"Automatic scoring of spoken language proficiency”, finanziata da Fondazione Bruno Kessler - FBK
Description: Automatic scoring of speech is gaining growing importance since an increasing number of students start to learn a second language at young age and therefore this demands techniques for the objective assessment of language proficiency. As such, many approaches have been proposed to develop suitable models for spoken answers and in particular it is of great interest to identify and extract features representing specific linguistic competences such as grammar correctness, fluency, quality of pronunciation. The aim of the Phd project is to investigate approaches for second language proficiency assessment using deep learning approaches. The task is very challenging since diverse categories of errors have to be recognized and classified (e.g. disfluencies, pronunciation errors, grammar errors, etc.). Moreover, from the automatic speech recognition perspective, recognition of child non-native speech still represents an open issue. The work will address the study of (a) suitable acoustic and linguistic features to develop the scoring tool and (b) end-to-end approaches to directly derive from the spoken answers the targeted proficiency indicators.
Skills: Good scientific background in speech/language processing;familiarity with speech recognition tools and popular deep learning platforms (e.g. tensorflow, pytorch).
Responsible of the Project: Marco Matassoni - matasso [at] fbk.eu
"Offline and online social relations in schools", finanziata da Fondazione Bruno Kessler - FBK
Description: The Phd project can evolve along either one of the two research lines described below.
The first research line has the goal to build dynamic models of face-to-face and proximity interactions among children in schools and to quantify their effect on performance, social and friendship relationships, motivation, integration, and other social aspects. Despite the large amount of communication means made available by our modern societies, direct face-to-face interactions between individuals remain an essential ingredient of our daily behaviors. Indeed, face-to-face interactions contribute to shape the friendship and collaboration networks, to define important channels of information propagation and opinion formation, to build social hierarchies and behavioral norms.
The second research line has the goal to test the hypothesis that an integration of Smart Classrooms with technologically-enhanced learning and exposure to social media, can increase academic performance across all tiers of education, from elementary level, to middle school, to university. While we live in an age where people are exposed to online content frequently, individuals differ in the rate of consumption of online content, and the usage of web platforms (i.e. social media), for work and leisure. Moreover, this pattern of consumption possibly changes according to different stages of development, which, in turn, may lead to differences in the effectiveness of technologically-facilitated learning that is increasingly employed in classrooms across various levels of education. However, little is known of how social media consumption patterns influence the effectiveness of technologically-driven pedagogy in influencing academic performance. A greater understanding of the mechanisms underlying this interaction will allow us to fully utilize technological advancements in a manner that optimally benefits learning.
Skills: The ideal candidate for the first line of research has a background in Computer Science, Cognitive Science, or Complex Systems and a strong interest on behavioral analyses and on designing field studies and interventions. Programming skills for data analysis (Python or similar) and knowledge of Machine Learning techniques are required.
The ideal candidate for the second line of research has a background in Computer Science, Cognitive Science, or Complex Systems and a strong interest on behavioral analyses and on designing field studies and interventions. Programming skills for data analysis (Python or similar) and knowledge of Machine Learning techniques are a plus.
Responsibles of the Project: Massimo Zancanaro - zancana [at] fbk.eu | massimo.zancanaro [at] unitn.it
"AI methods for tracing and optimizing the intervention process in autism", finanziata da Fondazione Bruno Kessler - FBK
Description: This project will explore the design of Artificial Intelligence solutions, and in particular of interpretable deep neural networks to elucidate the fundamental mechanisms of intervention process with patients with Autism Spectrum Disorders, in particular with children. The aim is to create real time support systems that can collaborate with experts by offering a quantitative information about key elements of attuning, reciprocity and general response in a general but structured therapeutic setting. This work will include the preparation of a novel data framework for managing video of sessions and their annotation, as a basis for training machine learning models. In particular we expect to provide new tools for observation in standardized or naturalistic set-ups, including the integration of video material, phenotype data, psychological tests. The ideal candidate will be strongly motivated in developing skills in data science and machine learning as well as a keen interest in the clinical aspects and application opportunities, e.g. offering support to caretakers in developing optimal strategies (e.g. interactive play). The PhD position is offered withih the TRAIN joint PhD program of UniTN/FBK, which aims at an interdisciplinary research in the domain of Autistic Spectrum Disorders. Joint positions will be offered at DIPSCO and Fondazione Kessler, Data Science Area, Predictive Models for Biomedicine & Environment Lab (FBK/MPBA http://mpbalab.fbk.eu/).
Skills: Skills requested include familiarity with quantitative data analysis, proven basic working knowledge of data science tools (e.g. Python), experience in analytics of developmental data combined with proven experience in a clinical setting, and knowledge of clinical psychology.
Responsibles of the Project: Cesare Furlanello - furlan [at] fbk.eu
The PhD student awarded the scholarship financed by FBK is obliged to maintain confidentiality in regard to the disclosure and use of any information, data, software, discovery, invention, idea, method, process (in any format including the source code) or other knowledge discovered, conceived, developed and/or implemented within the research activities financed relatively not only to the object of the PhD scholarship awarded to the PhD student, but also to possible changes made to the object of the grant as agreed with FBK.