Fighting Irrelevant Health Videos on Youtube: a Social Network Analysis Approach
|
If you are the presenter of this abstract (or if you cite this abstract in a talk or on a poster), please show the QR code in your slide or poster (QR code contains this URL). |
Abstract
Introduction:
Videos are among the most common types of information resources in the Web 2.0. For example, in YouTube hundreds of users are distributing thousands of health videos in their channels. These users include hospitals, medical organizations, clinicians, individual patients, and others. Automated techniques for effectively searching health videos on YouTube remain an unmet challenge, and there are YouTube users providing health videos with dubious claims and even misleading information. Finding trusted and relevant channels is sometimes hard using traditional search tools, as even highly "relevant" channels in YouTube may be irrelevant from a health point of view (e.g. singers, herbal cures).
In this study we explore the use of social network analysis on a community of channels about diabetes in YouTube. We studied how different metrics of social network analysis (e.g. prestige of the nodes) can help to determine the characteristics of the channels. Our hypothesis is that prestige within the diabetic community will correlate with relevance and quality of channels from a health point of view. For example, a video about a singer with diabetes is extremely popular and relevant overall within YouTube but less popular if only considering the diabetes community.
Methods:
Using the YouTube search API we searched for channels with the keyword "diabetes". A total of 219 channels with 5119 videos were collected, and an additional, 525 links between the channels (e.g. favorite videos, subscriptions, friendships) were extracted. The prestige between channels using algorithms such as HITS and PageRank was calculated based upon links between channels. We compared the top-20 most prestigious channels calculated using different social network analysis algorithms with the list of the 20 most relevant channels retrieved by YouTube. These lists of channels were evaluated by two physicians unaware of which technique was used to retrieve the channels. Channels were rated by the physicians for relevance, perceived quality, and if they would recommend the channel to a diabetic patient.
Results:
The lists created with of our social network analysis contained more recommended channels and higher average quality. The analysis of subscriptions and favorite videos turned out to be the most valuable links between the channels. Different limitations were also found. For example, some users are actively promoting the subscriptions to their channels while others have no promotion. Therefore, the best results were achieved when combining different links.
Conclusions:
Social network analysis applied to health online communities can be used to extract information about the health information providers. Our preliminary results show that this approach can outperform general-purpose information retrieval tools. In addition, the analysis of health communities using social network analysis it is a great tool to increase the knowledge about health content generation.
Our approach can be easily applied to other health-related conditions and is particularly appealing because it does not require manual intervention by human experts and its dynamically nature as additional valuable video resources are added. In our future work, we will study if this hypothesis is valid in other health communities. In addition, we are integrating our approach to create a video recommender system.
Videos are among the most common types of information resources in the Web 2.0. For example, in YouTube hundreds of users are distributing thousands of health videos in their channels. These users include hospitals, medical organizations, clinicians, individual patients, and others. Automated techniques for effectively searching health videos on YouTube remain an unmet challenge, and there are YouTube users providing health videos with dubious claims and even misleading information. Finding trusted and relevant channels is sometimes hard using traditional search tools, as even highly "relevant" channels in YouTube may be irrelevant from a health point of view (e.g. singers, herbal cures).
In this study we explore the use of social network analysis on a community of channels about diabetes in YouTube. We studied how different metrics of social network analysis (e.g. prestige of the nodes) can help to determine the characteristics of the channels. Our hypothesis is that prestige within the diabetic community will correlate with relevance and quality of channels from a health point of view. For example, a video about a singer with diabetes is extremely popular and relevant overall within YouTube but less popular if only considering the diabetes community.
Methods:
Using the YouTube search API we searched for channels with the keyword "diabetes". A total of 219 channels with 5119 videos were collected, and an additional, 525 links between the channels (e.g. favorite videos, subscriptions, friendships) were extracted. The prestige between channels using algorithms such as HITS and PageRank was calculated based upon links between channels. We compared the top-20 most prestigious channels calculated using different social network analysis algorithms with the list of the 20 most relevant channels retrieved by YouTube. These lists of channels were evaluated by two physicians unaware of which technique was used to retrieve the channels. Channels were rated by the physicians for relevance, perceived quality, and if they would recommend the channel to a diabetic patient.
Results:
The lists created with of our social network analysis contained more recommended channels and higher average quality. The analysis of subscriptions and favorite videos turned out to be the most valuable links between the channels. Different limitations were also found. For example, some users are actively promoting the subscriptions to their channels while others have no promotion. Therefore, the best results were achieved when combining different links.
Conclusions:
Social network analysis applied to health online communities can be used to extract information about the health information providers. Our preliminary results show that this approach can outperform general-purpose information retrieval tools. In addition, the analysis of health communities using social network analysis it is a great tool to increase the knowledge about health content generation.
Our approach can be easily applied to other health-related conditions and is particularly appealing because it does not require manual intervention by human experts and its dynamically nature as additional valuable video resources are added. In our future work, we will study if this hypothesis is valid in other health communities. In addition, we are integrating our approach to create a video recommender system.
Medicine 2.0® is happy to support and promote other conferences and workshops in this area. Contact us to produce, disseminate and promote your conference or workshop under this label and in this event series. In addition, we are always looking for hosts of future World Congresses. Medicine 2.0® is a registered trademark of JMIR Publications Inc., the leading academic ehealth publisher.
This work is licensed under a Creative Commons Attribution 3.0 License.