure, [Example 8] provides a text acquisition device for voice information, including:

[0119] An initial recognition text acquisition module is used to acquire the voice information to be recognized and obtain the initial recognition text through voice recognition technology;

[0120] The inverse text standardization execution module is used to input the initial recognition text into the trained inverse text standardization model to perform inverse text standardization processing on the initial recognition text, and use the processing result as the standard recognition text of the voice information.

[0121] According to one or more embodiments of the present disclosure, [Example 9] provides the apparatus of Example 8, further comprising:

[0122] An inverse text standardization model acquisition module is used to construct a first sequence-to-sequence model, and perform inverse text standardization training on the first sequence-to-sequence model through an inverse text corpus pair set to obtain a trained inverse text standardization model; wherein the inverse text corpus pair set includes at least one inverse text corpus pair.

[0123] According to one or more embodiments of the present disclosure, [Example 10] provides the apparatus of Example 9, further comprising:

[0124] The reverse text corpus pairs include natural number corpus pairs, number information corpus pairs, symbol information corpus pairs, moment scene corpus pairs and / or idiomatic collocation corpus pairs.

[0125] According to one or more embodiments of the present disclosure, [Example 11] provides the apparatus of Example 10, further comprising:

[0126] The natural number corpus pairs include corpus pairs that are divisible by ten thousand digits and corpus pairs that are not divisible by ten thousand digits.

[0127] According to one or more embodiments of the present disclosure, [Example 12] provides the apparatus of Example 9, further comprising:

[0128] The source corpus and the target corpus in the inverse text corpus pair are both sentences.

[0129] According to one or more embodiments of the present disclosure, [Example 13] provides the apparatus of Example 9, further comprising:

[0130] A standard text information acquisition module is used to construct a positive text corpus pair set according to the acquired inverse text mapping table and obtain standard text information;

[0131] A first normalized text model acquisition module is used to construct a second sequence-to-sequence model, and perform normalized text training on the second sequence-to-sequence model using the normal text corpus pair set to obtain a trained first normalized text model;

[0132] A first standard text mapping information acquisition module, used for inputting the standard text information into the first positive text standardization model to acquire corresponding first standard text mapping information;

[0133] The first inverse text corpus pair set is used to construct the inverse text corpus pair set according to the standard text information and the first standard text mapping information.

[0134] According to one or more embodiments of the present disclosure, [Example 14] provides the apparatus of Example 13, further comprising:

[0135] A second main text standardization model acquisition module is used to construct a plurality of third sequence-to-sequence models, and perform main text standardization training on each of the third sequence-to-sequence models through the main text corpus pair set to obtain a plurality of trained second main text standardization models; wherein each of the third sequence-to-sequence models is constructed based on a different neural network;

[0136] A second standard text mapping information acquisition module, used for inputting the standard text information into each of the second normal text standardization models respectively to obtain corresponding multiple second standard text mapping information;

[0137] A target standard text information acquisition module, used to select target second standard text mapping information with the same text content from each of the second standard text mapping information, and determine matching target standard text information according to the target second standard text mapping information;

[0138] The second inverse text corpus pair set is used to construct the inverse text corpus pair set according to the target standard text information and the target second standard text mapping information.

[0139] According to one or more embodiments of the present disclosure, [Example 15] provides an electronic device, including a memory, a processing device, and a computer program stored in the memory and executable on the processing device, wherein when the processing device executes the program, the method for acquiring text of voice information as described in any one of Examples 1-7 is implemented.

[0140] According to one or more embodiments of the present disclosure, [Example 16] provides a storage medium containing computer executable instructions, which, when executed by a computer processor, are used to execute the text acquisition method of voice information as described in any one of Examples 1-7.

[0141] The above description is only a preferred embodiment of the present disclosure and an explanation of the technical principles used. Those skilled in the art should understand that the scope of disclosure involved in the present disclosure is not limited to the technical solutions formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above disclosed concept. For example, the above features are replaced with the technical features with similar functions disclosed in the present disclosure (but not limited to) by each other to form a technical solution.

[0142] In addition, although each operation is described in a specific order, this should not be understood as requiring these operations to be performed in the specific order shown or in a sequential order. Under certain circumstances, multitasking and parallel processing may be advantageous. Similarly, although some specific implementation details are included in the above discussion, these should not be interpreted as limiting the scope of the present disclosure. Some features described in the context of a separate embodiment can also be implemented in a single embodiment in combination. On the contrary, the various features described in the context of a single embodiment can also be implemented in multiple embodiments individually or in any suitable sub-combination mode.

[0143] Although the subject matter has been described in language specific to structural features and / or methodological logical actions, it should be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or actions described above. On the contrary, the specific features and actions described above are merely example forms of implementing the claims.

Claims

1. A method for obtaining text of voice information, characterized in that: include: Acquire the voice information to be recognized, and obtain the initial recognition text through voice recognition technology; Inputting the initial recognition text into the trained inverse text standardization model to perform inverse text standardization processing on the initial recognition text, and using the processing result as the standard recognition text of the voice information; Before inputting the initial recognition text into the trained inverse text normalization model, the method further includes: Constructing a first sequence-to-sequence model, and performing inverse text normalization training on the first sequence-to-sequence model through an inverse text corpus pair set to obtain a trained inverse text normalization model; wherein the inverse text corpus pair set includes at least one inverse text corpus pair; The reverse text corpus pair includes a number information corpus pair and a common collocation corpus pair, wherein the common collocation corpus pair is a corpus pair that does not need to be normalized for reverse text, and the number information corpus pair is a corpus pair that is directly translated from Chinese character descriptions, and the number information corpus pair includes a date corpus pair and an address information corpus pair; Before performing inverse text normalization training on the first sequence-to-sequence model through the inverse text corpus pair set, the method further includes: According to the obtained inverse text mapping table, a positive text corpus pair set is constructed and standard text information is obtained; Constructing a plurality of third sequence-to-sequence models, and performing text standardization training on each of the third sequence-to-sequence models through the text corpus pair set to obtain a plurality of trained second text standardization models; wherein each of the third sequence-to-sequence models is constructed based on a different neural network; Inputting the standard text information into each of the second normal text standardization models respectively to obtain corresponding multiple second standard text mapping information; Selecting target second standard text mapping information with the same text content from each of the second standard text mapping information, and determining matching target standard text information according to the target second standard text mapping information; The inverse text corpus pair set is constructed according to the target standard text information and the target second standard text mapping information.

2. The method according to claim 1, characterized in that The inverse text corpus pairs also include natural number corpus pairs, symbol information corpus pairs and / or moment scene corpus pairs.

3. The method according to claim 2, characterized in that The natural number corpus pairs include corpus pairs that are divisible by ten thousand digits and corpus pairs that are not divisible by ten thousand digits.

4. The method according to claim 1, characterized in that The source corpus and the target corpus in the inverse text corpus pair are both sentences.

5. A device for acquiring text of voice information, characterized in that: include: An initial recognition text acquisition module is used to acquire the voice information to be recognized and obtain the initial recognition text through voice recognition technology; An inverse text standardization execution module, used for inputting the initial recognition text into the trained inverse text standardization model to perform inverse text standardization processing on the initial recognition text, and using the processing result as the standard recognition text of the voice information; The text acquisition device of the voice information also includes: An inverse text normalization model acquisition module is used to construct a first sequence-to-sequence model, and perform inverse text normalization training on the first sequence-to-sequence model through an inverse text corpus pair set to obtain a trained inverse text normalization model; wherein the inverse text corpus pair set includes at least one inverse text corpus pair; The reverse text corpus pair includes a number information corpus pair and a common collocation corpus pair, wherein the common collocation corpus pair is a corpus pair that does not need to be normalized for reverse text, and the number information corpus pair is a corpus pair that is directly translated from Chinese character descriptions, and the number information corpus pair includes a date corpus pair and an address information corpus pair; The text acquisition device of the voice information also includes: A standard text information acquisition module is used to construct a positive text corpus pair set according to the acquired inverse text mapping table and obtain standard text information; A second main text standardization model acquisition module is used to construct a plurality of third sequence-to-sequence models, and perform main text standardization training on each of the third sequence-to-sequence models through the main text corpus pair set to obtain a plurality of trained second main text standardization models; wherein each of the third sequence-to-sequence models is constructed based on a different neural network; A second standard text mapping information acquisition module, used for inputting the standard text information into each of the second normal text standardization models respectively to obtain corresponding multiple second standard text mapping information; A target standard text information acquisition module, used to select target second standard text mapping information with the same text content from each of the second standard text mapping information, and determine matching target standard text information according to the target second standard text mapping information; The second inverse text corpus pair set is used to construct the inverse text corpus pair set according to the target standard text information and the target second standard text mapping information.

6. An electronic device comprising a memory, a processing device, and a computer program stored in the memory and executable on the processing device, characterized in that: When the processing device executes the program, the text acquisition method of voice information as described in any one of claims 1-4 is implemented.

7. A storage medium comprising computer executable instructions, wherein the computer executable instructions are used to execute the method for obtaining text of speech information as claimed in any one of claims 1 to 4 when executed by a computer processor.

Images