Transformation de texte

Ingénierie des prompts avec l'API OpenAI

Fouad Trad

Machine Learning Engineer

Transformation de texte

Transforme le texte fourni pour créer un nouveau texte
Possède de nombreuses applications
- Traduction linguistique
- Ajustement du ton
- Amélioration de l'écriture

Image showing how we move from an original text to a transformed text having a very similar representation.

Traduction linguistique

Indiquer la langue d'entrée et de sortie dans le prompt

text = "XYZ Scooter is a cutting-edge electric scooter designed for urban 
adventurers. Its sleek design, long-lasting battery, and smart connectivity offer 
a seamless and eco-friendly way to navigate city streets."

prompt = f"""Translate the English text delimited by triple backticks to French:
           ```{text}```"""
print(get_response(prompt))

XYZ Scooter est une trottinette électrique de pointe conçue pour les aventuriers 
urbains. Avec son design élégant, sa batterie longue durée et sa connectivité 
intelligente, elle offre un moyen fluide et respectueux de l'environnement de se 
déplacer dans les rues de la ville.

Si la langue du texte est inconnue…

text = "Das Produkt ist wirklich schön, und der Preis ist fair."

Demander au modèle de préciser la langue du texte

prompt = f"""Tell me which language is the text delimited by triple backticks:
           ```{text}```"""

print(get_response(prompt))

The text delimited by triple backticks (```) is in the German language.

Traduction multilingue

Traduire simultanément vers plusieurs langues

text = "The product combines top quality with a fair price."
prompt =  f"""Translate the English text delimited by triple backticks to French, 
Spanish, and German:
```{text}```"""

French: Le produit allie une qualité supérieure à un prix équitable.
Spanish: El producto combina una calidad superior con un precio justo.
German: Das Produkt vereint Spitzenqualität mit einem fairen Preis.

Remarque : Les traductions modèles doivent être vérifiées si des clients sont impliqués

Ajustement du ton

Réécrire le texte dans un ton différent

text = "Hey there! Check out our awesome summer deals! They're super cool, and you 
won't want to miss them. Grab 'em now!"


prompt = f"""Write the text delimited by triple backticks using a formal and 
persuasive tone:
```{text}```"""

print(get_response(prompt))

Ajustement du ton

Dear [Recipient's Name],

We cordially invite you to explore our exceptional summer offers! These exclusive 
deals are truly extraordinary and should not be missed. Time is limited, so we 
encourage you to seize these incredible opportunities promptly.

[...]

Don't miss out on this chance! Take swift action to secure these remarkable summer 
deals before they expire. You won't find a more enticing selection of offers 
elsewhere.

[...]

Ajustement du ton : préciser le public cible

text = "Our cutting-edge widget employs state-of-the-art microprocessors and 
advanced algorithms, delivering unparalleled efficiency and performance for a wide 
range of applications."


prompt = f"""Write the text delimited by triple backticks to be suitable for
a non-technical audience:
```{text}```"""

print(get_response(prompt))

Our advanced widget uses the latest technology to provide exceptional efficiency 
and performance for a variety of uses.

Améliorations grammaticales et rédactionnelles

Corriger les erreurs d'orthographe, de grammaire et de ponctuation sans modifier les autres aspects

text = "Dear sir, I wanted too discus a potentiel opportunity for colabration. 
Pls let me know wen you're avaliable."

prompt = f"""Proofread the text delimited by triple backticks while keeping the 
original text structure intact:
           ```{text}```"""
print(get_response(prompt))

Dear sir, I wanted to discuss a potential opportunity for collaboration. Please 
let me know when you are available.

Améliorations grammaticales et rédactionnelles

Améliorer la clarté en modifiant la structure du texte

text = "Dear sir, I wanted too discus a potentiel opportunity for colabration. 
Pls let me know wen you're avaliable."

prompt = f"""Proofread and restructure the text delimited by triple backticks for 
enhanced readability, flow, and coherence:
           ```{text}```"""
print(get_response(prompt))

Dear sir,

I wanted to discuss with you a potential opportunity for collaboration. 
Please let me know your availability.

Thank you.

Transformations multiples

Demander plusieurs transformations simultanément -> prompts multi-étapes

text = "omg, I cant believe how awesome this product is! Its like the best thing ever! You guys gotta 
try it out!"

prompt = f"""Transform the text delimited by triple backticks with the following two steps:
             Step 1 - Proofread it without changing its structure
             Step 2 - Change the tone to be professional
           ```{text}```"""
response = get_response(prompt)

Step 1 - OMG, I can't believe how awesome this product is! It's like the best thing ever! You guys 
gotta try it out!
Step 2 - I am astounded by the exceptional qualities of this product. It represents a remarkable 
solution that surpasses expectations. I highly recommend trying it out, as its benefits are truly 
impressive.

Passons à la pratique !

Ingénierie des prompts avec l'API OpenAI