A voice evaluation scoring method based on scene requirement description, a storage medium and an equipment

Abstract

The application discloses a voice evaluation scoring method based on scene requirement description, a storage medium and equipment, and the method comprises the following steps: S1, inputting scene description data for training; S2, encoding the scene description data input in step S1; S3, fusing the scene vector obtained by encoding the scene description in step S2 with acoustic characteristics for training; S4, inputting the fusion result of step S3 into a voice evaluation model to train the voice evaluation model; and S5, inputting actual scene description data and voice file data to be scored into the trained voice evaluation model, and the voice evaluation model outputs a scoring result. The method can effectively adapt to various scene requirements by combining a large amount of data and scene description.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent education, in particular to a voice evaluation scoring method based on scene requirement description, a storage medium and equipment. BACKGROUND

[0002] In the current various oral examinations, voice evaluation models have been used for automatic scoring, which is of great significance to improve scoring efficiency. However, the existing voice evaluation model scoring has instability, especially for different schools and regions, and a single model is difficult to adapt to various scoring requirements, so the voice evaluation model needs to be retrained with each requirement adjustment. SUMMARY

[0003] In view of the defects of the prior art, the present application aims to provide a voice evaluation scoring method based on scene requirement description, a storage medium and equipment.

[0004] In order to achieve the above-mentioned purpose, the technical scheme adopted by the present application is as follows:

[0005] A voice evaluation scoring method based on scene requirement description, comprising the following steps:

[0006] S1, inputting scene description data for training;

[0007] S2, encoding the scene description input in step S1 to obtain a scene vector: for discrete scene dimension feature, using embedding coding technology for vectorization; for continuous scene dimension feature, dividing into N discrete features of equal level, and then using embedding coding technology to vectorize the discrete features obtained by division, N is a positive integer;

[0008] S3, fusing each scene vector obtained by encoding the scene description in step S2 with the acoustic feature for training;

[0009] S4, inputting the fusion result of the acoustic feature and the scene vector obtained in step S3 into the voice evaluation model to train the voice evaluation model;

[0010] S5, when using the voice evaluation model trained in step S4 to score, inputting the scene description data and the voice file data to be scored into the voice evaluation model, and the voice evaluation model outputs the scoring result.

[0011] Further, in step S1, the scene description includes vocabulary level, test paper difficulty level, grade, region code or dialect code, and scoring tightness.

[0012] Further, in step S2, the discrete scene dimension features include test paper difficulty level, grade, region code or dialect code, and vocabulary level, and the continuous scene dimension features include scoring tightness.

[0013] Further, in step S4, the speech evaluation model adopts a Transformer-based encoder-decoder network structure.

[0014] The application also provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the above method.

[0015] The application also provides a computer device, which comprises a processor and a memory, and the memory is used to store a computer program; and the processor is used to execute the computer program to implement the above method.

[0016] The application has the beneficial effect that in the method, a large amount of data is pre-trained, including vocabulary level, test paper difficulty level, grade, region code or dialect code, and scoring tightness, and the obtained speech evaluation model has the adaptation capability under various scene conditions, and only needs to adjust specific scene parameters to adapt to the scene, thereby effectively adapting to various scene requirements and reducing the work of model retraining. BRIEF DESCRIPTION OF DRAWINGS

[0017] Figure 1 The method flowchart of the embodiment of the application is shown. DETAILED DESCRIPTION

[0018] The application will be further described below with reference to the drawings, and it should be noted that the embodiment is based on the technical solution, and detailed implementation and specific operation process are given, but the protection scope of the application is not limited to the embodiment.

[0019] The embodiment provides a speech evaluation scoring method based on scene requirement description, as shown in Figure 1 The method comprises the following steps:

[0020] S1, input scene description data for training, including vocabulary level, test paper difficulty level, grade, region code or dialect code, and scoring tightness; and the scoring tightness can be any decimal between 0 and 1.

[0021] S2, encode the scene description input in step S1 to obtain a scene vector: for discrete scene dimension features (including test paper difficulty level, grade, region code or dialect code, and vocabulary level), use embedding encoding technology for vectorization; for continuous scene dimension features (including scoring tightness), divide into N discrete features, and then use embedding encoding technology for vectorization, N being a positive integer;

[0022] S3, fuse each scene vector obtained by encoding the scene description in step S2 with the acoustic feature used for training.

[0023] In this embodiment, the acoustic feature uses the FBank feature commonly used in speech evaluation and recognition modeling.

[0024] In this embodiment, each scene vector and acoustic feature are fused by splicing in the length direction. Let the acoustic feature be T*D, the scene vector be M*D, and the feature after splicing the acoustic feature and the scene vector be (T+M)*D, where T represents the length of the acoustic feature, M represents the total number of scene vectors obtained by encoding the scene description, D represents the dimension of the sound feature and the scene vector, and the dimension of the sound feature and the scene vector is the same.

[0025] S4, input the fusion result of the acoustic feature and the scene vector obtained in step S3 into a speech evaluation model to train the speech evaluation model.

[0026] In this embodiment, the speech evaluation model uses a Transformer-based encoder-decoder network structure. The encoder receives T acoustic features and M scene vectors and encodes them, and then the decoder outputs T scoring results. The decoder can have 1 to multiple, and when there are multiple, different scoring dimensions (such as accuracy, fluency, and rhythm) are respectively corresponding.

[0027] S5, when scoring using the speech evaluation model trained in step S4, input the scene description data and the speech file data to be scored into the speech evaluation model, and the speech evaluation model outputs the scoring result.

[0028] For those skilled in the art, various corresponding changes and modifications can be made according to the above technical solutions and concepts, and all these changes and modifications should be included in the protection scope of the claims of the present application.

Claims

1. A speech evaluation and scoring method based on scenario requirement description, characterized in that, Includes the following steps: S1. Input the scene description data for training, which includes vocabulary level, test paper difficulty level, grade, regional code or dialect code, and scoring leniency. S2. Encode the scene description input in step S1 to obtain a scene vector: For discrete scene dimensional features, use embedding encoding technology to vectorize them; for continuous scene dimensional features, divide them into N levels of discrete features, and then use embedding encoding technology to vectorize the divided discrete features, where N is a positive integer. S3. Fuse the scene vectors obtained by encoding the scene description in step S2 with the acoustic features used for training. S4. Input the fusion result of the acoustic features and scene vectors obtained in step S3 into the speech evaluation model to train the speech evaluation model. S5. When using the speech evaluation model trained in step S4 to score, input the scene description data and the speech file data to be scored into the speech evaluation model, and the speech evaluation model outputs the scoring results.

2. The method according to claim 1, characterized in that, In step S2, discrete scenario dimension features include test paper difficulty level, grade, regional code or dialect code, and vocabulary level, while continuous scenario dimension features include grading tightness.

3. The method according to claim 1, characterized in that, In step S4, the speech evaluation model adopts a Transformer codec network structure.

4. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the method described in any one of claims 1-3.

5. A computer device, characterized in that, It includes a processor and a memory, the memory being used to store a computer program; the processor being used to execute the computer program to implement the method of any one of claims 1-3.

Images