Abstract. The information conveyed via the social media, in addition to the content data, also contains social characteristics that come from the social network users. A special interesting data category concerns the data that come from the natural language present in the social media mainly in the form of video. Our study focuses on the speech content of the videos in the form of transcript and the opinion of the social network users that have watched them. The representation of content data is made through a vector space model that uses cosine similarity measure for the relevant ranking of the transcripts. In order for the ranking to be more comprehensive we suggest the addition of a new parameter that of social weight during the procedure, which will reflect the users’ opinion. There is an analytic presentation of the method being suggested; all the possible cases are being examined and the rules that define the new ranking are put forward. Furthermore, we apply this method to video lectures derived from YouTube. The findings of the experiments show that the addition of the social weight parameter reflects the users’ views without changing to great extent the content based ranking of the video lectures. Finally, a user evaluation experiment was carried out and showed that the ranking procedure that includes the social weight parameter is closer to the users’ ranking preferences.