Humans are a primate species with a unique social organization; we live in groups that are characterized by predominant kin and non-kin pair bonds. Although the nature of the relationship between individual members of the society is crucial to understand social network dynamics, this information is typically not available in “big data” datasets. Recent literature introduced a new methodology for identifying relationships types from automatically recorded mobile communication networks: age and gender of the ego’s callers together with the frequency of calls allows to identify parents, romantic partners and their offspring. In this paper, using a Chilean mobile communication data set, we introduce a refined version of the frequency-based methodology to detect relationship types. To do so, we exploit a feature of family names prevalent in most Iberoamerican cultures: each individual has two family names, one inherited from the father, one from the mother. Using anonymized both patrilineal and matrilineal family names as an additional filter together with the frequency of calls, we can identify which relationships were misidentified and which were missing from using frequency alone. We conclude that the refined methodology has similar patterns and small level of errors compared with the previous methodology.