A REVIEW OF AI-ASSISTED ACADEMIC SUBMISSION MANAGEMENT: WHATSAPP-BASED DEADLINE TRACKING AND CONTEXT-AWARE ASSIGNMENT SUPPORT

Academic submission work is distributed across classroom announcements, student group conversations, assignment instructions and individual study tasks. Important information such as due dates, deadline changes and required submission components can be difficult to track when embedded in long chat histories. This paper reviews AI and document-oriented information processing for academic submission workflows and presents an implementation-grounded case study of an AI Submission Assistant. The inspected Streamlit application imports exported WhatsApp conversations in text or ZIP form, parses message metadata, identifies task-related messages, resolves selected explicit and relative date expressions, categorizes messages by subject and stores messages and deadlines in SQLite. It also implements duplicate-message filtering, deadline urgency calculation, TF–IDF retrieval of relevant class messages and an optional Anthropic language-model interface for drafting assignment-oriented responses using retrieved context. The review relates these functions to research on large language models in academia, document-level information extraction, document understanding and hallucination in natural language generation. The analysis identifies limitations including rule-based date interpretation, imperfect event-to-assignment matching, dependence on announcement quality, the need for human verification and the absence of a demonstrated document-file export function in the inspected code. No quantitative accuracy or usability claims are made because a controlled evaluation dataset and measured results were not supplied. The paper concludes with a validation framework and development directions for improving reliability, transparency and usefulness.Academic submission work is distributed across classroom announcements, student group conversations, assignment instructions and individual study tasks. Important information such as due dates, deadline changes and required submission components can be difficult to track when embedded in long chat histories. This paper reviews AI and document-oriented information processing for academic submission workflows and presents an implementation-grounded case study of an AI Submission Assistant. The inspected Streamlit application imports exported WhatsApp conversations in text or ZIP form, parses message metadata, identifies task-related messages, resolves selected explicit and relative date expressions, categorizes messages by subject and stores messages and deadlines in SQLite. It also implements duplicate-message filtering, deadline urgency calculation, TF–IDF retrieval of relevant class messages and an optional Anthropic language-model interface for drafting assignment-oriented responses using retrieved context. The review relates these functions to research on large language models in academia, document-level information extraction, document understanding and hallucination in natural language generation. The analysis identifies limitations including rule-based date interpretation, imperfect event-to-assignment matching, dependence on announcement quality, the need for human verification and the absence of a demonstrated document-file export function in the inspected code. No quantitative accuracy or usability claims are made because a controlled evaluation dataset and measured results were not supplied. The paper concludes with a validation framework and development directions for improving reliability, transparency and usefulness.