The invention discloses a learning-based instant communication session segmentation technology and method and belongs to the field of big data analysis. Every two instant communication session users are divided into one group, and session sheet details of each group are classified and ordered based on time. According to session segmentation, two adjacent sheets R1 and R2 are sequentially selected, and a time interval delta t, text content similarity delta sim and a distance value F(R1,R2) are calculated; if F(R1,R2) is smaller than f, the sheets R1 and R2 belong to the same session; otherwise, the sheets R1 and R2 belong to two different sessions respectively; operation is performed on all the user groups and all similar sheet detail data of each user group parallelly at the same time through spark; and finally all sessions of all instant communication users are segmented. Through the instant communication session segmentation technology and method, session time distance impact factors and session text content distance impact factors are considered comprehensively, different session segmentation standards for different session user groups are realized, and accuracy and high efficiency of segmentation of mass instant communication text sessions in a big data background are effectively achieved.