Overview Si assicuri di leggere attentamente le informazioni relative a questa occasione prima di inviare la sua candidatura. ALMA MATER STUDIORUM - UNIVERSITA' DI BOLOGNA – DIPARTIMENTO DI INFORMATICA – SCIENZA E INGEGNERIA Offer Description The research activity will be organized into phases aimed at developing a system for the automatic extraction and classification of exercises from school textbooks. The first phase will focus on the construction of a multilingual dataset (French, English, and Italian) consisting of school textbooks in PDF format, either native or generated through OCR, together with the corresponding annotated HTML versions based on the project's data model. This will be followed by an analysis of the state of the art in Large Language Models, computer vision, and multimodal approaches, with particular attention to hybrid solutions combining textual, visual, and structural information. Based on these findings,



a software architecture will be defined and progressively implemented through iterative prototyping and systematic evaluation, gradually extending the system's ability to recognize, extract, and classify an increasing range of exercise types. xrdztoy Responsibilities The successful candidate will contribute to building a system for automatic extraction and classification of textbook exercises, including data collection, dataset construction, model analysis, software architecture definition, and iterative prototyping with systematic evaluation. Qualifications and Eligibility Eligibility of fellows: country/ies of residence: All Eligibility of fellows: nationality/ies: All Notes Number of offers available: 1 #J-18808-Ljbffr

📌 Artificial Intelligence Models and Tools for Extracting and Classifying Exercises from School T[...] (Italia)
🏢 Altro
📍 Italia