AAU student paper recommendation system
Author
Clausen, Sebastian Vissing
Term
4. term
Publication year
2020
Pages
78
Abstract
Aalborg University’s Digital Project Library hosts more than 40,000 student papers that are rarely used. This project explores how to encourage students to read and be inspired by these reports by developing a hybrid information retrieval and recommendation system that suggests relevant papers using content-based filtering. It tests the use of university courses as inputs to signal interest and outlines a process involving a literature review, user-centered requirements and iterative user testing, data collection and item representation (including processing of long documents), candidate generation, ranking, and evaluation. The solution is designed as a microservice architecture with cloud-based components to support rapid iteration and modular replacement. A prototype showcases student paper recommendations and augmentation of paper metadata via several microservices. Early tests indicate promising potential for course-based inputs, while detailed effectiveness results are not available in the provided excerpt.
Aalborg Universitetets Digitale Projektbibliotek rummer mere end 40.000 studenterpapirer, som sjældent bliver brugt. Dette projekt undersøger, hvordan flere studerende kan motiveres til at læse og lade sig inspirere af disse rapporter ved at udvikle et hybridsystem for informationssøgning og anbefaling, der foreslår relevante papirer med content-baseret filtrering. Projektet afprøver brugen af universitetets kurser som input til at indikere interesse, og beskriver en proces med litteraturstudie, brugercentreret kravindsamling og gentagen brugertest, dataindsamling og item-repræsentation (herunder tekstbehandling af lange dokumenter), kandidatgenerering, rangering og evaluering. Løsningen er designet som en mikrotjenestearkitektur med cloud-baserede komponenter for at muliggøre hurtig iteration og udskiftning af moduler. En prototype demonstrerer anbefaling af studenterrapporter og udvidelse af rapportmetadata via flere mikrotjenester. De første forsøg viser lovende potentiale for kursusbaserede input, mens detaljerede effektevalueringer ikke fremgår af det tilgængelige uddrag.
[This apstract has been generated with the help of AI directly from the project full text]
Keywords