Hot word determination method and device, storage medium and electronic device

Abstract

The present disclosure provides a hot word determination method, a hot word determination device, a storage medium and an electronic device, and relates to the field of speech recognition technology. The hot word determination method includes: obtaining current frame speech data, converting the current frame speech data into current frame text; determining candidate hot words using the current frame text and adjacent frame text corresponding to the current frame; determining at least one feature expression of the candidate hot word, and clustering the candidate hot words using at least one feature expression of the candidate hot word, and determining the hot word according to the clustering result. The present disclosure can improve the efficiency of generating hot words.

Description

Background Art

[0002] With the development of artificial intelligence technology, speech recognition technology has been applied to various industries, enabling the industrial development of various industries, and has been widely used in fields such as customer service, finance, and medical care. For speech recognition tasks in specific fields, it is crucial to successfully identify certain proprietary words. For example, in the customer service scenario of telecom operators, there are proprietary words such as "traffic package" and "King Card". The recognition of such words often plays an important role in downstream tasks such as customer intent analysis. We call such words hot words. However, in actual applications, it is found that the recognition effect of hot words is often biased due to the lack of training data containing the vocabulary and the frequent iteration of new hot words. Therefore, for R&D personnel, how to timely determine and generate hot words is an important issue to improve the accuracy of speech recognition of domain hot words.

[0003] At present, a lot of manpower is needed to mark hot words, which is inefficient and has poor accuracy.

[0004] It should be noted that the information disclosed in the above background technology section is only used to enhance the understanding of the background of the present disclosure, and therefore may include information that does not constitute the prior art known to ordinary technicians in the field. Summary of the invention

[0005] The purpose of the present disclosure is to provide a hot word determination method, a hot word determination device, a storage medium and an electronic device, thereby overcoming the problem of low efficiency and inaccuracy of manual hot word annotation at least to a certain extent.

[0006] According to the first aspect of the present disclosure, a hot word determination method is provided, including: obtaining current frame voice data, converting the current frame voice data into current frame text; determining candidate hot words using the current frame text and adjacent frame texts corresponding to the current frame; determining at least one feature expression of the candidate hot words, and clustering the candidate hot words using at least one feature expression of the candidate hot words, and determining the hot words according to the clustering results.

[0007] Optionally, determining candidate hot words using the current frame text and the adjacent frame text corresponding to the current frame includes: determining the original vocabulary using the current frame text and the adjacent frame text corresponding to the current frame; determining the coagulation degree and degree of freedom of the original vocabulary, the coagulation degree represents the closeness between characters in the original vocabulary, and the degree of freedom represents the degree to which the original vocabulary can be used independently and freely; if the coagulation degree of the original vocabulary is greater than the coagulation degree threshold and the degree of freedom of the original vocabulary is greater than the degree of freedom threshold, then determining the original vocabulary as a candidate hot word.

[0008] Optionally, determining at least one characteristic expression of a candidate hot word includes: determining a first characteristic expression of the candidate hot word using a text database and an existing hot word list in the field to which the hot word belongs; determining a second characteristic expression of the candidate hot word using a common vocabulary list; and performing word vector conversion on the candidate hot word to obtain a third characteristic expression of the candidate hot word.

[0009] Optionally, determining the first characteristic expression of a candidate hot word using a text database and an existing hot word table in the field to which the hot word belongs includes: determining frequency information in which the candidate hot word and an existing hot word in the existing hot word table simultaneously appear in the text database; and determining the first characteristic expression of the candidate hot word based on the frequency information.

[0010] Optionally, candidate hot words are clustered using at least one feature expression of the candidate hot words, and the hot words are determined based on the clustering results, including: initializing the class center; determining a first distance between the first feature expression and the class center; determining a second distance between the second feature expression and the class center; determining a third distance between the third feature expression and the class center; calculating the average distance of the first distance, the second distance, and the third distance, and iteratively updating the class center based on the average distance to obtain a clustering result; and determining the hot words based on the clustering results.

[0011] Optionally, the hot word determination method further includes: after determining the hot word, determining an incentive value corresponding to the hot word.

[0012] Optionally, determining the incentive value corresponding to the hot word includes: calculating the optimal solution of a speech recognition model with fixed parameters on a test set according to a current hot word table to determine the incentive value corresponding to the hot word.

[0013] According to a second aspect of the present disclosure, a hot word determination device is provided, including: a data conversion module, used to obtain current frame voice data and convert the current frame voice data into current frame text; a candidate hot word determination module, used to determine candidate hot words using the current frame text and adjacent frame texts corresponding to the current frame; a hot word determination module, used to determine at least one feature expression of the candidate hot word, and cluster the candidate hot words using at least one feature expression of the candidate hot word, and determine the hot word according to the clustering result.

[0014] Optionally, the candidate hot word determination module can be configured to perform: determine the original vocabulary using the current frame text and the adjacent frame text corresponding to the current frame; determine the coagulation degree and degree of freedom of the original vocabulary, the coagulation degree represents the closeness between the characters in the original vocabulary, and the degree of freedom represents the degree to which the original vocabulary can be used independently and freely; if the coagulation degree of the original vocabulary is greater than the coagulation degree threshold and the degree of freedom of the original vocabulary is greater than the degree of freedom threshold, then determine the original vocabulary as a candidate hot word.

[0015] Optionally, the process of the hot word determination module determining at least one characteristic expression of a candidate hot word can be configured to perform: determining a first characteristic expression of the candidate hot word using a text database of the field to which the hot word belongs and an existing hot word table; determining a second characteristic expression of the candidate hot word using a common vocabulary table; and performing word vector conversion on the candidate hot word to obtain a third characteristic expression of the candidate hot word.

[0016] Optionally, the process of determining the first characteristic expression of the candidate hot word by the hot word determination module can be configured to execute: determining frequency information of the candidate hot word and the existing hot words in the existing hot word table appearing simultaneously in the text database; and determining the first characteristic expression of the candidate hot word according to the frequency information.

[0017] Optionally, the hot word determination module can be configured to perform: initializing the class center; determining the first distance between the first feature expression and the class center; determining the second distance between the second feature expression and the class center; determining the third distance between the third feature expression and the class center; calculating the average distance of the first distance, the second distance, and the third distance, and iteratively updating the class center based on the average distance to obtain clustering results; determining the hot words based on the clustering results.

[0018] Optionally, the hot word determination device may further include an incentive value determination module.

[0019] Specifically, the incentive value determination module may be configured to execute: after determining the hot word, determine the incentive value corresponding to the hot word.

[0020] Optionally, the incentive value determination module may be configured to execute: calculating an optimal solution of a speech recognition model with fixed parameters on a test set according to a current hot word list to determine an incentive value corresponding to the hot word.

[0021] According to a third aspect of the present disclosure, a storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the above-mentioned hot word determination method is implemented.

[0022] According to a fourth aspect of the present disclosure, an electronic device is provided, comprising: a processor; and a memory for storing executable instructions of the processor; the processor is configured to implement the above-mentioned hot word determination method by executing the executable instructions.

[0023] In the technical solutions provided by some embodiments of the present disclosure, the current frame voice data is converted into the current frame text, the current frame text and the adjacent frame text corresponding to the current frame are used to determine the candidate hot words, at least one feature expression of the candidate hot words is determined, and the candidate hot words are clustered using at least one feature expression of the candidate hot words, and the hot words are determined according to the clustering results. The process of determining hot words in the present disclosure does not require human participation, which improves the efficiency of hot word generation, and in view of the lack of influence of human subjective factors, the accuracy of hot word determination is also improved, thereby improving the accuracy of speech recognition.

[0024] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] The drawings herein are incorporated into the specification and constitute a part of the specification, showing embodiments consistent with the present disclosure, and together with the specification, are used to explain the principles of the present disclosure. Obviously, the drawings described below are only some embodiments of the present disclosure, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. In the drawings:

[0026] Figure 1 A flowchart of a hot word determination method according to an exemplary embodiment of the present disclosure is schematically shown;

[0027] Figure 2 The flowchart of determining candidate hot words in an embodiment of the present disclosure is schematically shown;

[0028] Figure 3 A flowchart schematically shows the entire process of hot word recognition in an embodiment of the present disclosure;

[0029] Figure 4 A block diagram schematically shows a hot word determination device according to an exemplary embodiment of the present disclosure;

[0030] Figure 5 A block diagram schematically shows a hot word determination device according to another exemplary embodiment of the present disclosure;

[0031] Figure 6 A block diagram of an electronic device according to an exemplary embodiment of the present disclosure is schematically shown. DETAILED DESCRIPTION

[0032] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in a variety of forms and should not be construed as being limited to the examples set forth herein; on the contrary, these embodiments are provided so that the present disclosure will be more comprehensive and complete, and the concepts of the example embodiments are fully conveyed to those skilled in the art. The described features, structures, or characteristics may be combined in one or more embodiments in any suitable manner. In the following description, many specific details are provided to provide a full understanding of the embodiments of the present disclosure. However, those skilled in the art will appreciate that the technical solutions of the present disclosure may be practiced while omitting one or more of the specific details, or other methods, components, devices, steps, etc. may be adopted. In other cases, known technical solutions are not shown or described in detail to avoid obscuring various aspects of the present disclosure.

[0033] In addition, the accompanying drawings are only schematic illustrations of the present disclosure and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor devices and / or microcontroller devices.

[0034] The flowcharts shown in the accompanying drawings are only exemplary and do not necessarily include all the steps. For example, some steps may be decomposed, while some steps may be combined or partially combined, so the actual execution order may change according to the actual situation. In addition, all the terms "first", "second", "third", etc. below are only for the purpose of distinction and should not be used as limitations of the present disclosure.

[0035] At present, speech recognition technology is applied to various industries and enables various smart services. It is widely used in various fields such as operators, finance, medical care, e-commerce, etc. The discovery of hot words in the field of speech recognition and the correction of recognition results have always been a pain point in the industry. The successful recognition of certain hot words often has an important impact on downstream tasks. For example, the successful recognition of package hot words in telecommunications services can affect whether the service is successfully handled, and the successful recognition of disease hot words in medical scenarios may have an important impact on the diagnosis or further determination of the disease. The scheme of the embodiment of the present disclosure can automatically determine the hot words in speech recognition and generate a hot word table and its corresponding incentive value, which can be applied to various application scenarios involving speech recognition tasks.

[0036] Specifically, the disclosed embodiment is oriented to the speech recognition task in a specified field, and proposes a clustering-based adaptive generation scheme for hot words, determines and mines field hot words, adaptively determines the incentive value (weight) of field hot words, and continuously iterates and updates the hot word set. For example, the generated hot word set is used to assist the hot word speech recognition in the operator field, thereby improving the accuracy of the hot word speech recognition in the operator field.

[0037] The hot word determination scheme of the embodiment of the present disclosure can be implemented by an electronic device, that is, the electronic device can execute each step of the following hot word determination method, and the following hot word determination device can be configured in the electronic device. In addition, the present disclosure does not limit the type of electronic device, which may include but is not limited to smart phones, tablet computers, smart wearable devices, personal computers, servers, etc.

[0038] Figure 1 The flowchart of the hot word determination method of the exemplary embodiment of the present disclosure is schematically shown. Figure 1 , the hot word determination method may include the following steps:

[0039] S12. Obtain the current frame voice data, and convert the current frame voice data into the current frame text.

[0040] In an exemplary embodiment of the present disclosure, voice data may be data acquired by an electronic device through a sound receiving device (such as a microphone), or may be data contained in a voice file acquired by the electronic device from other devices, or may be data generated by voice generation software. The present disclosure does not impose any restrictions on the source, content, etc. of the voice data.

[0041] The electronic device can perform frame processing on the voice data to obtain the current frame voice data. The present disclosure does not limit the frame division method.

[0042] Next, the electronic device may convert the current frame voice data into the current frame text.

[0043] Specifically, the electronic device can convert the current frame voice data into the current frame text according to the dictionary, acoustic model and speech model. Among them, the acoustic model can be a knowledge representation of the differences in acoustics, phonetics, environmental variables, speaker gender, accent, etc., and the language model is a knowledge representation of a group of word sequences. The present disclosure does not limit the model structure, source, etc. of the acoustic model and the language model.

[0044] The electronic device can extract features from the current frame speech data, and use the above-mentioned acoustic model and language model to jointly encode the extracted features to obtain the current frame text.

[0045] It should be noted that when multiple texts are generated by jointly decoding the acoustic model and the language model, the first m of these multiple texts are determined as the current frame texts mentioned in the implementation mode of the present disclosure. That is, the number of current frame texts mentioned in the present disclosure can be multiple, for example, m, and m can be a positive integer greater than or equal to 2, and the present disclosure does not impose any restrictions on this.

[0046] S14. Determine candidate hot words using the current frame text and the adjacent frame text corresponding to the current frame.

[0047] In an exemplary embodiment of the present disclosure, the adjacent frames corresponding to the current frame may be the first n frames of the current frame. Based on the characteristics of Chinese, the granularity of frame recognition is usually the word granularity. In this case, n may be 3. However, n may also be 2, 4, 5, etc.

[0048] In addition, the adjacent frame corresponding to the current frame may also be a set of the previous a frame and the following b frame of the current frame. Alternatively, the adjacent frame corresponding to the current frame may also be the following c frame of the current frame. The present disclosure does not limit the values ​​of a, b, and c.

[0049] The adjacent frame text is the text obtained by converting the adjacent frame voice data. Similarly, the electronic device can convert the adjacent frame voice data into the adjacent frame text according to the dictionary, the acoustic model and the voice model. It should be noted that the adjacent frame text can be generated when the adjacent frame voice data is processed, rather than the processing process of the current frame voice data in this embodiment. In other words, the adjacent frame text can be determined in advance, and in step S14, the electronic device can directly obtain the adjacent frame text.

[0050] Regarding the process of determining candidate hot words, first, the electronic device can determine the original vocabulary using the current frame text and the adjacent frame text corresponding to the current frame. Specifically, at least one original vocabulary can be determined, or it can be called determining an original vocabulary set. The original vocabulary set can be a vocabulary set formed after eliminating non-Chinese vocabulary.

[0051] Taking the current frame text as the kth frame text as an example, the adjacent frame text may include the k-1th frame text, the k-2th frame text, and the k-3th frame text. In this case, the generated original vocabulary set composed of original vocabulary may be [“C k-1 C k ”,“C k-2 C k-1 C k ”,“C k-3 C k-2 C k-1 C k ”], where C k , C k-1 , C k-2 , C k-3They represent the k-th frame text, the k-1-th frame text, the k-2-th frame text and the k-3-th frame text respectively.

[0052] Next, the electronic device may filter the original vocabulary by word formation parameters to determine candidate hot words from the original vocabulary.

[0053] The word formation parameter may represent the word formation probability of the original vocabulary, and the word formation parameter may include cohesion and / or freedom. The cohesion may represent the closeness between characters in the original vocabulary, and the freedom may represent the degree to which the original vocabulary can be used independently and freely.

[0054] Taking the original vocabulary composed of three recognized texts as an example, the coagulation degree and the degree of freedom are explained. The coagulation degree C and the degree of freedom F can be expressed and calculated by the following formulas:

[0055]

[0056] F(C k-2 C k-1 C k )=min{H(C k-2 C k-1 ),H(C k-1 C k )}

[0057]

[0058] Among them, P, P i is the probability value, n is the number of characters, and H is the information entropy function.

[0059] When the solidification degree and the degree of freedom of the original word are determined, the electronic device can compare the calculated solidification degree with the solidification degree threshold and the calculated degree of freedom with the degree of freedom threshold respectively. If the solidification degree of the original word is greater than the solidification degree threshold and the degree of freedom of the original word is greater than the degree of freedom threshold, the original word can be determined as a candidate hot word. The present disclosure does not limit the specific values ​​of the solidification degree threshold and the degree of freedom threshold.

[0060] It should be noted that in other embodiments of the present disclosure, the electronic device may use the degree of solidification or the degree of freedom to determine the candidate hot words. That is, after comprehensively considering the algorithm accuracy and the device processing capability, the electronic device may use only one of the degree of solidification and the degree of freedom to determine the candidate hot words.

[0061] Figure 2 The flowchart of determining candidate hot words in an embodiment of the present disclosure is schematically shown. Figure 2First, the electronic device can combine the kth frame text, the k-1th frame text, the k-2th frame text, and the k-3th frame text to obtain a first original vocabulary set; next, the electronic device can remove non-Chinese vocabulary from the first original vocabulary set to obtain a second original vocabulary set; then, the electronic device can use the above-mentioned solidification degree and / or degree of freedom to filter the vocabulary in the second original vocabulary set to determine a candidate hot word set.

[0062] S16. Determine at least one characteristic expression of the candidate hot word, cluster the candidate hot words using the at least one characteristic expression of the candidate hot word, and determine the hot word according to the clustering result.

[0063] On the one hand, the electronic device can use the text database of the field to which the hot word belongs and the existing hot word table to determine the first characteristic expression of the candidate hot word. Taking the field to which the hot word belongs as the telecommunications operator field as an example, a text database of the telecommunications operator field can be established based on various data acquisition channels, wherein the data acquisition channels include but are not limited to Internet crawlers, voice annotated text data, downstream natural language processing tasks, etc.

[0064] Specifically, the electronic device can determine the frequency information of the candidate hot words and the existing hot words in the existing hot word table appearing in the text database at the same time, and determine the first feature expression of the candidate hot words according to the frequency information. For example, based on the text database in the field of telecommunications operators, the co-occurrence frequency matrix of the candidate hot words and the existing hot word table is calculated, and for each candidate hot word, a co-occurrence frequency row vector with the existing hot words is generated as the first feature expression of each candidate hot word.

[0065] On the other hand, the electronic device can use the general vocabulary to determine the second feature expression of the candidate hot word. Specifically, the electronic device can calculate the sound similarity between the candidate hot word and the general vocabulary based on the training data set of the speech recognition model adopted by the embodiment of the present disclosure, and generate a second feature expression with the same size as the general vocabulary. The sound similarity can be determined by accumulating the number of matches between the candidate hot word and the homophone in the general vocabulary.

[0066] On the other hand, the electronic device can perform word vector conversion on the candidate hot word to obtain the third feature expression of the candidate hot word. The word vector conversion can be implemented based on the Word2vec model, and the present disclosure does not limit this.

[0067] The disclosed embodiments may cluster candidate hot words using at least one of the first feature expression, the second feature expression, and the third feature expression.

[0068] In an embodiment of clustering candidate hot words by using the first feature expression, the second feature expression and the third feature expression of the candidate hot words, first, the electronic device may initialize the cluster center. Specifically, the cluster center may be initialized in a random manner.

[0069] Next, the electronic device may determine a first distance between the first feature expression and the class center, a second distance between the second feature expression and the class center, and a third distance between the third feature expression and the class center, wherein the first distance, the second distance, and the third distance may be calculated by calculating the cosine distance.

[0070] The electronic device may calculate an average distance of the first distance, the second distance, and the third distance as the distance between samples of the candidate hot word.

[0071] Subsequently, a clustering algorithm (eg, kmeans clustering algorithm) may be executed to iteratively update the cluster center in combination with the above average distance until the clustering model converges to obtain a clustering result, thereby determining hot words.

[0072] It should be noted that the above process of determining hot words is explained by taking three feature expressions as an example. In other embodiments of the present disclosure, the electronic device can also execute a clustering algorithm based on one or two of the first feature expression, the second feature expression and the third feature expression of the candidate hot words to determine the hot words.

[0073] After determining the hot word, the electronic device can also determine the incentive value corresponding to the hot word. Specifically, the determined hot word is added to the hot word table, and the optimal solution of the speech recognition model with fixed parameters on the training set is calculated according to the current hot word table to determine the incentive value corresponding to the hot word.

[0074] For the determined hot words, a seed incentive value is generated, the hot word table is updated, and the current frame text is decoded using the following formula:

[0075] P(c)=argmaxlogP(c|x)+λlog(P lm (c)+bi)

[0076] Among them, c is the decoded character of the current frame text, P(c) is the probability of the character, and b i is the incentive value of the hot word. If it is not a hot word, then b i is 0.

[0077] Calculate the optimal solution of the frozen speech recognition model on the test set according to the currently determined hot word list, and iteratively update the incentive value b i , and finally the optimal hot word list is obtained. Among them, the frozen speech recognition model is the speech recognition model with fixed parameters.

[0078] Figure 3 The flowchart schematically shows the entire process of hot word recognition according to an embodiment of the present disclosure.

[0079] In step S302, the electronic device may perform frame processing on the voice data.

[0080] In step S304, the electronic device may perform feature extraction on the framed voice data.

[0081] In step S306, the electronic device may jointly decode the extracted features in combination with the acoustic model and the language model to determine the recognized text of the speech data. Next, the electronic device may screen out candidate hot words.

[0082] In step S308, the electronic device determines whether the candidate hot word is in the hot word list. If it is in the hot word list, the optimized decoding process of step S314 is executed to obtain the recognition result; if it is not in the hot word list, step S310 is executed.

[0083] In step S310, the electronic device determines whether the candidate hot word is a generative hot word. If it is a generative hot word, the hot word table can be updated; if it is not a generative hot word, the decoding process of step S312 is executed to obtain the recognition result.

[0084] The disclosed embodiment adopts a method for selecting and clustering candidate hot words, and generates three feature expressions for the candidate hot words. It can automatically judge the hot words existing in the speech recognition process and determine the optimal incentive value, thereby ensuring the better effect of judging the domain hot words, improving the accuracy of hot word speech recognition and avoiding negative impact on general vocabulary.

[0085] It should be noted that although the steps of the method in the present disclosure are described in a specific order in the drawings, this does not require or imply that the steps must be performed in this specific order, or that all the steps shown must be performed to achieve the desired results. Additionally or alternatively, some steps may be omitted, multiple steps may be combined into one step, and / or one step may be decomposed into multiple steps, etc.

[0086] Furthermore, this example implementation also provides a hot word determination device.

[0087] Figure 4 FIG. 1 schematically shows a block diagram of a hot word determination device according to an exemplary embodiment of the present disclosure. Figure 4 The hot word determining device 4 according to an exemplary embodiment of the present disclosure may include a data conversion module 41 , a candidate hot word determining module 43 , and a hot word determining module 45 .

[0088] Specifically, the data conversion module 41 can be used to obtain the current frame voice data and convert the current frame voice data into the current frame text; the candidate hot word determination module 43 can be used to determine the candidate hot words using the current frame text and the adjacent frame text corresponding to the current frame; the hot word determination module 45 can be used to determine at least one feature expression of the candidate hot words, and cluster the candidate hot words using at least one feature expression of the candidate hot words, and determine the hot words based on the clustering results.

[0089] According to an exemplary embodiment of the present disclosure, the candidate hot word determination module 43 can be configured to perform: determining the original vocabulary using the current frame text and the adjacent frame text corresponding to the current frame; determining the coagulation degree and degree of freedom of the original vocabulary, the coagulation degree represents the closeness between the characters in the original vocabulary, and the degree of freedom represents the degree to which the original vocabulary can be used independently and freely; if the coagulation degree of the original vocabulary is greater than the coagulation degree threshold and the degree of freedom of the original vocabulary is greater than the degree of freedom threshold, then determining the original vocabulary as a candidate hot word.

[0090] According to an exemplary embodiment of the present disclosure, the process of the hot word determination module 45 determining at least one characteristic expression of a candidate hot word can be configured to perform: determining a first characteristic expression of the candidate hot word using a text database of the field to which the hot word belongs and an existing hot word table; determining a second characteristic expression of the candidate hot word using a common vocabulary table; and performing word vector conversion on the candidate hot word to obtain a third characteristic expression of the candidate hot word.

[0091] According to an exemplary embodiment of the present disclosure, the process of the hot word determination module 45 determining the first characteristic expression of a candidate hot word can be configured to execute: determining the frequency information of the candidate hot word and the existing hot words in the existing hot word table appearing simultaneously in the text database; determining the first characteristic expression of the candidate hot word according to the frequency information.

[0092] According to an exemplary embodiment of the present disclosure, the hot word determination module 45 can be configured to perform: initializing the class center; determining the first distance between the first feature expression and the class center; determining the second distance between the second feature expression and the class center; determining the third distance between the third feature expression and the class center; calculating the average distance of the first distance, the second distance, and the third distance, and iteratively updating the class center in combination with the average distance to obtain clustering results; and determining the hot words based on the clustering results.

[0093] According to an exemplary embodiment of the present disclosure, referring to Figure 5 Compared with the hot word determining device 4 , the hot word determining device 5 may further include an incentive value determining module 51 .

[0094] Specifically, the incentive value determination module 51 may be configured to execute: after determining a hot word, determining an incentive value corresponding to the hot word.

[0095] According to an exemplary embodiment of the present disclosure, the incentive value determination module 51 may be configured to execute: calculating the optimal solution of the speech recognition model with fixed parameters on the test set according to the current hot word list to determine the incentive value corresponding to the hot word.

[0096] Since the functional modules of the hot word determination device in the embodiment of the present disclosure are the same as those in the above-mentioned method implementation, they will not be described in detail here.

[0097] In an exemplary embodiment of the present disclosure, a computer-readable storage medium is also provided, on which a program product capable of implementing the above method of the present specification is stored. In some possible implementations, various aspects of the present disclosure may also be implemented in the form of a program product, which includes a program code, and when the program product is run on a terminal device, the program code is used to enable the terminal device to execute the steps according to various exemplary implementations of the present disclosure described in the above "Exemplary Method" section of the present specification.

[0098] The program product for implementing the above method according to the embodiment of the present disclosure can adopt a portable compact disk read-only memory (CD-ROM) and include program code, and can be run on a terminal device, such as a personal computer. However, the program product of the present disclosure is not limited thereto, and in this document, a readable storage medium can be any tangible medium containing or storing a program, which can be used by or in combination with an instruction execution system, an apparatus or a device.

[0099] The program product may use any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. The 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 (non-exhaustive list) of readable storage media include: an electrical connection with one or more wires, a portable 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 disk, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0100] Computer readable signal media may include data signals propagated in baseband or as part of a carrier wave, in which readable program code is carried. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Readable signal media may also be any readable medium other than a readable storage medium, which may send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0101] The program code embodied on the readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wired, optical cable, RF, etc., or any suitable combination of the foregoing.

[0102] Program code for performing the operations of the present disclosure may be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java, C++, etc., and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user computing device, partially on the user device, as a separate software package, partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server. In cases involving a remote computing device, the remote computing device may be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computing device (e.g., through the Internet using an Internet service provider).

[0103] In an exemplary embodiment of the present disclosure, an electronic device capable of implementing the above method is also provided.

[0104] Those skilled in the art will appreciate that various aspects of the present disclosure may be implemented as systems, methods or program products. Therefore, various aspects of the present disclosure may be specifically implemented in the following forms, namely: complete hardware implementation, complete software implementation (including firmware, microcode, etc.), or a combination of hardware and software, which may be collectively referred to herein as "circuits", "modules" or "systems".

[0105] Refer to the following Figure 6 The electronic device 600 according to this embodiment of the present disclosure is described. Figure 6 The electronic device 600 shown is merely an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0106] like Figure 6 As shown, the electronic device 600 is in the form of a general computing device. The components of the electronic device 600 may include but are not limited to: at least one processing unit 610, at least one storage unit 620, a bus 630 connecting different system components (including the storage unit 620 and the processing unit 610), and a display unit 640.

[0107] The storage unit stores program codes, which can be executed by the processing unit 610, so that the processing unit 610 executes the steps according to various exemplary embodiments of the present disclosure described in the above “Exemplary Method” section of this specification. For example, the processing unit 610 can execute various steps of the hot word determination method of the embodiment of the present disclosure.

[0108] The storage unit 620 may include a readable medium in the form of a volatile storage unit, such as a random access memory unit (RAM) 6201 and / or a cache memory unit 6202 , and may further include a read-only memory unit (ROM) 6203 .

[0109] The storage unit 620 may also include a program / utility 6204 having a set (at least one) of program modules 6205, such program modules 6205 including but not limited to: an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include an implementation of a network environment.

[0110] Bus 630 may represent one or more of several types of bus structures, including a memory unit bus or memory unit controller, a peripheral bus, an accelerated graphics port, a processing unit, or a local bus using any of a variety of bus architectures.

[0111] The electronic device 600 may also communicate with one or more external devices 700 (e.g., keyboards, pointing devices, Bluetooth devices, etc.), may also communicate with one or more devices that enable a user to interact with the electronic device 600, and / or communicate with any device that enables the electronic device 600 to communicate with one or more other computing devices (e.g., routers, modems, etc.). Such communication may be performed via an input / output (I / O) interface 650. Furthermore, the electronic device 600 may also communicate with one or more networks (e.g., local area networks (LANs), wide area networks (WANs), and / or public networks, such as the Internet) via a network adapter 660. As shown, the network adapter 660 communicates with other modules of the electronic device 600 via a bus 630. It should be understood that, although not shown in the figure, other hardware and / or software modules may be used in conjunction with the electronic device 600, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems, etc.

[0112] Through the description of the above implementation, it is easy for those skilled in the art to understand that the example implementation described here can be implemented by software, or by software combined with necessary hardware. Therefore, the technical solution according to the implementation of the present disclosure can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, including several instructions to enable a computing device (which can be a personal computer, a server, a terminal device, or a network device, etc.) to execute the method according to the implementation of the present disclosure.

[0113] In addition, the above-mentioned figures are only schematic illustrations of the processes included in the method according to the exemplary embodiments of the present disclosure, and are not intended to be limiting. It is easy to understand that the processes shown in the above-mentioned figures do not indicate or limit the time sequence of these processes. In addition, it is also easy to understand that these processes can be performed synchronously or asynchronously, for example, in multiple modules.

[0114] It should be noted that, although several modules or units of the device for action execution are mentioned in the above detailed description, this division is not mandatory. In fact, according to the embodiments of the present disclosure, the features and functions of two or more modules or units described above can be embodied in one module or unit. On the contrary, the features and functions of one module or unit described above can be further divided into multiple modules or units to be embodied.

[0115] Those skilled in the art will readily appreciate other embodiments of the present disclosure after considering the specification and practicing what is disclosed herein. This application is intended to cover any variations, uses, or adaptations of the present disclosure that follow the general principles of the present disclosure and include common knowledge or customary technical means in the art that are not disclosed in the present disclosure. The specification and embodiments are to be considered merely as exemplary, and the true scope and spirit of the present disclosure are indicated by the claims.

[0116] It should be understood that the present disclosure is not limited to the exact structures that have been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.

Claims

1. A method for determining hot words, characterized in that: include: Acquire current frame voice data, and convert the current frame voice data into current frame text; Determine candidate hot words using the current frame text and adjacent frame texts corresponding to the current frame; Determine frequency information of the candidate hot word and the existing hot words in the existing hot word table appearing simultaneously in the text database, and determine a first characteristic expression of the candidate hot word according to the frequency information; Determine a second characteristic expression of the candidate hot word by using a common vocabulary; Performing word vector conversion on the candidate hot word to obtain a third feature expression of the candidate hot word; Initialize class center; determining a first distance between the first feature expression and the class center; determining a second distance between the second feature expression and the class center; determining a third distance between the third feature expression and the class center; Calculating an average distance of the first distance, the second distance, and the third distance, and iteratively updating the cluster center based on the average distance to obtain a clustering result; Hot words are determined according to the clustering results.

2. The method for determining hot words according to claim 1, characterized in that: Determining candidate hot words using the current frame text and adjacent frame text corresponding to the current frame includes: Determine the original vocabulary using the current frame text and adjacent frame texts corresponding to the current frame; Determining the cohesion and freedom of the original vocabulary, wherein the cohesion represents the closeness between characters in the original vocabulary, and the freedom represents the degree to which the original vocabulary can be used independently and freely; If the solidification degree of the original word is greater than a solidification degree threshold and the degree of freedom of the original word is greater than a degree of freedom threshold, the original word is determined as the candidate hot word.

3. The hot word determination method according to claim 1 or 2, characterized in that: The hot word determination method also includes: After the hot word is determined, the incentive value corresponding to the hot word is determined.

4. The method for determining hot words according to claim 3, characterized in that: Determining the incentive value corresponding to the hot word includes: The optimal solution of the speech recognition model with fixed parameters on the test set is calculated according to the current hot word list to determine the incentive value corresponding to the hot word.

5. A hot word determination device, characterized in that: include: A data conversion module, used for acquiring the current frame voice data and converting the current frame voice data into the current frame text; A candidate hot word determination module, used to determine a candidate hot word by using the current frame text and adjacent frame text corresponding to the current frame; A hot word determination module, used to determine the frequency information of the candidate hot word and the existing hot words in the existing hot word table appearing simultaneously in the text database, and determine the first characteristic expression of the candidate hot word according to the frequency information; Determine a second characteristic expression of the candidate hot word by using a common vocabulary; Perform word vector conversion on the candidate hot word to obtain the third feature expression of the candidate hot word; initialize the class center; determine the first distance between the first feature expression and the class center; determine the second distance between the second feature expression and the class center; determine the third distance between the third feature expression and the class center; calculate the average distance of the first distance, the second distance, and the third distance, and iteratively update the class center in combination with the average distance to obtain a clustering result; determine the hot word according to the clustering result.

6. A storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the hot word determination method according to any one of claims 1 to 4 is implemented.

7. An electronic device, characterized in that: include: processor; as well as A memory, configured to store executable instructions of the processor; The processor is configured to implement the hot word determination method according to any one of claims 1 to 4 by executing the executable instructions.

Images