Summary and Ressources

We hope you found this lesson useful and informative. You now know how to correct a mispronounced word by changing its spelling or using phonetic writing. You can change the prosody with the punctuation and use tags to adjust various parameters such a speed, pitch and volume. You can add pauses, correct emphasis and active prompts to make the sentences more expressive, and even use expressive tags to change the mood they express. Your robot’s speech may still not be perfect, but it has certainly improved, and will keep doing so as you get more experienced in the art of tweaking!

Showing the result

You can find our tweak suggestion below, but it is only a suggestion, it does not mean that yours is wrong if it is different.

Sound file for One does not simply walk into Mordor.

Source code

You can clone the following code sample/project, which is the resulting code of this lesson.

\rspd=90\ \vct=80\ One \pau=20\ does not \emph=1\ \rspd=70\ sim\rspd=90\ply \bound=N\ walk into \emph=0\ \rspd=70\ \vct=70\ moar\rspd=90\dor.

Use Case Recap

As you’ve seen throughout the various examples, basically any use case that requires the robot to speak is applicable for this lesson, which means almost all of them! Indeed, it is important to have the robot speak clearly and eloquently when: * Welcoming clients in shops and malls * Making speeches on stage for various events * Showcasing a given brand * Answering specific Q&A

Another representative use case is the dialog part of the Basic Channel, called Basic Dialog: a non-proactive dialog enabling the robot to give technical informations, answer small talk and questions about himself, move on command, and launch applications.

And obviously any previously identified business use cases, as they all involve speech at one point : Pepper acts as a Host, a Brand Ambassador, an Advisor, a Sales Associate, a Promoter, a Service Provider, a Survey Conductor or an Entertainer.

Glossary

Basic Dialog: a non-proactive dialog enabling the robot to give technical informations, answer small talk and questions about himself, move on command, and launch applications.

Basic Channel: a set of essential applications to make NAO or Pepper operational (all Basic Dialog applications, and technical ones such as Fall Recovery or the Aldebaran soundset).

Emphasis: stress given to a word or a phrase to indicate particular importance.

Heteronym: a word that has a different pronunciation and meaning from another word but the same spelling.

Pitch: the degree of highness or lowness of a tone.

Prosodic boundary: defines a prosodic unit (a segment of speech that occurs with a single prosodic contour, aka pitch and rhythm contour).

Prosody: the patterns of stress and intonation in a language. You can tweak Pepper’s and NAO’s voices to make them sound more natural.

QiChat: the qiChat is a Softbank Robotics proprietary language forked from ChatScript, a combination of Natural Language engine and dialog management system designed initially for creating chatbots. QiChat provides specific robotic functions to manage both inputs (words and human actions) and outputs (the robot's speech and movements).

Sentence break: equivalent to an end punctuation.

Tweak: A fine adjustment to a mechanism or system, here the process of altering the robot’s voice.

Uncanny valley: The uncanny valley is an unsettling feeling people experience when androids (humanoid robots) and audio/visual simulations closely resemble humans in many respects but are not quite convincingly realistic.

List of the active prompts recorded by the vocal talents (PDFs, 2018): * French (FRF) * Joyful_FRF * Didactic_FRF * Neutral_FRF

Discovering QiChat

Linking QiChat and code

Integrating a chatbot: DialogFlow