Multi-Source Research Assistant
Start with a research question, gather selected social sources, track collection and explore summaries through evidence-based follow-ups.
Turn conventionally structured Word papers into reusable question records and question-and-answer images, then extract by question ID, assemble papers and prepare question-count summaries.
No public application link is provided. Explore the case study and screenshots.
Open image
For educational organisations, teaching-material teams and question-bank maintainers, this service connects splitting Word-based papers, checking questions, selecting content and producing new papers. It is intended for material with explicit question numbers and agreed markers. A complete document can be organised into reusable question records and recombined for teaching or editorial needs.
Duplicate question numbers are checked before processing and identified for correction. Longer parsing work can be submitted as a background task with queued, processing, completed or failed status. Content staff can see whether work has finished and which task produced the result. Output quality still depends on the document following the agreed conventions; answers or explanations absent from the original are not promised as existing content.
Questions and answers can be selected by number from one or several DOCX sources and assembled into a new Word file. Missing requested questions produce information for further checking.
PDF conversion and question parsing are separate tasks, not a promise that any scanned document can automatically enter a question bank. This project focuses on processing and reusing existing papers, complementing Lingkee AI’s exercise-generation and answer-image marking scenarios without claiming an automatic connection between them.
I independently analyzed the requirements, designed the processing workflow and architecture, and developed, deployed and iterated on the APIs and document parsing services. I implemented cross-page question and answer extraction, preserved image references during exam assembly, and added task deduplication, exclusive job claiming and timeout recovery for batch processing.
Explore more work with similar capabilities or uses.
Start with a research question, gather selected social sources, track collection and explore summaries through evidence-based follow-ups.
Manage one-way PostgreSQL synchronization, full or selected-table backups, schedules and failed-job follow-up across environments, keeping data copies and maintenance records organized.
Prepare questions from a target role and personal experience, then practise with spoken questions, typed or dictated answers, follow-ups and detailed feedback on what to work on next.