Evaluation of a Social Media Summarization System: A Multidimensional Approach

Automatic summarization in recent years has increasingly attracted the interest of researchers because it is a technique tested in many applications and is now one of the most promising branches of research. The social media domain presents a challenge for the implementation of automatic summarization due to the peculiar nature of their content. Automatic summarization of information posted and commented on social networks is now essential for timely, valid and consistent public information. Through the culmination of research in the creation of summaries but also the culmination of technology, the demands of automatic summarizing capabilities are increasing more and more demanding readability and fluency of formulation similar to that of a human. On the other hand, conventional assessment measurement methods have not kept pace with the rapidly evolving forms of summary assignments. Differentiated solutions address different challenges by varying the metrics of different but important language assessment factors. This paper presents a summarization system of social media comments on posts of individual topics and evaluates it across diverse evaluation criteria. Since natural language generation evaluation is a multidimensional process a comparison is made in order to assess the generated summary for fluency, consistence, readability and semantic coherency.