A computer implemented method for estimating a geographic location (geotag) from one or more short texts is described. Geo-tagging a short text post involves first determining location candidates from parsing text and then determining location resolutions (eg city, state) for each location candidate. Typically there are multiple resolutions for each candidate and then location resolution disambiguation is performed to identify the correct resolution from the set of potential resolutions This may be achieved by using context provided by other candidates within the post, as well as additional geographic context derived from the post author and their network. In particular, anchor locations may be obtained from metadata associated with the post to assist in resolution disambiguation.