Spaces:
Sleeping
Sleeping
| import gradio as gr | |
| from langchain_community.llms import OpenAI | |
| from langchain.prompts import PromptTemplate | |
| import os | |
| from dotenv import load_dotenv | |
| load_dotenv() | |
| open_api_key = os.getenv('OPENAI_API_KEY') | |
| os.environ["OPENAI_API_KEY"] = open_api_key | |
| system_prompt_1 = """ | |
| You are an advanced AI assistant tasked with helping to develop a system that automatically transcribes texts into | |
| simplified languages, specifically FALC (Facile à Lire et à Comprendre) and "Leichte Sprache" (Simple Language). | |
| This system is intended to streamline the creation of accessible content for government websites, where the current | |
| manual process is time-consuming and limits the deployment of simplified language texts. | |
| Requirements: | |
| 1. Input and Output Formats: | |
| - Input Formats: The AI tool must accept input in Rich Text Format (.rtf) and Free Text (.txt). | |
| - Output Formats: The output should be generated in the same format as the input file (i.e., if the input is .rtf, the | |
| output should be .rtf, and if the input is .txt, the output should be .txt). | |
| - Default Output Language: The output language must match the language detected in the input file. | |
| 2. Language Simplification Rules: | |
| - The transcription must adhere to the rules of FALC and "Leichte Sprache," ensuring the content is simple, clear, and | |
| accessible. | |
| - Use simple vocabulary and avoid complex terms. | |
| - Construct short, straightforward sentences with one main idea per sentence. | |
| - Structure information clearly, using bullet points or numbered lists where applicable. | |
| - Incorporate illustrations, icons, or symbols to support textual information if needed. | |
| 3. Accessibility Standards: | |
| - The final solution must comply with accessibility standards to ensure content is usable by individuals with intellectual | |
| disabilities and other target groups. | |
| - Ensure that the output is compatible with screen readers and other assistive technologies. | |
| 4. Scalability and Efficiency: | |
| - The tool should significantly reduce the time required for the transcription process compared to the current manual | |
| methods. | |
| - It should be capable of handling large volumes of text efficiently to support widespread deployment across various | |
| government websites. | |
| 5. User Collaboration: | |
| - The tool should allow for revisions and feedback from collaborators affected by intellectual disabilities to ensure the | |
| output meets the necessary standards of FALC and "Leichte Sprache." | |
| Instructions for AI Development: | |
| Implement a language detection mechanism to identify the language of the input text. | |
| Develop natural language processing (NLP) models trained specifically on FALC and "Leichte Sprache" guidelines to | |
| accurately transcribe complex texts into simplified language. | |
| Ensure the models are capable of maintaining the context and meaning of the original text while simplifying its language. | |
| Include features for user feedback and revisions to refine and improve the transcriptions based on real-world use and | |
| collaborator input. | |
| Test the tool rigorously to ensure it meets accessibility standards and performs well across different types of content | |
| and input formats. | |
| Your goal is to create an AI tool that makes the process of generating FALC and "Leichte Sprache" content more efficient, | |
| scalable, and accessible, ultimately facilitating better communication and inclusivity on government websites. | |
| User Text: {text} | |
| transcribes text: """ | |
| system_prompt_2 = """Please translate the following text field content in english langauge. | |
| text: {text} | |
| """ | |
| def translate_text(file, text_input): | |
| llm = OpenAI() | |
| with open(file.name, 'r', encoding='utf-8') as f: | |
| file_text = f.read() | |
| template_1 = PromptTemplate(input_variables=["text"], template=system_prompt_1) | |
| prompt_1 = template_1.format(text=file_text) | |
| file_translation = llm(prompt_1) | |
| template_2 = PromptTemplate(input_variables=["text"], template=system_prompt_2) | |
| prompt_2 = template_2.format(text=file_translation) | |
| text_translation_op = llm(prompt_2) | |
| template_3 = PromptTemplate(input_variables=["text"], template=system_prompt_2) | |
| prompt_3 = template_2.format(text=file_text) | |
| text_translation_ip = llm(prompt_3) | |
| output_file_path = "translated_file.txt" | |
| with open(output_file_path, 'w', encoding='utf-8') as f: | |
| f.write(file_translation) | |
| return text_translation_ip, file_translation, text_translation_op, output_file_path | |
| iface = gr.Interface( | |
| fn=translate_text, | |
| inputs=[ | |
| gr.File(label="Upload Text File") | |
| ], | |
| outputs=[ | |
| gr.Textbox(label="transcribes content in english Translated of input content"), | |
| gr.Textbox(label="transcribes content"), | |
| gr.Textbox(label="transcribes content in english Translated of output content"), | |
| gr.File(label="Download Translated File Text") | |
| ], | |
| title="Text Transcribes", | |
| description="Upload a text file and provide a text input to translate the text using LangChain and OpenAI with predefined system prompts.", | |
| allow_flagging="never" | |
| ) | |
| iface.launch(debug=True) | |