EKET 0.1.4

EKET: Educational Knowledge Extraction Toolkit

EKET

EKET is a package for loading the documents(.xls, .xlsx, .pdf, .csv, .md, youtube-links, .pptx, .txt, .json) and converting them into the chunks which has meaning for the further instructions from the user.

We can generate quiz, answers & explanations and answers for topic-based questions.

The Folder Structure

EKET/                         # Root directory of the project
├── EKET/                     # Main Python package (importable as EKET)
│   ├── __init__.py           # Initializes EKET as a package
│   ├── answer.py             # Answering the created quiz questions
│   ├── clean.py              # Functions for cleaning input/output data
│   ├── create.py             # Utilities for creating or generating quiz
│   ├── ingest.py             # Handles document ingestion (loading and chunking files)
│   ├── query.py              # Answering the specific question from the user
│   ├── utils.py              # Helper functions used across multiple modules
│   ├── data_ingest/          # Subpackage for modular ingestion logic
│   ├── evaluate_quiz/        # Subpackage to evaluate quiz answers, scoring, feedback
│   ├── generate_answer/      # Subpackage to generate answers
│   └── generate_quiz/        # Subpackage to create quiz questions from source material
├── example_usage/            # Example scripts or notebooks showing how to use the package
├── tests/                    # Unit and integration tests for all modules
├── README.md                 # Project description, usage instructions, and documentation
├── requirements.txt          # Requirements
└── setup.py                  # Setup script for packaging and installing the project

Installation

Step 0: Set Environment

First you should set your gemini api key into your environment system

setx TUTOR_API_KEY "GEMINI_API_KEY"
setx EMBEDDING_MODEL "gemini-embedding-001"

Step 1: Windows Installation Note (C++ Build Tools Required)

Some dependencies of EKET (such as PyMuPDF) include native C/C++ extensions. On Windows, if a precompiled wheel is not available for your Python version, pip may attempt to build the package from source.

If you encounter an error similar to:

please install Visual Studio C++ Build Tools: https://visualstudio.microsoft.com/visual-cpp-build-tools/

During installation, make sure to select:

• C++ build tools
• MSVC v143 (or latest)
• Windows 10/11 SDK

Step 3: HTML Rendering

For HTML rendering

pip install playwright
playwright install

Step 4: Pip Install

After installation, restart your terminal and rerun:

pip install EKET

How to Use

• For debugging and testing phase, in the current working place, you need to do files like below:

python example_usage/ingest.py --file "path/to/file.png" # supported files above.

• for YouTube:

python example_usage/ingest.py --youtube "Youtube-URL"   # YouTube URL  

• After starting example_usage/ingest.py, there will be folders that created in the current folder named chroma which is the vectorized database for chunks, saved_data which contains context_language.json formatted data in it.

• Then, if you want to generate quiz (multiple-choice and open-ended), you are going to initialize example_usage/generate_quiz.py. But it has some different kind of logic. Let me explain:

  • If you directly initialize the example_usage/generate_quiz.py, it will create the questions from all the documents the user is provided.
  • If you initialize the example_usage/generate_quiz.py after initializing example_usage/query_answer.py, then the questions will be much more relevant to the asked question.

• To initialize the example_usage/query_answer.py, you should use it in the terminal that:

python example_usage/query_answer.py --query "Your question here..."

• Lastly, example_usage/evaluation.py which indicates the evaluation of the solved generated quiz by the user. It informs the user how good one did and explains the question to make a better understanding.

Example Usage Scenario

Here is the simple designed FlowChart of the EKET-package:

1. Ingesting the documents that want to be studied on: For Files:

python example_usage/ingest.py --file "/path/to/file"

For YouTube:

python example_usage/ingest.py --youtube YOUTUBE-URL-LINK

It will yield context_language.json file. Other operations will be depend on this file. (e.g. creating questions, generating quiz).

2. Do you want to ask a specific question? If yes:

python example_usage/query_answer.py --query "Your Question" --input "Your context_language.jon" --output "Output Folder"

And it will yield context_question_answer.json. If it is created, the quiz generation(next step) will depend on this question.

3. Creating the quiz:

python example_usage/generate_quiz.py --input "Path to JSON file (query_answer.json or context_language.json)" --output "Output folder to save the quiz"

It will yield generated_quiz.json.

4. Evaluating the solved quiz:

python example_usage/evaluation.py --input "generated_quiz.json file path" --output "Output folder where evaluation results will be stored"

It will yield evaluation_results.json. It will create this file from generated_quiz.json and answers of the user.

• Optional:

python example_usage/summarize.py --input "Your context_language.json file" --output "Output Folder"

It will yield combined_summary.json. It has a summarization of context_language.json.

The example_usage folder contains the examples of the output files. While it all returns a .json file, you can also manipulate the files to the attributes we've assigned them to.

License

This project is licensed under the GNU General Public License v3.0 or later (GPL-3.0-or-later). See the LICENSE file for details.