Speech text reconstruction method and device, storage medium and computer device

Abstract

Embodiments of the present application provide a speech text reconstruction method and device, a storage medium and a computer device. The method comprises: obtaining a standard speech text; determining whether the standard speech text meets a sentence quantity condition; if it is determined that the standard speech text does not meet the sentence quantity condition, generating a first reconstructed speech text according to the standard speech text through a sentence reconstruction manner; and if it is determined that the standard speech text meets the sentence quantity condition, generating a second reconstructed speech text according to the standard speech text through a model cutting manner, so that the computer device can reconstruct the obtained standard speech text, and the complexity of the speech text is reduced.

Description

[Technical Field]

[0001] This invention relates to the field of text reconstruction technology, specifically to a speech-text reconstruction method, apparatus, storage medium, and computer device. [Background Technology]

[0002] Traditional interaction methods suffer from high learning costs and cumbersome interactions, while voice control systems are gaining increasing attention due to their more natural interaction methods. A voice control system includes a semantic understanding system. The process involves: first, acquiring audio signals; performing speech recognition on the audio signals to convert them into speech-to-text; then, using the semantic understanding system to interpret the speech-to-text and parse it into command information; finally, post-processing using natural language generation.

[0003] The standardization and complexity of speech text determine the complexity, stability, and construction cost of semantic understanding systems. While speech control systems can segment long speech texts, if the speech text contains multiple intentions, multiple sentences, and related combinations, the complexity of the speech text cannot be reduced, resulting in high complexity when the semantic understanding system performs semantic understanding on the speech text. [Summary of the Invention]

[0004] In view of this, embodiments of the present invention provide a speech-text reconstruction method, apparatus, storage medium, and computer device to solve the problem of high complexity of speech-text in the prior art.

[0005] In a first aspect, embodiments of the present invention provide a speech-text reconstruction method, including:

[0006] Obtain standard speech text;

[0007] Determine whether the standard speech text meets the statement quantity requirement;

[0008] If it is determined that the standard speech text does not meet the statement quantity condition, then a first reconstructed speech text is generated based on the standard speech text through statement reconstruction.

[0009] If the standard speech text is determined to meet the sentence quantity condition, then a second reconstructed speech text is generated based on the standard speech text using a model segmentation method.

[0010] In one possible implementation, determining whether the standard speech text meets the sentence quantity condition includes:

[0011] Based on the standard speech text, output the number of the first sentences;

[0012] Determine whether the number of the first statement is greater than the first quantity threshold;

[0013] If it is determined that the number of the first statement is greater than the first number threshold, then the standard speech text is determined to meet the statement number condition.

[0014] If it is determined that the number of the first statement is less than or equal to the first number threshold, then the number of the second statement in the standard speech text is calculated.

[0015] Determine whether the number of the second statement is greater than the first quantity threshold;

[0016] If it is determined that the number of the second statement is less than or equal to the first number threshold, then it is determined that the standard speech text does not meet the statement number condition.

[0017] If it is determined that the number of the second statement is greater than the first number threshold, then the standard speech text is determined to meet the statement number condition.

[0018] In one possible implementation, calculating the number of second sentences in the standard speech text includes:

[0019] The number of verbs in the standard spoken text is counted using a verb list, and this number is used as the number of sentences in the second sentence; or,

[0020] The number of the second statement is output based on the standard speech text using a machine learning model.

[0021] In one possible implementation, generating the second reconstructed speech text based on the standard speech text using a model segmentation method includes:

[0022] The standard speech text is segmented using a neural network model to generate segmented text sequences, which include the segmented sequences and the standard speech text.

[0023] If it is determined that the number of the segmented sequences is greater than the second number threshold, then a second reconstructed speech text is generated based on the segmented sequences;

[0024] If the number of the segmented sequences is determined to be less than or equal to the second quantity threshold, then the second reconstructed speech text is generated based on the segmented sequence text using the segmentation vocabulary.

[0025] In one possible implementation, generating the first reconstructed speech text based on the standard speech text through sentence reconstruction includes:

[0026] The standard speech text is analyzed to generate word segmentation results and dependency relationships;

[0027] Based on the word segmentation results and dependency relationships, the standard speech text is reorganized to generate multiple single-semantic sentences;

[0028] Multiple single semantic statements are used as the first reconstructed speech text.

[0029] In one possible implementation, the word segmentation result includes multiple word segments, the dependency relationship includes a parallel relationship between at least two word segments and a word attribute corresponding to at least one word segment, the word attribute including a modifier attribute; the step of reorganizing the standard speech text according to the word segmentation result and the dependency relationship to generate multiple single semantic sentences includes:

[0030] Delete at least one modifying segment from the at least one segment, wherein the word attribute corresponding to the modifying segment is a modifying attribute;

[0031] Each parallel segment is recombined with at least one first segment to generate multiple single semantic sentences, wherein the first segment is a segment that does not have a parallel relationship, and the parallel segment is a segment that has a parallel relationship.

[0032] In one possible implementation, obtaining standard speech text includes:

[0033] The synonym list is used to find words in the acquired speech text that do not meet the standard; these words are then replaced with standard words, and the resulting speech text is used as the standard speech text. The synonym list includes at least one word that does not meet the standard and a corresponding standard word for each word. And / or,

[0034] Remove modifiers from the acquired speech text and use the speech text after removing modifiers as the standard speech text.

[0035] Secondly, embodiments of the present invention provide a speech-text reconstruction apparatus, comprising:

[0036] The acquisition module is used to acquire standard speech text;

[0037] The judgment module is used to determine whether the standard speech text meets the statement quantity condition;

[0038] The first generation module is used to generate a first reconstructed speech text based on the standard speech text if it is determined that the standard speech text does not meet the statement quantity condition.

[0039] The second generation module is used to generate a second reconstructed speech text based on the standard speech text by means of a model segmentation method if it is determined that the standard speech text meets the sentence quantity condition.

[0040] Thirdly, embodiments of the present invention provide a storage medium including a stored program, wherein, when the program is running, it controls the device where the storage medium is located to execute the speech-text reconstruction method described in the first aspect or any possible implementation thereof.

[0041] Fourthly, embodiments of the present invention provide a computer device, including a memory and a processor. The memory is used to store information including program instructions, and the processor is used to control the execution of the program instructions. When the program instructions are loaded and executed by the processor, they implement the steps of the speech-text reconstruction method in the first aspect or any possible implementation of the first aspect.

[0042] The present invention provides a technical solution for a speech-to-text reconstruction method, apparatus, storage medium, and computer device. The method involves: acquiring standard speech text; determining whether the standard speech text meets the statement quantity condition; if the standard speech text does not meet the statement quantity condition, generating a first reconstructed speech text based on the standard speech text using a statement reconstruction method; and if the standard speech text meets the statement quantity condition, generating a second reconstructed speech text based on the standard speech text using a model segmentation method. This enables the computer device to reconstruct the acquired standard speech text, reducing the complexity of the speech text. [Attached Image Description]

[0043] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0044] Figure 1 A flowchart of a speech-to-text reconstruction method provided in an embodiment of the present invention;

[0045] Figure 2 A flowchart for determining whether standard speech text meets the statement quantity condition is provided in an embodiment of the present invention;

[0046] Figure 3 A flowchart for generating a first reconstructed speech text is provided as an embodiment of the present invention;

[0047] Figure 4 A flowchart for generating a second reconstructed speech text is provided as an embodiment of the present invention;

[0048] Figure 5 This is a schematic diagram of the structure of a speech-to-text reconstruction device provided in an embodiment of the present invention;

[0049] Figure 6This is a schematic diagram of a computer device provided in an embodiment of the present invention.

Detailed Implementation Methods

[0050] To better understand the technical solution of the present invention, the embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0051] It should be understood that the described embodiments are merely some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0052] The terminology used in the embodiments of this invention is for the purpose of describing particular embodiments only and is not intended to limit the invention. The singular forms “a,” “the,” and “the” as used in the embodiments of this invention and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise.

[0053] It should be understood that the term "and / or" used in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0054] It should be understood that although terms such as first, second, third, etc., may be used to describe numbers in embodiments of the present invention, these numbers should not be limited to these terms. These terms are only used to distinguish numbers from each other. For example, without departing from the scope of embodiments of the present invention, a first number may also be referred to as a second number, and similarly, a second number may also be referred to as a first number.

[0055] Depending on the context, the word "if" as used here can be interpreted as "when," "when," "in response to determination," or "in response to detection." Similarly, depending on the context, the phrase "if determination" or "if detection (of the stated condition or event)" can be interpreted as "when determination," "in response to determination," "when detection (of the stated condition or event)," or "in response to detection (of the stated condition or event)."

[0056] Figure 1 A flowchart of a speech-to-text reconstruction method provided in an embodiment of the present invention is shown below. Figure 1 As shown, the method includes:

[0057] Step 101: The computer device acquires the standard speech text.

[0058] In this embodiment of the invention, the computer equipment includes, but is not limited to, vehicle terminals, mobile phones, tablet computers, portable PCs, desktop computers, wearable devices, etc.

[0059] Before performing step 101, the computer device also includes: acquiring voice text.

[0060] Computer equipment includes a microphone, which collects speech. The computer equipment recognizes the speech and generates speech-to-text. Based on the speech-to-text, the computer equipment generates standard speech-to-text.

[0061] Step 102: The computer device determines whether the standard speech text meets the statement quantity condition. If yes, proceed to step 104; otherwise, proceed to step 103.

[0062] Step 103: The computer device generates the first reconstructed speech text based on the standard speech text through sentence reconstruction.

[0063] In this embodiment of the invention, the first reconstructed speech text includes multiple single semantic sentences. Thus, when the semantic understanding system performs semantic recognition on the speech text, it recognizes multiple single semantic sentences, which reduces the complexity and construction cost of the semantic understanding system and increases its stability.

[0064] Step 104: The computer device generates a second reconstructed speech text based on the standard speech text using a model segmentation method.

[0065] In this embodiment of the invention, the second reconstructed speech text includes multiple single semantic sentences. Thus, when the semantic understanding system performs semantic recognition on the speech text, it recognizes multiple single semantic sentences, which reduces the complexity and construction cost of the semantic understanding system and increases its stability.

[0066] This invention provides a speech-to-text reconstruction method that involves: acquiring standard speech text; determining whether the standard speech text meets the condition for the number of sentences; if the standard speech text does not meet the condition, generating a first reconstructed speech text based on the standard speech text using a sentence reconstruction method; and if the standard speech text meets the condition, generating a second reconstructed speech text based on the standard speech text using a model segmentation method. This enables a computer device to reconstruct the acquired standard speech text, generating multiple single-semantic sentences and reducing the complexity of the speech text.

[0067] In one possible implementation, step 101 may specifically include: the computer device searching for non-standard words in the acquired speech text through a synonym list; replacing the non-standard words in the speech text with standard words, and using the speech text after replacing the non-standard words with standard words as the standard speech text, wherein the synonym list includes at least one non-standard word and a standard word corresponding to each non-standard word; and / or deleting modifiers in the acquired speech text, and using the speech text after deleting modifiers as the standard speech text.

[0068] In this embodiment of the invention, modifiers include adjectives, modal particles, adverbs, etc. For example, the audio text is "Open the car window, umm, turn off the air conditioner, wow," which includes the modal particles "umm" and "wow." If the computer device removes "umm" and "wow," the audio text becomes "Open the car window, turn off the air conditioner." The computer device processes the audio text using a predefined thesaurus, replacing non-standard words in the audio text with standard words from the thesaurus, thus standardizing the audio text. Non-standard words include "turn off," and standard words include "close," with "close" being the standard word corresponding to "turn off." The computer device replaces "close" with "close" in the audio text, resulting in "Open the car window, turn off the air conditioner," i.e., the standard audio text is "Open the car window, turn off the air conditioner." However, this application does not limit the execution order of deleting modifiers or replacing non-standard words with standard words.

[0069] In one possible implementation, Figure 2 A flowchart for determining whether standard speech text meets the statement quantity condition is provided in an embodiment of the present invention, such as... Figure 2 As shown, step 102 may specifically include:

[0070] Step 1021: The computer device outputs the number of the first sentences based on the standard speech text.

[0071] In this embodiment of the invention, the computer device outputs the number of first sentences based on standard speech text using a neural network model. The neural network model includes a Recurrent Neural Network (RNN) model. The computer device inputs the standard speech text into the pre-trained RNN model to obtain the number of first sentences. For example, if the standard speech text is "Open the car window and turn off the air conditioner," the number of first sentences is 2; if the standard speech text is "Help me open the car window and sunroof," the number of first sentences is 1.

[0072] Step 1022: The computer device determines whether the number of the first statement is greater than the first quantity threshold. If yes, then proceed to step 1023; otherwise, proceed to step 1024.

[0073] In this embodiment of the invention, the first quantity threshold is 1. The computer device determines whether the number of the first statement is greater than 1. If yes, then step 1023 is executed; if no, then step 1024 is executed.

[0074] Step 1023: The computer equipment determines the number of sentences that the standard speech text meets.

[0075] Step 1024: The computer device calculates the number of second sentences in the standard speech text.

[0076] In this embodiment of the invention, the computer device counts the number of verbs in the standard speech text using a verb list and uses the number of verbs as the number of second sentences; or, it outputs the number of second sentences based on the standard speech text using a machine learning model.

[0077] For example, the computer device determines that the number of sentences in the first speech text is less than or equal to 1. The computer device can then recalculate the number of sentences in the standard speech text by traversing the verb list or using a machine learning model. When the computer device calculates the number of sentences in the standard speech text by traversing the verb list, if the standard speech text is "open the car window and turn off the air conditioner," then the verb list indicates that the verbs in the standard speech text are "open" and "turn off," with a verb count of 2, and therefore a second sentence count of 2. Machine learning models include tree models or ensemble learning models. The computer device can use a pre-trained tree model or ensemble learning model to output the number of sentences in the standard speech text based on the standard speech text.

[0078] Step 1025: The computer device determines whether the number of the second statement is greater than the first quantity threshold. If yes, proceed to step 1027; otherwise, proceed to step 1026.

[0079] In this embodiment of the invention, the first quantity threshold is 1. The computer device determines whether the number of the second statement is greater than 1. If yes, then step 1023 is executed; if no, then step 1024 is executed.

[0080] Step 1026: The computer equipment determines that the standard speech text does not meet the statement quantity condition.

[0081] Step 1027: The computer equipment determines the number of sentences that the standard speech text meets.

[0082] In one possible implementation, Figure 3 A flowchart for generating a first reconstructed speech text is provided as an embodiment of the present invention, such as... Figure 3 As shown, step 103 may specifically include:

[0083] Step 1031: The computer device analyzes the standard speech text to generate a word segmentation result and a dependency relationship.

[0084] In an embodiment of the present invention, the word segmentation result includes multiple word segments, and the dependency relationship includes a parallel relationship between at least two word segments and a word attribute corresponding to at least one word segment. The word attribute includes a modifying attribute. The modifying attribute includes a conjunctive adverbial, an adjective attribute, a modal particle attribute, etc.

[0085] The computer device may include word segmentation software. For example, the word segmentation software is the open-source software jieba. When the standard speech text is "Help me open the window and the sunroof", the computer device segments the standard speech text into ["Help", "me", "open", "window", "and", "sunroof"], and the multiple word segments are "Help", "me", "open", "window", "and", and "sunroof", and the word segmentation result is ["Help", "me", "open", "window", "and", "sunroof"].

[0086] The computer device may also include dependency relationship discrimination software. For example, the dependency relationship discrimination software is the open-source software hanlp. The computer device analyzes the multiple word segments through the dependency relationship discrimination software and analyzes that the relationship between "window" and "sunroof" is a parallel relationship; "and" is a conjunction, and the word attribute is a modifying attribute.

[0087] Step 1032: The computer device reorganizes the standard speech text according to the word segmentation result and the dependency relationship to generate multiple single-semantic sentences.

[0088] In an embodiment of the present invention, the computer device deletes at least one modifying word segment in at least one word segment, and the word attribute corresponding to the modifying word segment is a modifying attribute; reorganizes each parallel word segment with at least one first word segment to generate multiple single-semantic sentences. The first word segment is a word segment without a parallel relationship, and the parallel word segment is a word segment with a parallel relationship.

[0089] For example, the computer device analyzes that ["and"] in ["Help", "me", "open", "window", "and", "sunroof"] is a modifying word segment, ["window", "sunroof"] are both parallel word segments, and ["Help", "me", "open"] are all first word segments. The computer device deletes ["and"]; reorganizes each parallel word segment with 3 first word segments. For example, it reorganizes "window" with "Help", "me", "open" to generate the single-semantic sentence "Help me open the window"; it reorganizes "sunroof" with "Help", "me", "open" to generate the single-semantic sentence "Help me open the sunroof".

[0090] Step 1033: The computer device uses the multiple single-semantic sentences as the first reconstructed speech text.

[0091] In this embodiment of the invention, the reconstruction result of the standard speech text is multiple single semantic sentences. For example, the first reconstructed speech text is ["Help me open the car window", "Help me open the sunroof"].

[0092] This allows computer devices to reconstruct speech text, breaking down speech text with multiple intentions, multiple sentences, and parallel relationships and related combinations into multiple single semantic sentences. This facilitates the computer device's understanding of the user's speech input, reduces the complexity and construction cost of the semantic understanding system, and increases the stability of the semantic understanding system.

[0093] In one possible implementation, Figure 4 A flowchart for generating a second reconstructed speech text is provided as an embodiment of the present invention, such as... Figure 4 As shown, step 104 may specifically include:

[0094] Step 1041: The computer device segments the standard speech text using a neural network model to generate segmented text sequences, which include the segmented sequences and the standard speech text.

[0095] In this embodiment of the invention, the neural network model includes an RNN model. The cutting sequence is a sequence of “[I*E]”, where I is the cut start and E is the cut end. A cutting sequence may include at least one cut start and one cut end.

[0096] For example, given the standard speech text "Open the car window and turn off the air conditioner," a computer device uses an RNN model to segment "Open the car window and turn off the air conditioner," generating a segmented text sequence. This segmented text sequence is "Open / I Car Window / E Close / I Air Conditioner / E," containing two segmentation sequences. One sequence begins with the segment word "open" and ends with the segment word "car window." The other sequence begins with the segment word "close" and ends with the segment word "air conditioner." Alternatively, given the standard speech text "Open the car window and then turn off the air conditioner," a computer device uses an RNN model to segment "Open the car window and then turn off the air conditioner," generating a segmented text sequence. This segmented text sequence is "Open / I Car Window / I Again / I Put / I Air Conditioner / I Turn Off / E," containing one segmentation sequence with five segmentation starts and one segmentation end. The five segmentation starts correspond to the segment words "open," "car window," "again," "put," and "air conditioner," respectively, and the segmentation end corresponds to the segment word "turn off."

[0097] Step 1042: The computer device determines whether the number of cut sequences is greater than the second quantity threshold. If yes, proceed to step 1043; otherwise, proceed to step 1044.

[0098] In this embodiment of the invention, for example, the second quantity threshold is 1. The computer device determines whether the number of cut sequences is greater than 1. If yes, then step 1043 is executed; if no, then step 1044 is executed.

[0099] Step 1043: The computer device generates a second reconstructed speech text based on the segmentation sequence.

[0100] In this embodiment of the invention, for example, computer device I is the starting point for segmentation, and E is the ending point, segmenting the standard speech text into ["Open the car window", "Turn off the air conditioner"]. The second reconstructed speech text is ["Open the car window", "Turn off the air conditioner"]. Thus, the computer device reconstructs the speech text, reconstructing speech text with multiple intentions, multiple sentences, and parallel relationships and related combinations into multiple single semantic sentences. This facilitates the computer device's understanding of user-input speech, reduces the complexity and construction cost of the semantic understanding system, and increases the stability of the semantic understanding system.

[0101] Step 1044: The computer device generates a second reconstructed speech text based on the segmented word list and the segmented sequence text.

[0102] In this embodiment of the invention, the computer device obtains a pre-established segmentation word list, traverses the segmentation words included in the segmentation sequence text from the segmentation word list, deletes the segmentation words in the segmentation sequence text, and generates a second reconstructed speech text based on the segmentation sequence text after deleting the segmentation words.

[0103] For example, if the segmented text sequence is "open / I car window / I again / I turn off / I air conditioner / I turn it off / E", and the segmented word is "again", the computer device will delete "again", and the second reconstructed speech text will be ["open the car window", "turn off the air conditioner"].

[0104] This enables computer devices to reorganize speech text, transforming speech text with multiple intentions, multiple sentences, and parallel relationships and related combinations into multiple single semantic sentences. This facilitates the computer device's understanding of the user's speech input, reduces the complexity and construction cost of the semantic understanding system, and increases the stability of the semantic understanding system.

[0105] Figure 5 This is a schematic diagram of the structure of a speech-to-text reconstruction device provided in an embodiment of the present invention, as shown below. Figure 5 As shown, the device includes: an acquisition module 11, a judgment module 12, a first generation module 13, and a second generation module 14. The acquisition module 11 is connected to the judgment module 12, and the judgment module 12 is connected to the first generation module 13 and the second generation module 14.

[0106] The acquisition module 11 is used to acquire standard speech text; the judgment module 12 is used to judge whether the standard speech text meets the sentence quantity condition; the first generation module 13 is used to generate a first reconstructed speech text based on the standard speech text by means of sentence reconstruction if the standard speech text does not meet the sentence quantity condition; the second generation module 14 is used to generate a second reconstructed speech text based on the standard speech text by means of model segmentation if the standard speech text meets the sentence quantity condition.

[0107] In this embodiment of the invention, the judgment module 12 includes: an output submodule 121, a first judgment submodule 122, a first determination submodule 123, a calculation submodule 124, a second judgment submodule 125, a second determination submodule 126, and a third determination submodule 127. The output submodule 121 is connected to the first judgment submodule 122; the first judgment submodule 122 is connected to the first determination submodule 123 and the calculation submodule 124; the calculation submodule 124 is connected to the second judgment submodule 125; and the second judgment submodule 125 is connected to the second determination submodule 126 and the third determination submodule 127.

[0108] The output submodule 121 is used to output the first number of sentences based on the standard speech text; the first judgment submodule 122 is used to judge whether the first number of sentences is greater than a first number threshold; the first determination submodule 123 is used to determine that the standard speech text meets the sentence count condition if the first judgment submodule 122 judges that the first number of sentences is greater than the first number threshold; the calculation submodule 124 is used to calculate the second number of sentences in the standard speech text if the first judgment submodule 122 judges that the first number of sentences is less than or equal to the first number threshold; the second judgment submodule 125 is used to judge whether the second number of sentences is greater than the first number threshold; the second determination submodule 126 is used to determine that the standard speech text does not meet the sentence count condition if the second judgment submodule 125 judges that the second number of sentences is less than or equal to the first number threshold; the third determination submodule 127 is used to determine that the standard speech text meets the sentence count condition if the second judgment submodule 125 judges that the second number of sentences is greater than the first number threshold.

[0109] In this embodiment of the invention, the calculation submodule 124 is specifically used to count the number of verbs in the standard speech text through a verb list and use the number of verbs as the number of second sentences; or, through a machine learning model, output the number of second sentences based on the standard speech text.

[0110] In this embodiment of the invention, the second generation module 14 includes: a first generation submodule 141, a second generation submodule 142, and a third generation submodule 143. The first generation submodule 141 is connected to the second generation submodule 142 and the third generation submodule 143.

[0111] The first generation submodule 141 is used to segment the standard speech text using a neural network model to generate segmented sequence text, which includes the segmented sequence and the standard speech text; the second generation submodule 142 is used to generate a second reconstructed speech text based on the segmented sequence if the number of segmented sequences is determined to be greater than a second quantity threshold; the third generation submodule 143 is used to generate a second reconstructed speech text based on the segmented sequence text by segmenting a word list if the number of segmented sequences is determined to be less than or equal to the second quantity threshold.

[0112] In this embodiment of the invention, the first generation module 13 includes: a fourth generation submodule 131, a fifth generation submodule 132, and a submodule 133. The fourth generation submodule 131 is connected to the fifth generation submodule 132, and the fifth generation submodule 132 is connected to the submodule 133.

[0113] The fourth generation submodule 131 is used to analyze the standard speech text and generate word segmentation results and dependency relations; the fifth generation submodule 132 is used to reorganize the standard speech text according to the word segmentation results and dependency relations to generate multiple single semantic sentences; the submodule 133 is used to use the multiple single semantic sentences as the first reconstructed speech text.

[0114] In this embodiment of the invention, the word segmentation result includes multiple words, and the dependency relationship includes the parallel relationship between at least two words and the word attribute corresponding to at least one word. The word attribute includes the modification attribute. The fifth generation submodule 132 is specifically used to delete at least one modifying word in at least one word, and the word attribute corresponding to the modifying word is the modification attribute. Each parallel word is recombined with at least one first word to generate multiple single semantic sentences. The first word is a word that does not have a parallel relationship, and the parallel words are words that have a parallel relationship.

[0115] In this embodiment of the invention, the acquisition module is specifically used to find out the non-standard words in the acquired speech text through a synonym list; replace the non-standard words in the speech text with standard words, and use the speech text after replacing the non-standard words with standard words as the standard speech text, wherein the synonym list includes at least one non-standard word and the standard word corresponding to each non-standard word; and / or, delete the modifiers in the acquired speech text, and use the speech text after deleting the modifiers as the standard speech text.

[0116] This invention provides a speech-to-text reconstruction method that involves: acquiring standard speech text; determining whether the standard speech text meets the condition for the number of sentences; if the standard speech text does not meet the condition, generating a first reconstructed speech text based on the standard speech text using a sentence reconstruction method; and if the standard speech text meets the condition, generating a second reconstructed speech text based on the standard speech text using a model segmentation method. This enables a computer device to reconstruct the acquired standard speech text, generating multiple single-semantic sentences and reducing the complexity of the speech text.

[0117] This invention provides a storage medium that includes a stored program. When the program runs, it controls the device where the storage medium is located to execute the steps of the above-described speech-text reconstruction method. For a detailed description, please refer to the embodiments of the above-described speech-text reconstruction method.

[0118] This invention provides a computer device including a memory and a processor. The memory is used to store information including program instructions, and the processor is used to control the execution of the program instructions. When the program instructions are loaded and executed by the processor, they implement the steps of the above-described speech-text reconstruction method. For a detailed description, please refer to the above-described speech-text reconstruction method embodiments.

[0119] Figure 6 This is a schematic diagram of a computer device provided in an embodiment of the present invention. Figure 6 As shown, the computer device 30 in this embodiment includes a processor 31, a memory 32, and a computer program 33 stored in the memory 32 and executable on the processor 31. When the processor 31 executes the computer program 33, it implements the speech-to-text reconstruction method described in this embodiment. To avoid repetition, these details are not elaborated here. Alternatively, when the processor 31 executes the computer program, it implements the functions of each model / unit in the speech-to-text reconstruction apparatus described in this embodiment. To avoid repetition, these details are not elaborated here.

[0120] Computer device 30 includes, but is not limited to, processor 31 and memory 32. Those skilled in the art will understand that... Figure 6 This is merely an example of computer device 30 and does not constitute a limitation on computer device 30. It may include more or fewer components than shown, or combine certain components, or different components. For example, computer device 30 may also include input / output devices, network access devices, buses, etc.

[0121] The processor 31 may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor.

[0122] The memory 32 can be an internal storage unit of the computer device 30, such as a hard disk or RAM of the computer device 30. The memory 32 can also be an external storage device of the computer device 30, such as a plug-in hard disk, Smart Media Card (SMC), Secure Digital (SD) card, or Flash Card equipped on the computer device 30. Furthermore, the memory 32 can include both internal and external storage units of the computer device 30. The memory 32 is used to store computer programs and other programs and data required by the computer device 30. The memory 32 can also be used to temporarily store data that has been output or will be output.

[0123] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0124] In the embodiments provided by this invention, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection between devices or units through some interfaces, and may be electrical, mechanical, or other forms.

[0125] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0126] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional units.

[0127] The integrated units implemented as software functional units described above can be stored in a computer-readable storage medium. These software functional units, stored in a storage medium, include several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute some steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0128] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A speech-to-text reconstruction method, characterized in that, include: Obtain standard speech text; Determine whether the standard speech text meets the statement quantity requirement; If it is determined that the standard speech text does not meet the statement quantity condition, then a first reconstructed speech text is generated based on the standard speech text through statement reconstruction. If the standard speech text is determined to meet the sentence quantity condition, then a second reconstructed speech text is generated based on the standard speech text using a model segmentation method.

2. The method according to claim 1, characterized in that, The determination of whether the standard speech text meets the sentence quantity condition includes: Based on the standard speech text, output the number of the first sentences; Determine whether the number of the first statement is greater than the first quantity threshold; If it is determined that the number of the first statement is greater than the first number threshold, then the standard speech text is determined to meet the statement number condition. If it is determined that the number of the first statement is less than or equal to the first number threshold, then the number of the second statement in the standard speech text is calculated. Determine whether the number of the second statement is greater than the first quantity threshold; If it is determined that the number of the second statement is less than or equal to the first number threshold, then it is determined that the standard speech text does not meet the statement number condition. If it is determined that the number of the second statement is greater than the first number threshold, then the standard speech text is determined to meet the statement number condition.

3. The method according to claim 2, characterized in that, The calculation of the number of second sentences in the standard speech text includes: The number of verbs in the standard spoken text is counted using a verb list, and this number is used as the number of sentences in the second sentence; or, The number of the second statement is output based on the standard speech text using a machine learning model.

4. The method according to claim 1, characterized in that, The step of generating a second reconstructed speech text based on the standard speech text using a model segmentation method includes: The standard speech text is segmented using a neural network model to generate segmented text sequences, which include the segmented sequences and the standard speech text. If it is determined that the number of the segmented sequences is greater than the second number threshold, then a second reconstructed speech text is generated based on the segmented sequences; If the number of the segmented sequences is determined to be less than or equal to the second quantity threshold, then the second reconstructed speech text is generated based on the segmented sequence text using the segmentation vocabulary.

5. The method according to claim 1, characterized in that, The step of generating a first reconstructed speech text based on the standard speech text through sentence reconstruction includes: The standard speech text is analyzed to generate word segmentation results and dependency relationships; Based on the word segmentation results and dependency relationships, the standard speech text is reorganized to generate multiple single-semantic sentences; Multiple single semantic statements are used as the first reconstructed speech text.

6. The method according to claim 5, characterized in that, The word segmentation result includes multiple word segments, and the dependency relationship includes a parallel relationship between at least two word segments and a word attribute corresponding to at least one word segment. The word attribute includes a modification attribute. The standard speech text is reorganized based on the word segmentation result and the dependency relationship to generate multiple single-semantic sentences, including: Delete at least one modifying segment from the at least one segment, wherein the word attribute corresponding to the modifying segment is a modifying attribute; Each parallel segment is recombined with at least one first segment to generate multiple single semantic sentences, wherein the first segment is a segment that does not have a parallel relationship, and the parallel segment is a segment that has a parallel relationship.

7. The method according to claim 1, characterized in that, The acquisition of standard speech text includes: The synonym list is used to find words in the acquired speech text that do not meet the standard; these words are then replaced with standard words, and the resulting speech text is used as the standard speech text. The synonym list includes at least one word that does not meet the standard and a corresponding standard word for each word. And / or, Remove modifiers from the acquired speech text and use the speech text after removing modifiers as the standard speech text.

8. A speech-to-text reconstruction device, characterized in that, include: The acquisition module is used to acquire standard speech text; The judgment module is used to determine whether the standard speech text meets the statement quantity condition; The first generation module is used to generate a first reconstructed speech text based on the standard speech text if it is determined that the standard speech text does not meet the statement quantity condition. The second generation module is used to generate a second reconstructed speech text based on the standard speech text by means of a model segmentation method if it is determined that the standard speech text meets the sentence quantity condition.

9. A storage medium, characterized in that, The storage medium includes a stored program, wherein, when the program is executed, it controls the device where the storage medium is located to perform the speech-text reconstruction method according to any one of claims 1 to 7.

10. A computer device comprising a memory and a processor, the memory for storing information including program instructions, and the processor for controlling the execution of the program instructions, characterized in that, When the program instructions are loaded and executed by the processor, the speech-text reconstruction method according to any one of claims 1 to 7 is implemented.

Images