Developments in machine learning have enabled computers to make significant advances in identifying human faces as well as other objects in moving images. Determining whether or not machine learning can be applied as effectively to the analysis of film language – that is, to identifying film form rather than film content – is a question Kinolab is researching in partnership with Northeastern University’s Roux Institute, which specializes in the practical application of artificial intelligence and machine learning.
Automating elements of film language analysis is beneficial to quantitative analysis projects focusing on moving images for several reasons. Computer vision may, for example, be able to identify aspects of film language more efficiently than human researchers or curators, allowing film and media scholars to spend less time preparing data for interpretation and more time interpreting data.
By collaborating with computer scientists who focus on machine learning, Kinolab is helping to create a virtuous research cycle in which the project’s annotated clips serve as training data for machine learning projects that can subsequently assist Kinolab or other DH-inflected projects for moving image analysis.
If you are a computer scientist interested in utilizing Kinolab data for machine learning or a cinema and media studies scholar interested in collaborating with Kinolab to research a question related to film language and machine learning, please email project director Allison Cooper (firstname.lastname@example.org) to learn more.