NordicTranslator-GPT-PALM

Project Overview

This project focuses on translating article titles and descriptions from English to Norwegian using advanced language models. We leverage two leading models: GPT (versions 3.5 and 4) and Gemini Pro (including also Bison and Unicorn models), to achieve high-quality translations focusing on eCommerce platforms.

Methodology

• Input: English text (titles and descriptions).
• Processing: Utilizing GPT-3.5/GPT-4 and Gemini Pro models through their respective APIs.
• Output: Translated content in Norwegian, formatted in JSON.

Useful References:

• ChatGPT API Docs
• ChatGPT API Pricing
• ChatGPT API Rate Limits
• Gemini API Docs
• Gemini API Pricing
• Gemini API Rate Limits

Interaction:

• Playground from OpenAI
• Generative AI Studio from Google

Features

• Bilingual Translation: Converts English text into Norwegian while preserving the original context and nuances.
• Dual Model Approach: Uses the strengths of both GPT and PALM models for translation.
• JSON Output: Easy integration and usage in various applications due to JSON format.

Dependencies

To run this project, you need to install the following Python libraries:

!pip install openai
!pip install python-dotenv
!pip install "shapely<2.0.0"
!pip install python-telegram-bot
!pip install google-cloud-aiplatform==1.35.0

Setup and Authentication

1. ENVIRONMENT FILE

   Before using the OpenAI GPT API, you need to set up your environment with the necessary API key. Follow these steps to get started:

   Create a .env file in your project's root directory. This file will store your OpenAI API key securely. The content of the .env file should be as follows:

   OPENAI_API_KEY=your_api_key_here

   Replace your_api_key_here with your actual OpenAI API key.

2. OpenAI GPT API

   Set your OpenAI API key:

   import os
   from openai import OpenAI
   from dotenv import load_dotenv
   
   # Load environment variables from .env file
   load_dotenv()
   
   # Initialize the OpenAI client with your API key
   client = OpenAI(
    api_key=os.environ.get("OPENAI_API_KEY")
   )

3. Google Gemini Pro API using Google Colab

   Initialize Vertex AI with your Google Cloud project details:

   from google.colab import auth as google_auth
   google_auth.authenticate_user()

   import vertexai
   from vertexai.language_models import TextGenerationModel
   from google.api_core.exceptions import ResourceExhausted
   from vertexai.preview.generative_models import GenerativeModel, Part
   
   vertexai.init(project="your-gcp-project", location="your-gcp-location")
   
   parameters = {
    "candidate_count": 1,
    "max_output_tokens": 2048,
    "temperature": 0.9,
    "top_p": 0.8,
    "top_k": 40
   }
   
   # model = TextGenerationModel.from_pretrained("text-bison")
   # model = TextGenerationModel.from_pretrained("text-unicorn@001")
   model = GenerativeModel("gemini-pro") ### from the Gemini models family

Using the APIs

To use both APIs, you will be sending prompts to each model and then processing their outputs.

GPT API Usage

1. Create a prompt for the GPT model.

2. Send the prompt to the GPT API and retrieve the response:

   prompt = "your-prompt-for-gpt"
   title_response = client.chat.completions.create(
       messages=[{"role": "user","content": prompt}],
       # max_tokens=600,
       # temperature=0.9,
       model="gpt-4-0125-preview"
   )
   json_output_gpt = title_response.choices[0].message.content

3. Bonus: Calculating number of tokens in a prompt.

   The function num_tokens_from_string takes a string (the prompt) and an encoding name. It first tries to get the encoding directly using tiktoken.get_encoding.

   import tiktoken
   
   def num_tokens_from_string(string: str, encoding_name: str) -> int:
       """
       Returns the number of tokens in a text string according to a specified encoding.
   
       Parameters:
       string (str): The text string to be tokenized.
       encoding_name (str): The name of the encoding or model to use for tokenization. 
           Encoding values: cl100k_base, r50k_base, p50k_base.
           Model values: gpt-3.5-turbo, gpt-3.5-turbo-1106, gpt-3.5-turbo-0125, gpt-4, gpt-4-1106-preview, gpt-4-0125-preview
   
       Returns:
       int: The number of tokens in the text string.
       """
       try:
           # Attempt to get the encoding directly
           ### cl100k_base, r50k_base, p50k_base as encoding_name
           encoding = tiktoken.get_encoding(encoding_name) 
       except ValueError:
           # If direct retrieval fails, attempt to map the model name to an encoding
           try:
               ### gpt-3.5-turbo, gpt-3.5-turbo-0125, gpt-4, gpt-4-1106-preview, gpt-4-0125-preview as encoding_name
               encoding = tiktoken.encoding_for_model(encoding_name)
           except KeyError as e:
               # If mapping also fails, raise the error
               raise e
   
       # Encode the string and return the number of tokens
       l_tokens = encoding.encode(string)
       num_tokens = len(l_tokens)
       return num_tokens, l_tokens

Note

Don't forget to install tiktoken by running pip install tiktoken.

Gemini Pro API Usage

1. Set parameters for the Gemini model.
2. Send the same or a modified prompt to the PALM API and retrieve the response:

   prompt = "your-prompt-for-gemini"
   response_gemini = model.generate_content(prompt, generation_config=parameters)
   json_output_palm = response_gemini.text

Combining the Outputs

After receiving responses from both APIs, you can compare, contrast, or combine the outputs as needed for your project. This might involve additional processing or data manipulation based on your specific requirements.

Script Overview

The main Python script performs several key functions:

1. GPT Notebook
   • Translation and Generation: It takes English titles, sends them to the LLM API, and retrieves Norwegian translations.
   • Data Handling: Combines CSV files, processes data, and handles errors efficiently.
   • Custom Functions: Includes functions like replace_mayusculas_noruego for specific text manipulations in Norwegian.
2. PALM (Gemini Pro) Notebook
   • Translation and Generation: Takes titles and descriptions, sends them to the PALM API (models: text-bison, text-unicorn and gemini-pro), and retrieves translations.
   • Vertex AI Integration: Uses Vertex AI for managing model interactions.
   • Custom Functions: Includes specific text manipulations and data handling routines.

Acknowledgements

• GPT-3.5 and GPT-4 by OpenAI
• Gemini Pro (also including Bison and Unicorn models) by Google

Contributing

Contributions to NordicTranslator-GPT-PALM are welcome! Here's how you can contribute:

Fork the repository on GitHub. Create a new branch for your proposed feature or fix. Commit your changes with an informative description. Push your branch and submit a pull request. We appreciate your input!

License

NordicTranslator-GPT-PALM is open source software licensed under the MIT License. See the LICENSE file for more details.