CN110827803A - Method, device and equipment for constructing dialect pronunciation dictionary and readable storage medium - Google Patents

Abstract

The application discloses a method, a device, equipment and a readable storage medium for constructing a dialect pronunciation dictionary, wherein the method for constructing the dialect pronunciation dictionary comprises the steps of inputting a plurality of dialect pronunciation data of vocabularies into a voice recognition device; receiving a candidate phoneme sequence group output by a voice recognition device based on multiple dialect pronunciation data; selecting a correct phoneme sequence from the candidate phoneme sequence group; and constructing a dialect pronunciation dictionary according to the vocabulary and the correct phoneme sequence. The method comprises the steps of obtaining various dialect pronunciation data for words which can be added to a dialect pronunciation dictionary, inputting the various dialect pronunciation data into a voice recognition device to obtain a candidate phoneme sequence group, selecting a correct phoneme sequence from the candidate phoneme sequence group to establish a mapping relation with the words to obtain the dialect pronunciation dictionary, inputting the various dialect pronunciation data, obtaining more diverse results in the candidate phoneme sequence group, selecting the correct phoneme sequence from the candidate phoneme sequence group to ensure more correct results, and improving the accuracy of the recognition result of the finally constructed dialect pronunciation dictionary.

Description

Method, device, device and readable storage medium for constructing dialect pronunciation dictionary

technical field

The present invention relates to the field of speech recognition, and in particular, to a method, device, device and readable medium for constructing a dialect pronunciation dictionary.

Background technique

The pronunciation dictionary is an important part of the speech recognition model. The received speech information can be converted into processable phoneme information through the dialect dictionary.

In the process of building the pronunciation dictionary, the pronunciation of Mandarin is relatively uniform, and the recognition accuracy is high, but there are large differences between dialect pronunciations. There are no small differences, so the pronunciation dictionary has a low accuracy for dialect recognition.

SUMMARY OF THE INVENTION

The main purpose of the present application is to provide a method, device, device and readable storage medium for constructing a dialect pronunciation dictionary, aiming at solving the problem of low accuracy of the recognition result of the dialect pronunciation dictionary.

In order to achieve the above object, the construction method of a dialect pronunciation dictionary provided by the application, the construction method of the dialect pronunciation dictionary comprises the following steps:

Input the pronunciation data of multiple dialects of the vocabulary into the speech recognition device;

receiving a candidate phoneme sequence group output by the speech recognition device based on the pronunciation data of the multiple dialects;

selecting the correct phoneme sequence from the candidate phoneme sequence group;

According to the vocabulary and the correct phoneme sequence, a dialect pronunciation dictionary is constructed.

Optionally, before the step of inputting the pronunciation data of multiple dialects of the vocabulary into the speech recognition device, it includes:

The standard phoneme sequence of the Mandarin pronunciation corresponding to the vocabulary is added to the speech recognition device.

Optionally, the step of inputting the multiple dialect pronunciation data of the vocabulary into the speech recognition device includes:

Input the pronunciation data of the various dialects of the vocabulary into the speech recognition device;

Repeatingly input the dialect pronunciation data of the same source in each dialect category of the vocabulary into the speech recognition device;

The dialect pronunciation data of different origins in each dialect category of the vocabulary are input into the speech recognition device.

Optionally, after the step of receiving the candidate phoneme sequence groups output by the speech recognition apparatus based on the pronunciation data of the multiple dialects, the following steps are included:

Associating the candidate phoneme sequence with the dialect category to which it belongs;

The probability distribution of each candidate phoneme sequence in the candidate phoneme sequence group of the same type of dialect pronunciation data is determined.

Optionally, the step of selecting the correct phoneme sequence from the candidate phoneme sequence group includes:

determining the maximum value in the probability distribution;

The candidate phoneme sequence in the candidate phoneme sequence group corresponding to the maximum value is used as the correct phoneme sequence.

Optionally, after the step of selecting the correct phoneme sequence from the candidate phoneme sequence, it includes:

comparing the correct phoneme sequence with the standard phoneme sequence;

If the correct phoneme sequence is different from the standard phoneme sequence, a mapping is established between the previous phoneme sequence and the standard phoneme sequence.

Optionally, the construction method of the dialect pronunciation dictionary includes:

If there is a dialect vocabulary that cannot correspond to Mandarin in the dialect category, the dialect vocabulary and the corresponding dialect phoneme sequence are directly mapped.

The present application also provides a device for constructing a dialect pronunciation dictionary, and the device for constructing the dialect pronunciation dictionary includes:

an input module for inputting the pronunciation data of various dialects of the vocabulary into the speech recognition device;

a receiving module, configured to receive a candidate phoneme sequence group output by the speech recognition device based on the dialect pronunciation data;

a selection module for selecting the correct phoneme sequence from the candidate phoneme sequence group;

The building module is used for building a dialect pronunciation dictionary according to the vocabulary and the correct phoneme sequence.

The present application also provides a device for constructing a dialect pronunciation dictionary. The device for constructing the dialect pronunciation dictionary includes: a memory, a processor, and a program for constructing a dialect pronunciation dictionary that is stored on the memory and can be run on the processor. , when the program for constructing the dialect pronunciation dictionary is executed by the processor to implement the steps of the above method for constructing the dialect pronunciation dictionary.

The present application also provides a readable storage medium, where a computer program is stored on the readable storage medium, and when the computer program is executed by a processor, the steps of the above method for constructing a dialect pronunciation dictionary are implemented.

In the present application, the pronunciation data of various dialects of the vocabulary is input into the speech recognition device; the candidate phoneme sequence group output by the speech recognition device based on the pronunciation data of the various dialects is received; the correct phoneme is selected from the candidate phoneme sequence group sequence; construct a dialect pronunciation dictionary according to the vocabulary and the correct phoneme sequence. For the vocabulary to be added to the dialect pronunciation dictionary, obtain its various dialect pronunciation data and input it into the speech recognition device, obtain a candidate phoneme sequence group, select the correct phoneme sequence and the vocabulary to establish a corresponding mapping relationship, thereby obtaining the dialect pronunciation dictionary, Inputting various dialect pronunciation data, the result in the candidate phoneme sequence group is more diverse. At the same time, selecting the correct phoneme sequence from the candidate phoneme sequence group ensures the result is more correct and improves the accuracy of the recognition result of the final constructed dialect pronunciation dictionary.

Description of drawings

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

In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the accompanying drawings that need to be used in the description of the embodiments or the prior art. In other words, on the premise of no creative labor, other drawings can also be obtained from these drawings.

FIG. 1 is a schematic diagram of the device structure of the hardware operating environment involved in the solution of the embodiment of the present application;

2 is a schematic flowchart of the first embodiment of the construction method of the dialect pronunciation dictionary of the application;

Fig. 3 is the refinement flow chart of step S10 of Fig. 2 and the previous steps in the second embodiment of the construction method of the dialect pronunciation dictionary of the present application;

FIG. 4 is a detailed flow chart of the steps after step S20 of FIG. 2 in the third embodiment of the construction method of the dialect pronunciation dictionary of the present application;

5 is a detailed flow chart of step S30 and subsequent steps in FIG. 2 in the fourth embodiment of the construction method of the dialect pronunciation dictionary of the present application;

FIG. 6 is a schematic diagram of the system structure of an embodiment of an apparatus for constructing a dialect pronunciation dictionary of the present application.

The realization, functional characteristics and advantages of the purpose of the present application will be further described with reference to the accompanying drawings in conjunction with the embodiments.

Detailed ways

It should be understood that the specific embodiments described herein are only used to explain the present application, but not to limit the present application.

In the following description, suffixes such as 'module', 'component' or 'unit' used to represent elements are used only to facilitate the description of the present invention and have no specific meaning per se. Thus, "module", "component" or "unit" may be used interchangeably.

As shown in FIG. 1 , FIG. 1 is a schematic structural diagram of a terminal of a hardware operating environment involved in the solution of the embodiment of the present application.

The terminal in the embodiment of the present application is a device for constructing a dialect pronunciation dictionary.

As shown in FIG. 1 , the terminal may include: a processor 1001 , such as a CPU, a network interface 1004 , a user interface 1003 , a memory 1005 , and a communication bus 1002 . Among them, the communication bus 1002 is used to realize the connection and communication between these components. The user interface 1003 may include a display screen (Display), an input unit such as a keyboard (Keyboard), and the optional user interface 1003 may also include a standard wired interface and a wireless interface. Optionally, the network interface 1004 may include a standard wired interface and a wireless interface (eg, a WI-FI interface). The memory 1005 may be high-speed RAM memory, or may be non-volatile memory, such as disk memory. Optionally, the memory 1005 may also be a storage device independent of the aforementioned processor 1001 .

Optionally, the terminal may further include a camera, an RF (Radio Frequency, radio frequency) circuit, a sensor, an audio circuit, a WiFi module, and the like. Among them, sensors such as light sensors, motion sensors and other sensors. Specifically, the light sensor may include an ambient light sensor and a proximity sensor, wherein the ambient light sensor may adjust the brightness of the display screen according to the brightness of the ambient light, and the proximity sensor may turn off the display screen and/or when the terminal device moves to the ear Backlight. Of course, the terminal device may also be configured with other sensors such as a gyroscope, a barometer, a hygrometer, a thermometer, an infrared sensor, etc., which will not be repeated here.

Those skilled in the art can understand that the terminal structure shown in FIG. 1 does not constitute a limitation on the terminal, and may include more or less components than the one shown, or combine some components, or arrange different components.

As shown in FIG. 1 , the memory 1005 as a computer storage medium may include an operating system, a network communication module, a user interface module, and a program for constructing a dialect pronunciation dictionary.

In the terminal shown in FIG. 1 , the network interface 1004 is mainly used to connect to the background server and perform data communication with the background server; the user interface 1003 is mainly used to connect to the client (client) and perform data communication with the client; and the processor 1001 can be used to invoke the construction program of the dialect pronunciation dictionary stored in the memory 1005, and perform the following operations:

Input the dialect pronunciation data of the dialect vocabulary into the speech recognition device;

receiving a candidate phoneme sequence output by the speech recognition device based on the dialect pronunciation data;

selecting the correct phoneme sequence from the candidate phoneme sequence;

According to the dialect vocabulary and the correct phoneme sequence, a dialect pronunciation dictionary is constructed.

Based on the above-mentioned terminal hardware structure, various embodiments of the present application are proposed.

The present application provides a method for constructing a dialect pronunciation dictionary.

Referring to Fig. 2, in the first embodiment of the construction method of the dialect pronunciation dictionary, the method includes:

Step S10, input the multiple dialect pronunciation data of the vocabulary into the speech recognition device;

The pronunciation dictionary contains a collection of words that the system can handle, with their pronunciations marked. The mapping relationship between the modeling unit of the acoustic model and the modeling unit of the language model is obtained through the pronunciation dictionary, so that the acoustic model and the language model are connected to form a searched state space for the decoder to decode. The pronunciation dictionary contains the mapping from words to phonemes, which is used to connect the acoustic model and the language model. Therefore, the pronunciation dictionary contains the mapping relationship between a single word and the corresponding phoneme. For words commonly used in dialects, the pronunciation data of various dialects corresponding to each word are input into the speech recognition device. The dialect pronunciation data can include the pronunciation data of a variety of different types of dialects, and can also include the pronunciation data of multiple different speakers in the same type of dialect. Entering multiple dialect pronunciation data increases the diversity of the input dialect pronunciation data. A more diverse set of candidate phoneme sequences can be obtained. The dialect pronunciation dictionary obtained by inputting the pronunciation data of multiple different dialects at the same time will be more complete.

Step S20, receiving the candidate phoneme sequence groups output by the speech recognition device based on the pronunciation data of the multiple dialects;

The conventional method for converting vocabulary into corresponding phonemes is G2P (Grapheme-to-PhonemeConversion, word-to-phoneme conversion). For the input dialect pronunciation data, the speech recognition device first recognizes the pinyin data in the dialect pronunciation data, and then according to the identification According to the input dialect pronunciation data, the speech recognition device will output a plurality of possible phoneme sequences to form a candidate phoneme sequence group, and the probability values corresponding to each phoneme sequence in the candidate phoneme sequence group are different. When converting the identified pinyin data to phoneme data, different conversion rules or mapping relationships may also affect the obtained phoneme sequence, so a unified conversion rule or mapping relationship needs to be determined in the speech recognition device.

Step S30, selecting the correct phoneme sequence from the candidate phoneme sequence group;

The candidate phoneme sequence group contains multiple different phoneme sequences, and the correct phoneme sequence refers to the phoneme sequence that is closest to the standard dialect pronunciation in the dialect. In the present application, the probability of each phoneme sequence in the candidate phoneme sequence group is used as the evaluation criterion, and the preset sequence with the highest occurrence probability in the candidate phoneme sequence group is selected as the correct phoneme sequence. Meanwhile, the probability calculation method of each phoneme sequence in the candidate phoneme sequence is not unique. At the same time, according to the different acoustic models in the speech recognition device, the obtained probability may also be different. The acoustic model commonly used in speech recognition devices is generally a hybrid model of hidden Markov-deep neural network, and some acoustic models are obtained only by training the deep neural network model.

Step S40, constructing a dialect pronunciation dictionary according to the vocabulary and the correct phoneme sequence;

The correct phoneme sequence is selected from the candidate phoneme sequence group, each vocabulary and the corresponding correct phoneme sequence are formed into a lexical entry, and a dialect pronunciation dictionary is constructed. In the constructed dialect pronunciation dictionary, for a word, there may be multiple corresponding multiple correct phoneme sequences, because the same word may exist in different dialect types, and the pronunciation of the same word in different dialect types may be different. Therefore, the dialect pronunciation dictionary also includes the mapping relationship between dialect words and corresponding dialect types. When finally using the dialect pronunciation dictionary, it is necessary to first determine the corresponding dialect type, and then find the corresponding correct phoneme sequence.

In this embodiment, the pronunciation data of multiple dialects of the vocabulary is input into the speech recognition device; the candidate phoneme sequence group output by the speech recognition device based on the pronunciation data of the multiple dialects is received; from the candidate phoneme sequence group Select the correct phoneme sequence from the lexicon and the correct phoneme sequence, construct a dialect pronunciation dictionary. For the vocabulary to be added to the dialect pronunciation dictionary, obtain its various dialect pronunciation data and input it into the speech recognition device, obtain a candidate phoneme sequence group, select the correct phoneme sequence and the vocabulary to establish a corresponding mapping relationship, thereby obtaining the dialect pronunciation dictionary, Inputting various dialect pronunciation data, the result in the candidate phoneme sequence group is more diverse. At the same time, selecting the correct phoneme sequence from the candidate phoneme sequence group ensures the result is more correct and improves the accuracy of the recognition result of the final constructed dialect pronunciation dictionary.

Further, referring to FIG. 2 and FIG. 3, on the basis of the first embodiment of the construction method of the dialect pronunciation dictionary of the present application, a second embodiment of the construction method of the dialect pronunciation dictionary is provided. In the second embodiment,

Before step S10, it includes:

Step S11, adding the standard phoneme sequence of the Mandarin pronunciation corresponding to the vocabulary to the speech recognition device;

Dialect pronunciation dictionaries are relatively difficult, but the existing Chinese pronunciation dictionaries are constructed based on Mandarin pronunciation, and can cover the phoneme sequences of most common words. Here, the standard phoneme sequences of Mandarin pronunciation corresponding to the words are used. Adding to the speech recognition device means adding the existing Chinese pronunciation dictionary to the speech recognition device. Standard phoneme sequences are more general and cover a wider vocabulary for speech recognition devices.

Step S10 includes:

Step S12, input the multiple dialect pronunciation data of the vocabulary into the speech recognition device;

There are various types of dialects. For a word, the pronunciations in different dialects may be quite different. Therefore, it is necessary to input the pronunciation data of different dialects into the speech recognition device to obtain candidate phonemes in different dialect types. At the same time, the dialect pronunciation data of different dialect types can be used to improve the dialect types of the constructed dialect pronunciation dictionary, expand the use range of the dialect pronunciation dictionary, and do not need to rebuild the dialect pronunciation dictionary for different types of dialect data.

Step S13, repeatedly input the dialect pronunciation data of the same origin in each dialect type of the vocabulary into the speech recognition device;

For each dialect pronunciation data, it needs to be repeatedly input into the speech recognition device. For the same dialect pronunciation data, because the acoustic model in the speech recognition device is the neural network that outputs the corresponding results for the input data, because the neural network There is a certain probability uncertainty, so for the same dialect pronunciation data, the output phoneme sequence may also be different. Therefore, the repeated input of the same dialect pronunciation data also increases the diversity of candidate phoneme sequences.

Step S14, input the dialect pronunciation data from different sources in each dialect type of the vocabulary into the speech recognition device;

Dialect pronunciation data from different sources can be dialect pronunciation data collected from different people. There are differences between different types of dialects. For groups speaking the same dialect, different speakers will also have differences. Vocabulary and pronunciation pitches of different people may also be different. Entering the dialect pronunciation data of multiple different individuals can avoid errors in the acquired phoneme sequence caused by only inputting the dialect pronunciation data of the same individual.

In this embodiment, the input dialect pronunciation data needs to include different types of dialect pronunciation data, dialect pronunciation data of different individuals, and at the same time, each input dialect pronunciation data must be repeatedly input multiple times, so it is guaranteed that the candidates obtained later Accuracy of phoneme sequence groups.

Further, referring to FIG. 2 and FIG. 4 , on the basis of the second embodiment of the construction method of the dialect pronunciation dictionary of the present application, a third embodiment of the construction method of the dialect pronunciation dictionary is provided. In the third embodiment,

After step S20, it includes:

Step S21, associating the candidate phoneme sequence with the dialect category to which it belongs;

The input dialect pronunciation data contains the pronunciation data of different types of dialects. The candidate phoneme sequences corresponding to the different types of dialect pronunciation data need to be distinguished. Therefore, the candidate phoneme sequences and the corresponding dialect types are marked together. The dialect type refers to Cantonese. , Shanghai dialect, Sichuan dialect, etc. That is, the candidate phoneme sequences obtained from the same vocabulary should be classified into the corresponding dialect categories.

Step S22, determining the probability distribution of each candidate phoneme sequence in the candidate phoneme sequence group of the same type of dialect pronunciation data;

After the candidate phoneme sequences are classified, for the candidate phoneme sequences of the same dialect type, the candidate phoneme sequence groups under the dialect type are formed, and the probability distribution of the candidate phoneme sequence groups can be calculated by using the Bayesian formula. Formula, using the acoustic model and the speech model in the speech recognition device, for the dialect pronunciation data from different individual sources, the obtained candidate phoneme sequences can also be introduced to calculate the probability distribution of the candidate phoneme sequence groups.

In this embodiment, each obtained candidate phoneme sequence is classified into the corresponding dialect type to form a candidate phoneme sequence group, and the probability distribution in the candidate phoneme sequence group is calculated by using the Bayesian formula.

Further, referring to FIG. 2 and FIG. 5, on the basis of the third embodiment of the construction method of the dialect pronunciation dictionary of the present application, a fourth embodiment of the construction method of the dialect pronunciation dictionary is provided. In the fourth embodiment,

Step S30 includes:

Step S31, determining the maximum value in the probability distribution;

There are two possible ways to determine the maximum value in the probability distribution. The first method is, for each dialect pronunciation data from different sources, for the repeated input of each dialect pronunciation data, a candidate phoneme sequence will be obtained for each input, and each candidate phoneme sequence regardless of whether it is related to other candidate phoneme sequences or not. are the same as an independent candidate phoneme sequence, and then classify the same candidate phoneme sequence into one category, count the number of candidate phoneme sequences in the same category, and compare the number of candidate phoneme sequences in each category with all candidate phoneme sequences. The ratio of the number of sequences is used as the probability distribution in the candidate phoneme sequence group, and the maximum value of the probability distribution represents the phoneme sequence with the largest number of occurrences obtained after the same word is processed by the speech recognition device, and the phoneme sequence should also be The most matching phoneme sequence for this word in this dialect category. The second method is to substitute the obtained different candidate phoneme sequences into the Bayesian formula according to the Bayesian formula to obtain the value of the Bayesian formula, thereby obtaining the maximum value of the Bayesian formula, that is, the probability distribution. the maximum value in . The value obtained by the Bayesian formula is the degree of matching between the candidate phoneme sequence and the input lexical information.

Step S32, taking the candidate phoneme sequence in the candidate phoneme sequence group corresponding to the maximum value as the correct phoneme sequence;

The candidate phoneme sequence corresponding to the maximum value is the correct phoneme sequence obtained after processing by the speech recognition device, and the correct phoneme sequence represents the most likely phoneme sequence obtained according to the dialect pronunciation dictionary when the same vocabulary is received. However, the correct phoneme sequence will be selected, but other candidate phoneme sequences will still be retained. Other candidate phoneme sequences are reserved, which can be used if the corresponding vocabulary needs to be found through the phoneme sequence. Retaining other candidate phoneme sequences can also improve the accuracy of the recognition results. In normal daily communication, the pronunciation of words before and after a word will also affect the pronunciation characteristics of the word itself. Therefore, other candidate phoneme sequences that are different from the correct phoneme sequence are reserved. When using the dialect pronunciation dictionary to match words in continuous speech data , the results obtained are more accurate.

Step S33, comparing the correct phoneme sequence with the standard phoneme sequence;

In general, for the same vocabulary, the pronunciations in different dialects are different, and the pronunciation between dialects and Mandarin is even more different. If the pronunciations of dialects and Mandarin are the same, the standard phoneme sequence can be used as the correct phoneme sequence.

Step S34, if the correct phoneme sequence is different from the standard phoneme sequence, then establish a mapping between the correct phoneme sequence and the standard phoneme sequence;

The correct phoneme sequence obtained through the input dialect pronunciation data is different from the standard phoneme sequence. The correct phoneme sequence is mapped with the standard phoneme sequence, that is, the standard phoneme sequence can be found through the correct phoneme sequence. In steps S33 to S34, the correct phoneme sequence and the standard phoneme sequence are mapped to connect the dialect pronunciation of the vocabulary with the Mandarin, because compared with the dialect, the Mandarin pronunciation dictionary is more complete, and the recognition result will be more accurate at the same time. At the same time, if the voice information needs to be converted into text information, it is also convenient to convert into dialect text and Mandarin text at the same time. Of course, steps S33 to S34 are optional, and there is no need to establish a mapping relationship between the correct phoneme sequence and the standard phoneme sequence.

In this embodiment, the candidate phoneme sequence corresponding to the maximum probability distribution in the candidate phoneme sequence group is selected as the correct phoneme sequence, and the vocabulary and the correct phoneme sequence are mapped, and the correct phoneme sequence and the standard phoneme sequence can also be mapped. The conversion between the phoneme sequence of dialect pronunciation and the phoneme sequence of Mandarin pronunciation improves the accuracy of the recognition result of the dialect pronunciation dictionary.

Further, on the basis of the fourth embodiment of the construction method of the dialect pronunciation dictionary of the present application, a fifth embodiment of the construction method of the dialect pronunciation dictionary is provided. In the fifth embodiment,

The construction method of the dialect pronunciation dictionary includes:

Step A, if there is a vocabulary that cannot be corresponding to Mandarin in the dialect category, then directly establish a mapping between the vocabulary and the corresponding dialect phoneme sequence;

Some words may be exclusive words in dialects, that is, words with the same meaning cannot be found in Mandarin. For such words, the corresponding dialect phoneme sequence needs to be added to the dialect pronunciation dictionary separately, and the corresponding mapping is established with the corresponding words. However, for words that do not exist in Mandarin but are used in dialects, it is not necessary to add all of them to the dialect pronunciation dictionary. You can choose to add commonly used words such as modal particles or other words similar to mantras to the dialect pronunciation dictionary. The determination of the dialect phoneme sequence is determined according to the dialect habits of most people in the dialect, and some common other phoneme sequences can be added as supplements.

In this embodiment, for the special vocabulary in the dialect, the mapping relationship between the corresponding phoneme sequence and the vocabulary is additionally added, which expands the vocabulary included in the dialect pronunciation dictionary and improves the recognition accuracy of the dialect pronunciation dictionary.

In addition, referring to FIG. 6 , an embodiment of the present application also proposes a device for constructing a dialect pronunciation dictionary, and the device for constructing the dialect pronunciation dictionary includes:

an input module for inputting the pronunciation data of various dialects of the vocabulary into the speech recognition device;

a receiving module, configured to receive the candidate phoneme sequence groups output by the speech recognition device based on the pronunciation data of the multiple dialects;

a selection module for selecting the correct phoneme sequence from the candidate phoneme sequence group;

The building module is used for building a dialect pronunciation dictionary according to the vocabulary and the correct phoneme sequence.

The expanded contents of the specific implementations of the device and the readable storage medium (ie, the computer-readable storage medium) of the present application are basically the same as the above-mentioned embodiments of the construction method of the dialect pronunciation dictionary, and will not be repeated here.

It should be noted that, herein, the terms "comprising", "comprising" or any other variation thereof are intended to encompass non-exclusive inclusion, such that a process, method, article or device comprising a series of elements includes not only those elements, It also includes other elements not expressly listed or inherent to such a process, method, article or apparatus. Without further limitation, an element qualified by the phrase "comprising a..." does not preclude the presence of additional identical elements in a process, method, article or apparatus that includes the element.

The above-mentioned serial numbers of the embodiments of the present invention are only for description, and do not represent the advantages or disadvantages of the embodiments.

From the description of the above embodiments, those skilled in the art can clearly understand that the method of the above embodiment can be implemented by means of software plus a necessary general hardware platform, and of course can also be implemented by hardware, but in many cases the former is better implementation. Based on this understanding, the technical solutions of the present invention can be embodied in the form of software products in essence or the parts that make contributions to the prior art, and the computer software products are stored in a storage medium (such as ROM/RAM, magnetic disk, CD), including several instructions to make a terminal (which may be a mobile phone, a computer, a server, an air conditioner, or a network device, etc.) execute the methods described in the various embodiments of the present invention.

The embodiments of the present invention have been described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific embodiments, which are merely illustrative rather than restrictive. Under the inspiration of the present invention, without departing from the scope of protection of the present invention and the claims, many forms can be made, which all belong to the protection of the present invention.

Claims (10)

1. a construction method of dialect pronunciation dictionary, is characterized in that, the construction method of described dialect pronunciation dictionary comprises the following steps: Input the pronunciation data of multiple dialects of the vocabulary into the speech recognition device; receiving a candidate phoneme sequence group output by the speech recognition device based on the pronunciation data of the multiple dialects; selecting the correct phoneme sequence from the candidate phoneme sequence group; According to the vocabulary and the correct phoneme sequence, a dialect pronunciation dictionary is constructed. 2. the construction method of the dialect pronunciation dictionary as claimed in claim 1 is characterized in that, before the described step of inputting the multiple dialect pronunciation data of vocabulary in the speech recognition device, comprises: The standard phoneme sequence of the Mandarin pronunciation corresponding to the vocabulary is added to the speech recognition device. 3. the construction method of dialect pronunciation dictionary as claimed in claim 2 is characterized in that, the described step of inputting the multiple dialect pronunciation data of vocabulary into speech recognition device comprises: Input the pronunciation data of the various dialects of the vocabulary into the speech recognition device; Repeatingly input the dialect pronunciation data of the same source in each dialect category of the vocabulary into the speech recognition device; The dialect pronunciation data of different origins in each dialect category of the vocabulary are input into the speech recognition device. 4. the construction method of dialect pronunciation dictionary as claimed in claim 3, is characterized in that, after described receiving the step of the candidate phoneme sequence group that described speech recognition device outputs based on described multiple dialect pronunciation data comprises: Associating the candidate phoneme sequence with the dialect category to which it belongs; The probability distribution of each candidate phoneme sequence in the candidate phoneme sequence group of the same type of dialect pronunciation data is determined. 5. the construction method of dialect pronunciation dictionary as claimed in claim 4 is characterized in that, the described step of selecting correct phoneme sequence from described candidate phoneme sequence group comprises: determining the maximum value in the probability distribution; The candidate phoneme sequence in the candidate phoneme sequence group corresponding to the maximum value is used as the correct phoneme sequence. 6. the construction method of dialect pronunciation dictionary as claimed in claim 5, is characterized in that, after the described step of selecting correct phoneme sequence from described candidate phoneme sequence group, comprises: comparing the correct phoneme sequence with the standard phoneme sequence; If the correct phoneme sequence is different from the standard phoneme sequence, a mapping is established between the correct phoneme sequence and the standard phoneme sequence. 7. the construction method of dialect pronunciation dictionary as claimed in claim 6 is characterized in that, the construction method of described dialect pronunciation dictionary comprises: If there is a vocabulary that cannot be corresponding to Mandarin in the dialect category, the vocabulary is directly mapped to the corresponding dialect phoneme sequence. 8. a construction device of a dialect pronunciation dictionary, is characterized in that, the construction device of described dialect pronunciation dictionary comprises: an input module for inputting the pronunciation data of various dialects of the vocabulary into the speech recognition device; a receiving module, configured to receive the candidate phoneme sequence groups output by the speech recognition device based on the pronunciation data of the multiple dialects; a selection module for selecting the correct phoneme sequence from the candidate phoneme sequence group; The building module is used for building a dialect pronunciation dictionary according to the vocabulary and the correct phoneme sequence. 9. a construction device of a dialect pronunciation dictionary, is characterized in that, described device comprises: memory, processor and the construction program of the dialect pronunciation dictionary that is stored on described memory and can run on described processor, described When the program for constructing the dialect pronunciation dictionary is executed by the processor, the steps of the method for constructing the dialect pronunciation dictionary according to any one of claims 1 to 7 are implemented. 10. A readable storage medium, wherein a computer program is stored on the readable storage medium, and when the computer program is executed by a processor, the dialect pronunciation according to any one of claims 1 to 7 is realized The steps of the dictionary building method.

Images