sanaemahtal / geolocation-1

geolocation of Twitter users based on text and network information

#Geolocation of Social Media Users Using Text and Network Information Afshin Rahimi, [email protected]

This tool implements the geolocation models described in the following two papers:

@InProceedings{rahimi2015exploiting,
author="Rahimi, Afshin and Vu, Duy and Cohn, Trevor and Baldwin, Timothy",
title="Exploiting Text and Network Context for Geolocation of Social Media Users",
booktitle="Proceedings of the 2015 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies",
year="2015",
publisher="Association for Computational Linguistics",
pages="1362--1367",
location="Denver, Colorado",
url="http://aclweb.org/anthology/N15-1153"
}

@InProceedings{rahimi2015twitter,
author="Rahimi, Afshin and Cohn, Trevor and Baldwin, Timothy",
title="Twitter User Geolocation Using a Unified Text and Network Prediction Model",
booktitle="Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics and the 7th International Joint Conference on Natural Language Processing (Volume 2: Short Papers)",
year="2015",
publisher="Association for Computational Linguistics",
pages="630--636",
location="Beijing, China",
url="http://aclweb.org/anthology/P15-2104"
}

##Models The models include text-based classification, network-based label propagation (regression) and network-based label propagation (classification). To run a model add the model name from params.all_models to the params.models_to_run list.

##Data ###Clusters (regions) for classification models For the classification models, the real-valued coordinates of the training points are clustered using k-d tree and each cluster (region) is assigned a label. This implementation expects the clusters to be written into a text file where training point members of each cluster are written in one line tab-separated. For example int the following example we have 2 clusters (regions) each with two training points:

lat1,lon1	lat2,lon2
lat3,lon3	lat4,lon4

###Training, development and test input files This program expects three separate training, dev and test gzipped files in each, a user is represented as a line with the following format:

username	latitude	longitude	aggregated-tweets-of-user-including-mentions

Note: The datasets used are different from those used in: Wing, Benjamin, and Jason Baldridge. "Hierarchical Discriminative Classification for Text-Based Geolocation." EMNLP. 2014. in that we have rebuilt the datasets to include mention information.

##Configuration All the configuration parameters are in params.py which should be changed according to the environment.