Creating MCQs from your PDFs by leveraging LLMs.
Exminer aims to ease the tedious process of test creation. It utilizes RAG-based LLMs to generate questions on the topics you specify from the PDFs you provide.
- Processes the PDF and saves it as chunks into a database.
- Generates 5 MCQs based on the specified topic when a user enters a query.
- Allow users to choose the number of questions.
- Enhance the pre-prompt to generate more relevant questions.
- Improve chunk saving based on topics for better indexing and question generation.
- Automatically extract PDF topics to offer users options to choose from different topics.
flowchart TD
subgraph User Interface
PFDup[PDF Upload]
listoftitles[List of Titles in the PDF] -->selectedtitles[User: selected titles <5]
selectquestioncharecter[User: level of test taker + taxonomy + question type + question standard practice]
questionandrationale
end
subgraph Preprocessing
PFDup--> PDF[PDF knowledge source] --best: spliting [chunk_size = 5000, chunk_overlap = 1000]--> chunks[N * chunks]
chunks --vectore store--> database(vector database)
chunks --LLM based cleaning-->cleanedchunk[cleaned chunk]
cleanedchunk --LLM-based title & summary generator--> chunkmetadata[chunk's title and summary]
chunkmetadata -->tiab_list[tiab list]--LLM based summary-->alltiab[all, tilte & abstract]
end
subgraph Question generation
tiabselected[Title and Abstract of selected title] --RAG--> retrived[retrived similar chunks]
retrived ---> finalprompt
finalprompt[title+context+command]--LLM-->questions[3 questions]
selectquestioncharecter-->prompt_of_question[Q charectar prompt using taxonomy]
prompt_of_question --> finalprompt
end
selectedtitles-->tiabselected
alltiab --> listoftitles
subgraph Fact Check
questions-->Factcheck[Each option location and rationale of truth or false] --fail--> excludeQ[excludequestion]
end
Factcheck--pass-->questionandrationale[Question, Options, and rationales]