1  Scope of One-to-Many Speech Translation

2  Reference Architecture of One-to-Many Speech Translation

3  I/O Data of One-to-Many Speech Translation

4  Functions of AI Modules of One-to-Many Speech Translation

5  I/O Data of AI Modules of One-to-Many Speech Translation

6  JSON Metadata of One-to-Many Speech Translation

1      Scope of One-to-Many Speech Translation

The goal of the One-to-Many Speech Translation (MMC-MST) Use Case is to enable one person speaking his or her language to broadcast to two or more audience members, each listening and responding in a different language, presented as speech or text. If the desired output is speech, users can specify whether their speech features (voice colour, emotional charge, etc.) should be preserved in the translated speech.

The flow of control (from Recognised Text to Translated Text to Output Speech) is identical to that of the Unidirectional case. However, rather than one such flow, multiple paired flows are provided – the first pair from language A to language B and B to A; the second from A to C and C to A; and so on.

Depending on the value of Input Selector (text or speech):

  1. Input Text in Language A is translated into Translated Text in and pronounced as Speech of all Requested Languages.
  2. The Speech features (voice colour, emotional charge, etc.) in Language A are preserved in all Requested Languages.

2      Reference Architecture of One-to-Many Speech Translation

Figure 1 depicts the AIMs and the data exchanged between AIMs.

Figure 1 – Reference Model of One-to-Many Speech Translation (MST)

3      I/O Data of One-to-Many Speech Translation

The input and output data of the One-to-Many Speech Translation Use Case are given by Table 1:

Table 1 – I/O Data of One-to-Many Speech Translation

Input Descriptions
Input Selector Determines whether:
1.     The input will be in Text or Speech
2.     The Input Speech features are preserved in the Output Speech.
Language Preferences User-specified input language and translated languages
Input Speech Speech produced by human desiring translation and interpretation in a specified set of languages.
Input Text Alternative textual source information.
Output Descriptions
Translated Speech Speech translated into the Requested Languages.
Translated Text Text translated into the Requested Languages.

4      Functions of AI Modules of One-to-Many Speech Translation

Table 2 gives the functions of One-to-Many Speech Translation AIMs.

Table 2 – Functions of One-to-Many Speech Translation AI Modules

AIM Functions
Automatic Speech Recognition Recognises Speech
Text-to-Text Translation Translates Recognised Text
Input Speech Description Extracts Speech Features
Text-to-Speech (Features) Synthesises Translated Text adding Speech Features

5      I/O Data of AI Modules of One-to-Many Speech Translation

Table 3 gives the I/O Data of the AI Modules.

Table 33 – AI Modules of One-to-Many Speech Translation

AIM Receives Produces
Automatic Speech Recognition Input Speech Segment Recognised Text
Text-to-Text Translation Text input Translated Texts in the Requested Languages.
Input Speech Description Input Speech Speaker-specific Speech Descriptors.
Text-to-Speech (Features) 1.     Translated Texts

2.    Speech Descriptors (based on Input Selector)

 Speech Segments in the Desired Languages.

6     Specification of One-to-Many Speech Translation AIW, AIMs, and JSON Metadata

Table 4 – AIW, AIMs, and JSON Metadata

AIW and AIMs Name JSON
MMC-BST One-to-Many Speech Translation X
– MMC-ASR Audio Scene Description X
– MMC-TTT Text-to-Speech X
– MMC-ISD Input Speech Description X
– MMC-TTS Text-to-Speech X