Speech Mentoring

Inspiration

I often give speeches and presentations at work and in my Toastmasters public speaking club. I need a speech mentor who could look at my speech, tell me what it could be improved, and even demonstrate how to do it.

This Speech Mentoring app does just that.

What it does

You can enter a draft speech of any length, and the criteria to be used to evaluate the speech.

Then an evaluation of the speech will be generated, by highlighting the strengths of the speech, and potential areas of improvement to make it a better speech.

The user will also be presented with an improved version of the speech, by taking the suggestions given in the speech evaluation above into consideration.

Finally, the user can ask the app any question on the speech. For example, he or she can ask "What have you changed in my speech?"

How I built it

The app is built using widgets provided in PartyRock, careful prompting and repetitive testing.

All types of widgets have been used, including "User Input" for the user to enter the speech text, "Static Text" for notices, "Text Generation" for generating the speech evaluation and improved speech, "Image Generation" for generating the representative image, and "Chatbot" for the interactive chat assistant.

For prompting, these have been used:

• Asking the AI to pretend to be a speech coach or mentor.
• Specifying the format or limiting the kind of info generated. In particular, the speech evaluation is typical of the style for evaluating a speech in the Toastmasters organization.
• Making each widget run sequentially after the previous widget has completed running to avoid error and provide user the info on a logical manner.
• Ensuring the chatbot take into account all info provided in other widgets to be able to answer questions on the info given.

Challenges I ran into

The widgets start responding as soon as I start typing in the input box. This might lead to errors in the subsequent widgets to run, and confuse the user using the app. I think an "Enter" or "Submit" button could have been useful. Or at least, the widgets could start responding only after certain lag time such as 3 seconds.

Accomplishments

I'm proud to have made this app to act as an impartial speech mentor to help me improve my speeches. It is also helping my Toastmasters members and proteges on improving their individual speeches.

What I learned

I've learned to create something useful for practical use with prompt engineering, careful prompting and sequencing, and having the users in mind when developing the app. I've learned to replicate the function of evaluating a speech in Toastmasters using the app.

What's next for Speech Mentoring

I wish to program using AWS Bedrock, Python and other web technologies to make it a more official and user-friendly public-facing app.

Built With

• partyrock