Speech translation method and device
A technology of speech translation and target speech, applied in the field of speech translation, can solve problems such as wrong translation results and inaccurate translation results
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Method used
Image
Examples
no. 1 example
[0079] see figure 1 , which is a schematic flowchart of a voice translation method provided in this embodiment, the method includes the following steps:
[0080] S101: Obtain a target speech to be translated.
[0081] In this embodiment, any speech that is used for speech translation in this embodiment is defined as the target speech. And, the present embodiment does not limit the language type of the target voice, for example, the target voice can be Chinese voice or English voice, etc.; meanwhile, the present embodiment does not limit the length of the target voice, for example, the target voice can be a sentence, or More words and so on.
[0082] It can be understood that the target voice can be obtained by recording according to actual needs. For example, the voice of telephone conversations in people's daily life, or conference recordings, etc. can be used as the target voice. After the target voice is obtained, this implementation can be used to Example realizes the t...
no. 2 example
[0147] In this embodiment, by translating the first translation object (i.e., the recognized text of the target speech) through step S102 in the above-mentioned first embodiment, the first probability corresponding to the kth word in the final translated text of the target speech can be generated distribution, and this first probability distribution can be defined as P text (y k ), where y k refers to the kth word in the final translated text of the target speech.
[0148] Among them, the first probability distribution P text (y k ) can include the kth word y obtained after decoding the recognized text of the target speech k is the first decoding probability of each candidate word in the vocabulary. The larger the value of the first decoding probability, it indicates that the kth word y obtained after decoding the recognized text of the target speech k The probability of corresponding to the word to be selected is greater.
[0149] combine Figure 4 The network structu...
no. 3 example
[0173] In this embodiment, the first translated text can be obtained by translating the first translation object (ie, the recognized text of the target speech) through step S102 in the above-mentioned first embodiment.
[0174] Among them, the first translation text is the text of the target translation language, and the first translation text can be defined as Among them, K 1 Indicates the number of single characters (or words) contained in the first translated text. For example, assuming that the target speech is Chinese and the target translation language is English, that is, the target speech needs to be translated into English text, then the first translated text is the English text Among them, K 1 Indicates the number of words contained in the English text.
[0175] An optional implementation is to combine Figure 4 The network structure shown can use the decoding vector obtained by the text decoder to generate the first translated text of the target speech Among...
PUM
Login to View More Abstract
Description
Claims
Application Information
Login to View More