Voice transcription hot word automatic extraction method and device, electronic equipment and storage medium

Abstract

The embodiment of the invention discloses a voice transcription hot word automatic extraction method and device, electronic equipment and a storage medium, and the method comprises the steps: calculating the word accuracy of a corresponding transcription text based on a standard text of a target audio file, and judging whether the standard text is a hot word extraction object or not based on the word accuracy, the corresponding transliteration text is a text obtained by transliteration of the target audio file by the voice transliteration model; when the standard text is the hot word extraction object, obtaining word segmentation processing results of the standard text and the corresponding transliteration text to correspondingly obtain a plurality of standard segmented words and a plurality of transliteration segmented words; obtaining the standard segmented words which are different from the corresponding transliteration segmented words and are not in the transliteration text in the standard segmented words to obtain alternative hot words; and determining a hot word file of the voice transcription model based on the alternative hot words. According to the embodiment of the invention, the cost for extracting the hot words can be reduced, and the hot word extraction efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a method, device, electronic device and storage medium for automatically extracting hot words in speech transcription. Background Art

[0002] When using a speech recognition model (Automatic Speech Recognition, ASR) to recognize and transcribe speech, it is usually not possible to completely recognize speech correctly, and recognition errors often occur in the process of recognizing non-common words such as specific professional terms and place names. In the prior art, a hot word file is set for the speech recognition model to help correctly recognize words that are prone to errors, but the hot word file requires manual extraction of hot words to determine the hot word file, which requires high labor costs and low extraction efficiency. Summary of the invention

[0003] The embodiments of the present invention provide a method, device, electronic device and storage medium for automatically extracting hot words in speech transcription, which can reduce the cost required for extracting hot words and improve the efficiency of hot word extraction.

[0004] In a first aspect, an embodiment of the present invention provides a method for automatically extracting hot words in speech transcription, comprising:

[0005] Calculating the word accuracy of the corresponding transcribed text based on the standard text of the target audio file, and judging whether the standard text is a hot word extraction object based on the word accuracy, the corresponding transcribed text is the text obtained by transcribing the target audio file by the speech transcription model;

[0006] When the standard text is a hot word extraction object, the word segmentation processing results of the standard text and the corresponding transcribed text are obtained to obtain a plurality of standard word segments and a plurality of transcribed word segments;

[0007] Obtaining standard participles that are different from corresponding transcribed participles and are not in the transcribed text to obtain candidate hot words; and

[0008] A hot word file of the speech transcription model is determined based on the candidate hot words.

[0009] In a second aspect, an embodiment of the present invention provides a device for automatically extracting hot words by voice transcription, comprising:

[0010] An extraction object judgment module is used to calculate the word accuracy of the corresponding transcribed text based on the standard text of the target audio file, and judge whether the standard text is a hot word extraction object based on the word accuracy, and the corresponding transcribed text is the text obtained by transcribing the target audio file by the speech transcription model;

[0011] A word segmentation acquisition module, used for obtaining the word segmentation processing results of the standard text and the corresponding transcribed text to obtain a plurality of standard word segments and a plurality of transcribed word segments when the standard text is a hot word extraction object;

[0012] A candidate hot word acquisition module is used to acquire standard participles that are different from corresponding transcribed participles and are not in the transcribed text to obtain candidate hot words; and

[0013] A hot word file determination module is used to determine the hot word file of the speech transcription model based on the candidate hot words.

[0014] In a third aspect, an embodiment of the present invention further provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the method for automatically extracting hot words from speech transcription as described in any one of the embodiments of the present invention is implemented.

[0015] In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements a method for automatically extracting hot words from speech transcription as described in any one of the embodiments of the present invention.

[0016] The embodiments of the present invention provide a method, device, electronic device and storage medium for automatically extracting hot words in speech transcription. The method calculates the accuracy of the corresponding transcribed text transcribed by the speech transcription model, and judges whether the standard text of the target audio file is a hot word extraction object based on the accuracy. When the standard text is a hot word extraction object, the segmentation of the standard text and the corresponding transcribed text is obtained, and then the standard segmentation that is different from the corresponding transcribed segmentation and different from each transcribed segmentation is obtained as an alternative hot word. The hot word file of the speech transcription model is further determined based on the alternative hot words. The method can automatically extract the standard vocabulary corresponding to the incorrectly transcribed vocabulary in the standard text based on the transcribed text, thereby eliminating the need for manual extraction of hot words, reducing the cost required for extracting hot words, and improving the efficiency of hot word extraction. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solution of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0018] Figure 1 It is a flow chart of a method for automatically extracting hot words from speech transcription provided by an embodiment of the present invention;

[0019] Figure 2It is another flow chart of the method for automatically extracting hot words from speech transcription provided by an embodiment of the present invention;

[0020] Figure 3 It is another flow chart of the method for automatically extracting hot words from speech transcription provided by an embodiment of the present invention;

[0021] Figure 4 It is another flow chart of the method for automatically extracting hot words from speech transcription provided by an embodiment of the present invention;

[0022] Figure 5 It is a structural schematic diagram of a device for automatically extracting hot words from speech transcription provided by an embodiment of the present invention;

[0023] Figure 6 It is a structural schematic diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0024] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0025] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0026] Figure 1 A flowchart of a method for automatically extracting hot words from speech transcription provided in an embodiment of the present invention is provided. This embodiment can be applied to extracting hot word scenarios from a speech transcription model hot word file. The method can be performed by an automatic extraction device for hot words from speech transcription provided in an embodiment of the present invention, and the device can be implemented in software and / or hardware. In a specific embodiment, the device can be integrated in an electronic device, such as a computer, a server, etc. The following embodiments will be described by taking the device integrated in an electronic device as an example. Reference Figure 1 , the method may specifically include the following steps:

[0027] Step 101, based on the standard text of the target audio file, the word accuracy of the corresponding transcribed text is calculated, and based on the word accuracy, whether the standard text is a hot word extraction object is determined, and the corresponding transcribed text is the text obtained by transcribing the target audio file by the speech transcription model. This step can facilitate the acquisition of a standard text that can be used as a hot word extraction object, and further facilitates hot word extraction based on the standard text that can be used as a hot word extraction object.

[0028] Optionally, the target audio file may include multiple target audio files or one target audio file.

[0029] Optionally, the process of calculating the word accuracy of the corresponding transcribed text based on the standard text of the target audio file includes: for each standard text in multiple audio files, calculating the word accuracy of the transcribed text corresponding to the current text based on the current standard text.

[0030] Specifically, before step 101, a suitable recording device, such as a microphone, can be selected, a suitable sampling rate and bit rate can be set, and the original audio file can be recorded in a quiet environment. After that, the original audio file can be subjected to noise reduction, invalid segments can be removed, the volume can be increased, the clarity can be improved, and / or segmentation processing can be performed to obtain the above-mentioned target audio file.

[0031] Specifically, the target audio file may also be obtained from a database storing audio files.

[0032] Specifically, after obtaining the target audio file, the target audio file may be transcribed using a speech transcription model to obtain the transcribed text.

[0033] Specifically, the standard text of the target file can be obtained by manually annotating and correcting the written text.

[0034] Optionally, the process of calculating the word accuracy of the corresponding transcribed text based on the standard text of the target audio file, and judging whether the standard text is a hot word extraction object based on the word accuracy includes:

[0035] Based on the first normalization strategy, the standard text and the corresponding transcribed text are normalized to obtain a first normalized standard text and a first normalized transcribed text.

[0036] Specifically, the first normalization strategy mentioned above can be understood as a text processing strategy determined based on the text norm required by the method for calculating word accuracy.

[0037] Optionally, the above-mentioned process of normalizing the standard text and the corresponding transcribed text based on the first normalization strategy includes: removing punctuation marks, unifying the case of letters, deleting unnecessary spaces and line breaks, etc. from the standard text and the corresponding written text.

[0038] Optionally, the process of calculating the word accuracy of the corresponding transcribed text based on the standard text of the target audio file, and judging whether the standard text is a hot word extraction object based on the word accuracy includes:

[0039] Based on the dynamic standardization algorithm, the number of correct words, deleted words and inserted words in the first standard transcribed text are obtained, and the word accuracy is calculated based on the number of correct words, deleted words, inserted words and the number of words in the first standard text.

[0040] Optionally, the process of calculating the word accuracy based on the correct word count, the deleted word count, the inserted word count and the word count of the first standard text is performed based on the following formula:

[0041] Acc=(HI) / N×100%

[0042] Among them, Acc represents the word accuracy rate, H represents the correct number of words, D represents the number of deleted words, I represents the number of inserted words; N represents the number of words in the first standard text.

[0043] Specifically, the above character accuracy rate can also be calculated based on the calculation formula of the Yinxie character accuracy rate.

[0044] Specifically, when both the standard text and the transcribed text meet the corresponding text specifications, the correct number of words, deleted number of words and inserted number of words of the transcribed text can be directly obtained based on the dynamic specification algorithm, and the word accuracy can be calculated based on the corresponding correct number of words, deleted number of words, inserted number of words and the number of words of the standard text.

[0045] Optionally, the process of calculating the word accuracy of the corresponding transcribed text based on the standard text of the target audio file, and judging whether the standard text is a hot word extraction object based on the word accuracy includes:

[0046] When the word accuracy is less than a preset accuracy threshold, the standard text is determined to be a hot word extraction object.

[0047] Specifically, the accuracy threshold can be set based on empirical data, or can be set according to test results after multiple tests. Specifically, the accuracy threshold can be set to 0.8.

[0048] It can be understood that the lower the character accuracy rate of the transcribed text, the greater the difference between the standard text and the transcribed text, that is, the greater the probability that the vocabulary of the standard text corresponding to the transcribed text has a transcription error, and the greater the probability of becoming a hot word. Therefore, judging whether the standard text is an object for extracting hot words based on whether the character accuracy rate is less than a preset threshold can facilitate the accurate and efficient extraction of hot words.

[0049] Specifically, it is also possible to judge whether the standard text is an object for extracting hot words based on the magnitude of the difference between the character accuracy rate and the target character accuracy rate.

[0050] Step 102, when the standard text is an object for extracting hot words, obtain the word segmentation processing results of the standard text and the corresponding transcribed text to obtain multiple standard word segments and multiple transcribed word segments. This step can facilitate obtaining candidate hot words among multiple standard word segments based on multiple transcribed word segments.

[0051] Optionally, the process of obtaining the word segmentation processing results of the standard text and the corresponding transcribed text to obtain multiple standard word segments and multiple transcribed word segments includes:

[0052] Perform normalization processing on the standard text and the corresponding transcribed text based on the second normalization strategy to obtain the second canonical standard text and the second canonical transcribed text.

[0053] Specifically, the above-mentioned second normalization strategy can be understood as a text processing strategy determined based on the text norms required by the word segmentation method.

[0054] Optionally, the process of performing normalization processing on the standard text and the corresponding transcribed text based on the second normalization strategy includes: removing stop words and punctuation marks in the standard text and the corresponding written text.

[0055] Specifically, the above-mentioned stop words are common words that are usually ignored in text processing because they appear frequently in the text but contribute little to the theme or meaning of the text. These words usually include common words such as "de", "a", "o", etc.

[0056] It can be understood that stop words and punctuation marks usually do not help much in text analysis, but they will introduce noise and increase the computational burden. Removing stop words and punctuation marks can facilitate improving the word segmentation accuracy and saving computational resources.

[0057] Optionally, the process of obtaining the word segmentation processing results of the standard text and the corresponding transcribed text to obtain multiple standard word segments and multiple transcribed word segments includes: using a word segmentation tool to perform word segmentation processing on the second canonical standard text and the second canonical transcribed text to obtain multiple standard word segments and multiple transcribed word segments.

[0058] Specifically, when the standard text and the corresponding transcribed text meet the corresponding text specifications, the standard text and the corresponding transcribed text may be directly segmented to obtain standard segmented words and transcribed segmented words respectively.

[0059] In an optional specific example, for the following standard text and transcribed text:

[0060] Standard text: First Branch Product Development Center, Software Development Department 2, Product Quality Inspection Group.

[0061] Transcribed text: The first branch focuses on product development, not software development, and is a group of products.

[0062] After normalization based on the second normalization strategy and segmentation using the segmentation tool, the corresponding standard segmentation and transcribed segmentation are as follows:

[0063] Standard segmentation: First Branch Product Development Center Software Development Second Product Quality Inspection Group Translated segmentation: First Branch Product Development focuses on software development rather than product development

[0064] Step 103, obtaining standard segmentations that are different from corresponding transcribed segmentations and not in the transcribed text from each standard segmentation to obtain candidate hot words. This step can help determine the hot word file of the speech transcription model based on the candidate hot words.

[0065] Optionally, the process of obtaining the standard participles that are different from the corresponding transcribed participles and are not in the transcribed text to obtain candidate hot words includes:

[0066] For each standard participle, determine whether there is a participle that is the same as the current participle in each transcribed participle.

[0067] Step 104, determining the hot word file of the speech transcription model based on the alternative hot words. This step is based on steps 101 to 103, by calculating the accuracy of the corresponding transcribed text transcribed by the speech transcription model, and judging whether the standard text of the target audio file is a hot word extraction object based on the accuracy, obtaining the segmentation of the standard text and the corresponding transcribed text when the standard text is a hot word extraction object, and then obtaining the standard segmentation that is different from the corresponding transcribed segmentation and different from each transcribed segmentation as the alternative hot word, and further determining the hot word file of the speech transcription model based on the alternative hot word, which can automatically extract the standard vocabulary corresponding to the vocabulary with transcription errors in the standard text based on the transcribed text, thereby eliminating the need for manual extraction of hot words, reducing the cost required for extracting hot words, and improving the efficiency of hot word extraction.

[0068] Optionally, the process of determining the hot word file of the speech transcription model based on the candidate hot words includes: determining the candidate hot words as hot words, and constructing the hot word file using the hot words.

[0069] Optionally, the method for automatic extraction of hot words in speech transcription provided by an embodiment of the present invention also includes: transcribing a sample audio file based on a hot word file through a speech transcription model to obtain a sample transcribed text; determining the word accuracy of the sample text based on the standard text of the sample audio file to obtain a sample word accuracy; and verifying whether the hot word file meets the requirements based on the sample word accuracy.

[0070] In an optional specific embodiment of the present invention, the embodiment of the present invention further provides a speech transcription method, comprising:

[0071] Acquire a hot word file of a speech transcription model based on the method for automatically extracting hot words from speech transcription provided by any embodiment of the present invention; and

[0072] Based on the above hot word file, the target audio file is recognized and transcribed through the above speech transcription model.

[0073] The following further introduces the method for automatically extracting hot words in speech transcription provided by an embodiment of the present invention.

[0074] Optional, such as Figure 2 As shown, Figure 1 Step 103 in the method may include the following steps:

[0075] Step 1031 , using a text difference comparison tool, obtain the segmentations in each standard segmentation that are different from the corresponding transcribed segmentation, and the segmentations in each transcribed segmentation that are different from the corresponding standard segmentation, to obtain differential segmentations.

[0076] In an optional specific embodiment, a text difference comparison tool is used to obtain the difference attributes of each standard segmentation and each transcribed segmentation.

[0077] Optionally, through the text difference comparison tool, for the participles that need to be deleted from the transcribed text, set their difference attributes to "delete"; for the participles that need to be added to the transcribed text, set their difference attributes to "insert"; for the participles that need to remain unchanged, set their difference attributes to "remain unchanged".

[0078] In a specific example, for the following standard participles and transcribed participles:

[0079] Standard participle: First Branch Product Development Center Software Development Department Two Product Quality Inspection Group Transcribed participle: First Branch Product Development Focus Software Development Rather Than Product Between Group Set the difference attribute of "center", "department two" and "quality inspection" to "insert", set the difference attribute of "focus", "rather than" and "between" to "delete", and set the difference attribute of other participles to "remain unchanged".

[0080] Step 1032, determining whether there are differential segmentations in the standard text and the transcribed text, and determining the differential segmentations as candidate hot words when the differential segmentations exist in the standard text but not in the transcribed text.

[0081] Optionally, the process of determining whether there are different segmentations between the standard text and the transcribed text includes:

[0082] Based on the fuzzy matching method, an attempt is made to obtain the same or similar participles as the difference participles in multiple standard participles and multiple transcribed participles; when the same or similar participles as the difference participles in multiple standard participles are obtained, it is judged that the difference participles exist in the standard text, otherwise it is judged that the difference participles do not exist in the standard text; when the same or similar participles as the difference participles in multiple transcribed participles are obtained, it is judged that the difference participles exist in the transcribed text, otherwise it is judged that the difference participles do not exist in the standard text.

[0083] Specifically, it can be determined whether there are different word segmentations between the standard text and the transcribed text based on the prefix and suffix fuzzy matching method.

[0084] Specifically, missing matches can be avoided based on the fuzzy matching method.

[0085] Specifically, based on the exact matching method, an attempt may be made to obtain the same segmentation as the differential segmentation in multiple standard segmentations and multiple transcribed segmentations, and based on the exact matching result, it may be determined whether there is a differential segmentation in the standard text and the transcribed text.

[0086] Specifically, the embodiment of the present invention can more accurately and efficiently determine the candidate hot words in each standard participle by first comparing and determining the difference words, and then judging whether the difference words are only standard participles and not transcribed participles.

[0087] The following further introduces the method for automatically extracting hot words in speech transcription provided by an embodiment of the present invention.

[0088] Optional, such as Figure 3 As shown, Figure 1 Step 104 in may include the following steps:

[0089] Step 1041, determining a word frequency file based on the candidate hot words.

[0090] Optionally, the process of determining a word frequency file based on the candidate hot words includes: creating a word frequency file based on the candidate hot words, or updating the frequency value of the corresponding candidate hot words in the word frequency file.

[0091] Specifically, the word frequency file may be a text file containing words and frequencies, and may be set to two columns, the first column being words and the second column being frequencies.

[0092] Optionally, the above process of updating the frequency value of the corresponding candidate hot word in the word frequency file includes:

[0093] Traverse each candidate hot word to see if it is in the word frequency file. If it is in the word frequency file, increase its word frequency number by 1. If it is not in the word frequency file, append it to the word frequency file and set its word frequency to 1.

[0094] Step 1042: determine a hot word file based on candidate hot words in the word frequency file whose frequency values ​​are greater than a frequency value threshold.

[0095] Specifically, the candidate hot words whose frequency values ​​in the word frequency file are greater than the frequency threshold can be added to the hot word file, or the hot word file can be created by using the candidate hot words whose frequency values ​​in the word frequency file are greater than the frequency threshold.

[0096] Specifically, the frequency threshold may be set based on experience, for example, may be set to 5.

[0097] It is understandable that the appearance of some low-frequency words may be accidental and inappropriate as hot words, which may affect the recognition of other texts by the speech writing model, so it is necessary to screen the candidate hot words. The embodiment of the present invention screens the candidate hot words by the word frequency of each candidate hot word to create a hot word file, which can help improve the accuracy of the speech transcription model in recognizing text, thereby improving the transcription accuracy of the speech transcription model.

[0098] The following further introduces the method for automatically extracting hot words in speech transcription provided by an embodiment of the present invention.

[0099] Optional, such as Figure 4 As shown, the method for automatically extracting hot words from speech transcription provided by the embodiment of the present invention may include the following steps:

[0100] Step 401, determining a performance score of the speech transcription model based on the user feedback word accuracy and / or the user feedback sentence accuracy corresponding to the speech transcription model.

[0101] Specifically, the above-mentioned user-feedback word accuracy and / or user-feedback sentence accuracy may be obtained regularly.

[0102] Specifically, when determining the above performance score based on the user feedback word accuracy and the user feedback sentence accuracy, you can first obtain the word accuracy score corresponding to the user feedback word accuracy and the sentence accuracy score corresponding to the user feedback sentence accuracy, and then perform weighted summation on the word accuracy score and the sentence accuracy score to obtain the above performance score.

[0103] Step 402, when the performance score is less than the score threshold, the word accuracy of the corresponding transcribed text is calculated based on the standard text of the target audio file, and whether the standard text is a hot word extraction object is determined based on the word accuracy.

[0104] Specifically, the score threshold may be set based on user requirements or empirical data.

[0105] Step 403, when the standard text is a hot word extraction object, the word segmentation processing results of the standard text and the corresponding transcribed text are obtained to obtain a plurality of standard word segments and a plurality of transcribed word segments.

[0106] Step 404: Obtain standard segmentations that are different from corresponding transcribed segmentations and are not in the transcribed text to obtain candidate hot words.

[0107] Step 405: determine a hot word file of the speech transcription model based on the candidate hot words.

[0108] The embodiment of the present invention can ensure the effect of speech recognition and transcription of the speech transcription model by evaluating the model performance according to the word accuracy and sentence accuracy fed back by users, and optimizing the hot word file based on the evaluation results.

[0109] Figure 5 1 is a structural diagram of a device for automatically extracting hot words from speech transcription provided by an embodiment of the present invention, and the device is suitable for executing a method for automatically extracting hot words from speech transcription provided by an embodiment of the present invention. Figure 5 As shown, the device may specifically include:

[0110] The extraction object judgment module 501 is used to calculate the word accuracy of the corresponding transcribed text based on the standard text of the target audio file, and judge whether the standard text is a hot word extraction object based on the word accuracy, and the corresponding transcribed text is the text obtained by transcribing the target audio file by the speech transcription model. It is beneficial to obtain a standard text that can be used as a hot word extraction object, and further facilitate hot word extraction based on the standard text that can be used as a hot word extraction object.

[0111] Optionally, the above-mentioned extraction object judgment module 501 can be specifically used to normalize the standard text and the corresponding transcribed text based on the first normalization strategy to obtain a first normalized standard text and a first normalized transcribed text; obtain the correct number of words, the number of deleted words and the number of inserted words in the first normalized transcribed text based on the dynamic normalization algorithm, and calculate the word accuracy based on the correct number of words, the number of deleted words, the number of inserted words and the number of words in the first normalized standard text; and determine whether the standard text is a hot word extraction object based on whether the word accuracy is less than a preset accuracy threshold.

[0112] The word segmentation acquisition module 502 is used to obtain the word segmentation processing results of the standard text and the corresponding transcribed text when the standard text is the hot word extraction object, and obtain multiple standard word segments and multiple transcribed word segments, which can be helpful for obtaining candidate hot words in multiple standard word segments based on multiple transcribed word segments.

[0113] Optionally, the above-mentioned word segmentation acquisition module 502 can be specifically used to normalize the standard text and the corresponding transcribed text based on the second normalization strategy to obtain a second standard standard text and a second standard transcribed text; and use a word segmentation tool to perform word segmentation on the second standard standard text and the second standard transcribed text, and obtain multiple standard word segments and multiple transcribed word segments accordingly.

[0114] The candidate hotword acquisition module 503 is used to acquire the standard segmentation words that are different from the corresponding transcribed segmentation words and are not in the transcribed text to obtain the candidate hotwords, which can be helpful for determining the hotword file of the speech transcription model based on the candidate hotwords.

[0115] Optionally, the above-mentioned alternative hot word acquisition module 503 can be specifically used to use a text difference comparison tool to obtain the participles in each standard participle that are different from the corresponding transcribed participles, and the participles in each transcribed participle that are different from the corresponding standard participles, to obtain differential participles; and to determine whether there are differential participles in the standard text and the transcribed text, and when the differential participles exist in the standard text but not in the transcribed text, determine the differential participles as alternative hot words.

[0116] Optionally, the above-mentioned alternative hot word acquisition module 503 can be specifically used to try to obtain the same or similar participles as the difference participles in multiple standard participles and multiple transcribed participles based on the fuzzy matching method; when the same or similar participles as the difference participles in multiple standard participles are obtained, it is judged that the difference participles exist in the standard text, otherwise it is judged that the difference participles do not exist in the standard text; when the same or similar participles as the difference participles in multiple transcribed participles are obtained, it is judged that the difference participles exist in the transcribed text, otherwise it is judged that the difference participles do not exist in the standard text.

[0117] Hot word file determination module 504 is used to determine the hot word file of the speech transcription model based on the alternative hot words. It can be combined with modules 501 to 503, by calculating the accuracy of the corresponding transcribed text transcribed by the speech transcription model, and judging whether the standard text of the target audio file is a hot word extraction object based on the accuracy, when the standard text is a hot word extraction object, the segmentation of the standard text and the corresponding transcribed text is obtained, and then the standard segmentation that is different from the corresponding transcribed segmentation and different from each transcribed segmentation is obtained as an alternative hot word, and the hot word file of the speech transcription model is further determined based on the alternative hot words, and the standard vocabulary corresponding to the vocabulary with transcription errors in the standard text can be automatically extracted based on the transcribed text, thereby eliminating the need for manual extraction of hot words, reducing the cost required for extracting hot words, and improving the efficiency of hot word extraction.

[0118] Optionally, the hot word file determination module 504 can be specifically used to determine a word frequency file based on the candidate hot words; and

[0119] A hot word file is determined based on candidate hot words in the word frequency file whose frequency values ​​are greater than a frequency value threshold.

[0120] Optionally, the automatic extraction device for hot words in speech transcription provided by an embodiment of the present invention further includes: a performance score determination module, which is used to determine the performance score of the speech transcription model based on the user feedback word accuracy and / or user feedback sentence accuracy corresponding to the speech transcription model.

[0121] Optionally, the aforementioned extraction object determination module 501 can be specifically used to start executing the calculation of the word accuracy of the corresponding transcribed text based on the standard text of the target audio file when the performance score is less than the score threshold.

[0122] Those skilled in the art can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional modules is used as an example for illustration. In actual applications, the above-mentioned functions can be assigned to different functional modules as needed, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. The specific working process of the functional modules described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0123] An embodiment of the present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the method for automatically extracting hot words from speech transcription provided in any of the above embodiments is implemented.

[0124] An embodiment of the present invention further provides a computer-readable medium on which a computer program is stored. When the program is executed by a processor, the method for automatically extracting hot words in speech transcription provided in any of the above embodiments is implemented.

[0125] An embodiment of the present invention further provides a computer program product, including a computer program, which, when executed by a processor, implements the method for automatically extracting hot words in speech transcription as described in any one of the embodiments of the present invention.

[0126] Reference below Figure 6 , which shows a schematic diagram of the structure of a computer system 600 of an electronic device suitable for implementing an embodiment of the present invention. Figure 6 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 invention.

[0127] like Figure 6As shown, the computer system 600 includes a central processing unit (CPU) 601, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 602 or a program loaded from a storage part 608 into a random access memory (RAM) 603. In the RAM 603, various programs and data required for the operation of the system 600 are also stored. The CPU 601, the ROM 602, and the RAM 603 are connected to each other via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0128] The following components are connected to the I / O interface 605: an input section 606 including a keyboard, a mouse, etc.; an output section 607 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 608 including a hard disk, etc.; and a communication section 609 including a network interface card such as a LAN card, a modem, etc. The communication section 609 performs communication processing via a network such as the Internet. A drive 610 is also connected to the I / O interface 605 as needed. A removable medium 611, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 610 as needed, so that a computer program read therefrom is installed into the storage section 608 as needed.

[0129] In particular, according to the embodiments disclosed in the present invention, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments disclosed in the present invention include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 609, and / or installed from the removable medium 611. When the computer program is executed by the central processing unit (CPU) 601, the above-mentioned functions defined in the system of the present invention are executed.

[0130] It should be noted that the computer-readable medium shown in the present invention 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 invention, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, device or device. In the present invention, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable program code. This propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, 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 appropriate medium, including but not limited to: wireless, wire, optical cable, RF, etc., or any suitable combination of the above.

[0131] 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 invention. In this regard, each box in the flow chart or block diagram can represent a module, a program segment, or a part of a code, and the above-mentioned module, program segment, or a part of a code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two 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 box in the block diagram or flow chart, and the combination of the boxes in the block diagram 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.

[0132] The modules and / or units involved in the embodiments of the present invention may be implemented in software or hardware. The modules and / or units described may also be set in a processor, for example, it may be described as: a processor includes an extraction object judgment module, a word segmentation acquisition module, an alternative hot word acquisition module and a model update module. The names of these modules do not, in some cases, constitute limitations on the modules themselves.

[0133] As another aspect, the present invention also provides a computer-readable medium, which may be included in the device described in the above embodiment; or it may exist independently and not be assembled into the device. The above computer-readable medium carries one or more programs, and when the above one or more programs are executed by a device, the device implements: calculating the word accuracy of the corresponding transcribed text based on the standard text of the target audio file, and judging whether the standard text is a hot word extraction object based on the word accuracy, and the corresponding transcribed text is the text obtained by transcribing the target audio file by the speech transcription model; when the standard text is a hot word extraction object, obtaining the word segmentation processing results of the standard text and the corresponding transcribed text to obtain a plurality of standard participles and a plurality of transcribed participles; obtaining the standard participles in each standard participle that are different from the corresponding transcribed participles and are not in the transcribed text to obtain alternative hot words; and determining the hot word file of the speech transcription model based on the alternative hot words.

[0134] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions may occur depending on design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for automatically extracting hot words from speech transcription, characterized in that: include: Calculating the word accuracy of the corresponding transcribed text based on the standard text of the target audio file, and judging whether the standard text is a hot word extraction object based on the word accuracy, the corresponding transcribed text is the text obtained by transcribing the target audio file by the speech transcription model; When the standard text is a hot word extraction object, the word segmentation processing results of the standard text and the corresponding transcribed text are obtained to obtain a plurality of standard word segments and a plurality of transcribed word segments; Obtaining standard participles that are different from corresponding transcribed participles and are not in the transcribed text from among the standard participles to obtain candidate hot words; as well as A hot word file of the speech transcription model is determined based on the candidate hot words.

2. The method for automatically extracting hot words from speech transcription according to claim 1, characterized in that: The step of calculating the word accuracy of the corresponding transcribed text based on the standard text of the target audio file, and judging whether the standard text is a hot word extraction object based on the word accuracy, includes: Based on the first normalization strategy, the standard text and the corresponding transcribed text are normalized to obtain a first normalized standard text and a first normalized transcribed text; Obtaining the number of correct words, deleted words, and inserted words in the first canonical transcription text based on a dynamic canonical algorithm, and calculating the word accuracy rate based on the number of correct words, deleted words, inserted words, and the number of words in the first canonical standard text; and When the character accuracy is less than a preset accuracy threshold, the standard text is determined to be a hot word extraction object.

3. The method for automatically extracting hot words from speech transcription according to claim 1, characterized in that: The word segmentation processing results of the standard text and the corresponding transcribed text are obtained to obtain a plurality of standard word segments and a plurality of transcribed word segments, including: Normalizing the standard text and the corresponding transcribed text based on a second normalization strategy to obtain a second normalized standard text and a second normalized transcribed text; and The second canonical standard text and the second canonical transcribed text are processed by word segmentation using a word segmentation tool, and the plurality of standard word segments and the plurality of transcribed word segments are obtained accordingly.

4. The method for automatically extracting hot words from speech transcription according to claim 1, characterized in that: The step of obtaining standard participles that are different from corresponding transcribed participles and are not in the transcribed text to obtain candidate hot words includes: Using a text difference comparison tool, obtain the segmentations in each standard segmentation that are different from the corresponding transcribed segmentation, and the segmentations in each transcribed segmentation that are different from the corresponding standard segmentation, to obtain differential segmentations; and It is determined whether the difference participle exists in the standard text and the transcribed text, and when the difference participle exists in the standard text but not in the transcribed text, the difference participle is determined as the candidate hot word.

5. According to the method for automatic extraction of hot words in speech transcription according to claim 4, it is characterized in that: The determining whether the difference participle exists in the standard text and the transcribed text comprises: Attempting to obtain a word that is identical or similar to the difference word from among the plurality of standard word segments and the plurality of transcribed word segments based on a fuzzy matching method; When a participle that is identical or similar to the difference participle is obtained from the plurality of standard participles, it is determined that the difference participle exists in the standard text; otherwise, it is determined that the difference participle does not exist in the standard text; When a participle that is identical or similar to the difference participle in the multiple transcribed participles is obtained, it is determined that the difference participle exists in the transcribed text; otherwise, it is determined that the difference participle does not exist in the standard text.

6. The method for automatically extracting hot words from speech transcription according to claim 1, characterized in that: The step of determining the hot word file of the speech transcription model based on the candidate hot words includes: Determine a word frequency file based on the candidate hot words; and The hot word file is determined based on the candidate hot words in the word frequency file whose frequency values ​​are greater than a frequency value threshold.

7. The method for automatically extracting hot words from speech transcription according to claim 1, characterized in that: Before calculating the word accuracy of the corresponding transcribed text based on the standard text of the target audio file, the method for automatically extracting hot words from speech transcription further includes: Determining a performance score of the speech transcription model based on a user feedback word accuracy rate and a user feedback sentence accuracy rate corresponding to the speech transcription model; The step of calculating the word accuracy of the corresponding transcribed text based on the standard text of the target audio file comprises: When the performance score is less than the score threshold, the step of calculating the word accuracy of the corresponding transcribed text based on the standard text of the target audio file is started.

8. A device for automatically extracting hot words from speech transcription, characterized in that: include: An extraction object judgment module is used to calculate the word accuracy of the corresponding transcribed text based on the standard text of the target audio file, and judge whether the standard text is a hot word extraction object based on the word accuracy, and the corresponding transcribed text is the text obtained by transcribing the target audio file by the speech transcription model; A word segmentation acquisition module, used for obtaining the word segmentation processing results of the standard text and the corresponding transcribed text to obtain a plurality of standard word segments and a plurality of transcribed word segments when the standard text is a hot word extraction object; The candidate hot word acquisition module is used to acquire the standard participles that are different from the corresponding transcribed participles and are not in the transcribed text to obtain the candidate hot words; as well as A hot word file determination module is used to determine the hot word file of the speech transcription model based on the candidate hot words.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the method for automatically extracting hot words from speech transcription as described in any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method for automatically extracting hot words from speech transcription as described in any one of claims 1 to 7 is implemented.

Images