For the last decade, social networking websites have boosted interaction among people through the use of digital communication such as chats, comments, discussion boards and exchange of documentation. This lead to mutual learning and sharing of all kind of information. This phenomenon has attracted many researchers and techniques aiming at discovering and prediction links between people have been developed. Most existing solutions are based on the similarity in the profiles using the declared personal information. We present in this paper an approach to discover and predict semantic links between members of a social network based on the content analysis. Our approach uses a textual aggregation function to aggregate keywords extracted from people’s textual production. The result of this aggregation is then used to predict semantic links between the members of the network. Experiments were carried out using a real scientific corpus extracted from ReseachGate Web Site. The obtained results showed that our approach, compared to others, achieves better performances in terms of recall, precision, F-measure, complexity and runtime.

在过去十年中，社交网站通过使用数字通信（例如聊天、评论、讨论板和文档交换）促进了人与人之间的互动。这导致相互学习和共享各种信息。这种现象吸引了许多研究人员，并开发了旨在发现和预测人与人之间联系的技术。大多数现有解决方案基于使用声明的个人信息的配置文件的相似性。我们在本文中提出了一种基于内容分析来发现和预测社交网络成员之间语义链接的方法。我们的方法使用文本聚合函数来聚合从人们的文本产品中提取的关键字。然后使用这种聚合的结果来预测网络成员之间的语义链接。实验是使用从 ReseachGate 网站提取的真实科学语料库进行的。获得的结果表明，与其他方法相比，我们的方法在召回率、精度、F 度量、复杂性和运行时间方面取得了更好的性能。