MSc thesis proposals
Two parallel projects on Development of medical L2 language practice with a social robot (1) or virtual agent (2).
1. Development of medical L2 language practice with a social robot
Background
Swedish healthcare is ever more dependent on personnel born outside Sweden (27% of generalist medical doctors, 37% of specialists, 34% of nurses and 46% of dentists), many of whom (4,000 doctors and over 35,000 nurses and care assistants) have arrived in Sweden during the last ten years. To be certified to work in the medical sector, these persons need to master high school level Swedish, which requires extensive practice. This may be difficult to achieve within language courses focused on second language (L2) learning of Swedish in the medical context (SFM: Swedish for medical personel). In a joint effort between KTH and Karolinska Institutet, we therefore aim to develop realistic practice scenarios for L2 medical Swedish in which a doctor, nurse or dentist can interact with a simulated patient in the form of a social robot.
The work builds on previous work on using the social robot Furhat to practice a) L2 Swedish (see https://www.kth.se/profile/engwall) and b) clinical reasoning in meetings with patients (see https://news.ki.se/ai-driven-robot-patient-can-train-medical-students-in-clinical-reasoning). Additional descriptions of the long-term goals and methods of the efforts are available on request.
Target
The MSc thesis project aims at extending/modifying existing implementations in the social robot to focus on practice of L2 medical Swedish in meetings with a simulated patient, evalaute the effectiveness of the implementation in a field trial and provide suggestions for future development of scenarios and robot implementation.
Description of the assignment
Tasks to be performed in collaboration with the supervisor and other partners:
- Define and implement a set of patient meeting scenarios relevant for practising medical Swedish
- Define intended learning outcomes (ILO) of the practice (e.g., vocabulary use)
- Define and implement pre- and post-test surveys to assess ILO and investigate influencing factors
- Conduct a field-test with the social robot at KI and/or VuX Huddinge with students in collaborating SFM courses
- Analyse quantitative learning outcomes and qualitative learner feedback
In addition, tasks related to the MSc thesis course (https://canvas.kth.se/courses/61869) are to be performed independently.
Education
The project is to be performed as a MSc thesis at the EECS school, KTH, signifying that the student should primarily belong to a master program at or affiliated with EECS, including, but not limited to, in Machine Learning, Computer Science, Interactive Media Technology, Systems Control and Robotics, Embedded Systems, ICT Innovation, Industrial Management and Engineering sepcialising in Computer Science.
The thesis student should have a strong background in programming and statistics. Relevant courses for the project include: DD2413 Social Robotics, DT2151 Project in Conversational Systems, DD2417 Langauge Engineering , DT2140 Multimodal Interaction and Interfaces, DT2119 Speech and Speaker Recognition, DT2112 Speech Technology, DD2601 Deep Generative Models and Synthesis, DD2380 Artificial Intelligence, and other courses in machine learning and interactive systems.
Number of students: 1 or 2
Start date for the Thesis project: Flexible
Estimated time needed: 20 weeks
Note: A parallel project, in which a computer-animated screen-based agent is used as the simulated patient, is also advertised. The projects are closely connected and may collaborate to achieve synergetic effects.
Contact person and/or supervisor
Olov Engwall, Professor in Speech communication, engwall@kth.se
======================================================
2. Development of medical L2 language practice with a virtual patient
Background
Swedish healthcare is ever more dependent on personnel born outside Sweden (27% of generalist medical doctors, 37% of specialists, 34% of nurses and 46% of dentists), many of whom (4,000 doctors and over 35,000 nurses and care assistants) have arrived in Sweden during the last ten years. To be certified to work in the medical sector, these persons need to master high school level Swedish, which requires extensive practice. This may be difficult to achieve within language courses focused on second language (L2) learning of Swedish in the medical context (SFM: Swedish for medical personel). In a joint effort between KTH and Karolinska Institutet, we therefore aim to develop realistic practice scenarios for L2 medical Swedish in which a doctor, nurse or dentist can interact with a simulated patient in the form of an animated screen-based agent.
The work builds on previous work on using screen-based agents and social robots to practice a) L2 Swedish (see https://www.kth.se/profile/engwall) and b) clinical reasoning in meetings with patients (see https://news.ki.se/ai-driven-robot-patient-can-train-medical-students-in-clinical-reasoning). Additional descriptions of the long-term goals and methods of the efforts are available on request.
Target
The MSc thesis project aims at modifying existing implementations for a social robot to, firstly, transfer the practice to a virtual agent on screen, and, secondly, focus on practice of L2 medical Swedish in meetings with a simulated patient, evalaute the effectiveness of the implementation in a field trial and provide suggestions for future development of scenarios and agent implementation.
Description of the assignment
Tasks to be performed in collaboration with the supervisor and other partners:
- Define and implement a set of patient meeting scenarios relevant for practising medical Swedish
- Define intended learning outcomes (ILO) of the practice (e.g., vocabulary use)
- Define and implement pre- and post-test surveys to assess ILO and investigate influencing factors
- Conduct a field-test with the virtual agent system at KI and/or VuX Huddinge with students in collaborating SFM courses
- Analyse quantitative learning outcomes and qualitative learner feedback
In addition, tasks related to the MSc thesis course (https://canvas.kth.se/courses/61869) are to be performed independently.
Education
The project is to be performed as a MSc thesis at the EECS school, KTH, signifying that the student should primarily belong to a master program at or affiliated with EECS, including, but not limited to, in Machine Learning, Computer Science, Interactive Media Technology, Systems Control and Robotics, Embedded Systems, ICT Innovation, Industrial Management and Engineering sepcialising in Computer Science.
The thesis student should have a strong background in programming and statistics. Relevant courses for the project include: DD2413 Social Robotics, DT2151 Project in Conversational Systems, DD2417 Langauge Engineering , DT2140 Multimodal Interaction and Interfaces, DT2119 Speech and Speaker Recognition, DT2112 Speech Technology, DD2601 Deep Generative Models and Synthesis, DD2380 Artificial Intelligence, and other courses in machine learning and interactive systems.
Number of students: 1 or 2
Start date for the Thesis project: Flexible
Estimated time needed: 20 weeks
Note: A parallel project, in which a social robot is used as the simulated patient, is also advertised. The projects are closely connected and may collaborate to achieve synergetic effects.
Contact person and/or supervisor
Olov Engwall, Professor in Speech communication, engwall@kth.se
