A SURVEY OF LOCATION PREDICTION ON TWITTER

Abstract

Places such as countries, states, cities, and points-of-interest play
an essential role in news coverage, emergency situations, and
people's daily activities. They are also crucial in politics.
Researchers have been experimenting with automated
recognition of locations that are related to or referenced in
documents for several decades. Because of the vast number of
users that send millions of tweets every day, Twitter has risen to
become one of the most popular social media platforms available
today. Geographic prediction has gained a great deal of attention
in recent years, owing to Twitter's global reach as well as the
real-time freshness of the information included in tweets in real
time. The majority of the research is devoted to identifying and
solving the new challenges and opportunities given by the loud,
quick, and contextually rich nature of Twitter messages. In
addition, we hope that this survey will give a more comprehensive
picture of location prediction on Twitter than we now have. To be
more specific, we're looking for user home location forecasts,
tweet location predictions, and mentioned location predictions.
We begin by identifying the three tasks and going over the
assessment criteria one more time. When we summarise and
analyse the Twitter network as well as the tweet content and
context as possible inputs, we can more systematically explain
how these inputs have an impact on the issues in question.
Detailed analyses of the solutions that have been implemented in
current best practises are offered for each dependency to support
the point being made. In addition, we provide a high-level
description of two related challenges, semantic location
prediction and point-of-interest recommendation, which are
treated in further depth later in this section. We then draw a
conclusion based on the facts and offer some suggestions for
further research.