Article type
Abstract
"Background
IDESR is a repository of protocols and published systematic reviews in Education. Initially focused on language education solely, it now aims to broaden its scope and encompass reviews across all topics of education. Maintaining such a diverse database up to date as well as expanding the range of topics that IDESR covers poses a great challenge and requires a lot of manual work and resources. In this context, machine learning techniques for identifying potentially relevant research publications present a promising solution to streamline this process.
Objectives
Applying machine learning to maintaining living reviews or identifying potentially relevant research is becoming a common practice in evidence synthesis. However, these novel approaches are usually used for identifying publications on a given topic or a series of closely related questions, and their potential in navigating a wider range of subjects within a broad field is less explored. Our aim is to investigate whether machine learning-based approaches to systematic searching and screening literature could be a more sustainable way to maintain, update and expand IDESR than relying on manual submissions from research producers.
Methods
All records from IDESR will be used as a seeding sample to identify other potentially relevant publications on the same or similar topics in education using EPPI-Reviewer and its integration with OpenAlex. All references will be screened on relevance to education using machine learning-based prioritisation and classification. All identified references will be included to the database and their topics will be explored and classified to investigate whether the strategy leads to the growth in both the number of items and the breadth of topics.
Expected Results and Conclusions
Our work showcases the potential of using machine learning approaches to identify research publications on a wide range of topics in education in the context of limited resources and funding. We expect that although machine learning applied to searching and screening potentially relevant publications cannot fully substitute manual work, it will lead to significant reductions in the amount of human labour and will allow IDESR to grow and develop serving as an up-to-date source of knowledge to research producers and consumers."
IDESR is a repository of protocols and published systematic reviews in Education. Initially focused on language education solely, it now aims to broaden its scope and encompass reviews across all topics of education. Maintaining such a diverse database up to date as well as expanding the range of topics that IDESR covers poses a great challenge and requires a lot of manual work and resources. In this context, machine learning techniques for identifying potentially relevant research publications present a promising solution to streamline this process.
Objectives
Applying machine learning to maintaining living reviews or identifying potentially relevant research is becoming a common practice in evidence synthesis. However, these novel approaches are usually used for identifying publications on a given topic or a series of closely related questions, and their potential in navigating a wider range of subjects within a broad field is less explored. Our aim is to investigate whether machine learning-based approaches to systematic searching and screening literature could be a more sustainable way to maintain, update and expand IDESR than relying on manual submissions from research producers.
Methods
All records from IDESR will be used as a seeding sample to identify other potentially relevant publications on the same or similar topics in education using EPPI-Reviewer and its integration with OpenAlex. All references will be screened on relevance to education using machine learning-based prioritisation and classification. All identified references will be included to the database and their topics will be explored and classified to investigate whether the strategy leads to the growth in both the number of items and the breadth of topics.
Expected Results and Conclusions
Our work showcases the potential of using machine learning approaches to identify research publications on a wide range of topics in education in the context of limited resources and funding. We expect that although machine learning applied to searching and screening potentially relevant publications cannot fully substitute manual work, it will lead to significant reductions in the amount of human labour and will allow IDESR to grow and develop serving as an up-to-date source of knowledge to research producers and consumers."