Telegram Medical MCQ Matching Bot
Budget: $750 – $1,500 USD
I need a Telegram bot that listens to questions posted in a channel or group and instantly finds the closest‐matching multiple-choice items from nine medical textbooks I own. All reference questions are in PDF files, so the first task is to parse those books, pull out each stem, options and any author-provided explanation, then store everything in a searchable index.
Once the database is ready, the bot should accept an “imperfect” MCQ—anything from a photo-to-text snippet to a paraphrased plain-text question—run a similarity check and return every match whose score exceeds a configurable threshold. Each hit must show:
• the similarity percentage
• the full textbook question and its options
• any explanation that is actually present in the book (if none exists, the bot simply skips that part)
A single top match isn’t enough; the user should see all qualifying matches, ordered by score, so they can judge which one is truly relevant.
Key points to cover in your build
• Accurate PDF text extraction, including tables or special medical symbols where they appear
• Robust fuzzy-matching / NLP logic capable of coping with typos and partial wording
• Easy deployment to an existing Telegram channel, with commands for start, help, set-threshold and reindex
• Clean, commented code (Python + aiogram, Node, or another mainstream stack you prefer) plus a brief setup guide so I can run the bot on my VPS or a cloud function
Acceptance will be based on a live demo in my test channel, correct similarity scoring on a sample set I provide, and clear hand-over of all source code and instructions.
Once the database is ready, the bot should accept an “imperfect” MCQ—anything from a photo-to-text snippet to a paraphrased plain-text question—run a similarity check and return every match whose score exceeds a configurable threshold. Each hit must show:
• the similarity percentage
• the full textbook question and its options
• any explanation that is actually present in the book (if none exists, the bot simply skips that part)
A single top match isn’t enough; the user should see all qualifying matches, ordered by score, so they can judge which one is truly relevant.
Key points to cover in your build
• Accurate PDF text extraction, including tables or special medical symbols where they appear
• Robust fuzzy-matching / NLP logic capable of coping with typos and partial wording
• Easy deployment to an existing Telegram channel, with commands for start, help, set-threshold and reindex
• Clean, commented code (Python + aiogram, Node, or another mainstream stack you prefer) plus a brief setup guide so I can run the bot on my VPS or a cloud function
Acceptance will be based on a live demo in my test channel, correct similarity scoring on a sample set I provide, and clear hand-over of all source code and instructions.