[Defense] Automatic Content Assessment for Safe and Healthy Media Experiences
Tuesday, May 17, 2022
5:00 pm - 6:30 pm
will defend his proposal
Automatic Content Assessment for Safe and Healthy Media Experiences
In this proposal, we introduce the task of automatic content assessment for media. Digital media products such as movies, songs, and books not only entertain people but also produce knowledge. They appear everywhere in people’s daily life but sometimes media content is not suitable for users of all ages. For example, movies with sexually suggestive languages or violent scenes might not be suitable for children to watch. Policymakers keep an eye on media suitability. They developed systems such as the Motion Picture Association (MPA) film rating system for movies and the Parental Advisory Label (PAL) for music to classify the age suitability for their media products. Service providers like streaming websites also put efforts into giving suitability suggestions to their customers. Current rating systems rely on experts to do the classification, the process of which will be expensive and inefficient. Meanwhile, such ratings only provide a general age-based suitability label. They are not informative enough if a user would like to know more aspects of the content. We address these shortcomings by automating the media content assessment process from different perspectives using natural language processing (NLP) techniques.
First, we investigate rating movie severity in different age-restricted aspects (such as violence and sex) to provide perceptible level information as a compliment for the general suitability category. Then we expand the scope from movies to music products to assess the song lyrics on not only the risky aspects but also positive messages. Moreover, we propose to go one step further to study how and to what extent the NLP model can interpret themes and educational values from the literature. All of our efforts will go to one goal: providing comprehensive information for media products. In line with those research topics, we first formulate the media content assessment tasks as machine learning problems and then create benchmark datasets for movie and music assessment. Based on the benchmarks, we successfully devise, implement, evaluate, and analyze the proposed methods to address the research problems. We plan to further explore the task of story theme comprehension to complete this dissertation, which contributes to the goal of providing people with safe and healthy media experiences.
5:00PM - 6:30PM CT
Virtual via Zoom
Dr. Thamar Solorio, dissertation advisor
Faculty, students and the general public are invited.