Base_JSON/prefixW/json/wanlp/2021.wanlp-1.15.json

{
    "paper_id": "2021",
    "header": {
        "generated_with": "S2ORC 1.0.0",
        "date_generated": "2023-01-19T05:59:02.949947Z"
    },
    "title": "UL2C: Mapping User Locations to Countries on Arabic Twitter",
    "authors": [
        {
            "first": "Hamdy",
            "middle": [],
            "last": "Mubarak",
            "suffix": "",
            "affiliation": {
                "laboratory": "",
                "institution": "Qatar Computing Research Institute Doha",
                "location": {
                    "country": "Qatar"
                }
            },
            "email": "[email protected]"
        },
        {
            "first": "Sabit",
            "middle": [],
            "last": "Hassan",
            "suffix": "",
            "affiliation": {
                "laboratory": "",
                "institution": "Qatar Computing Research Institute Doha",
                "location": {
                    "country": "Qatar"
                }
            },
            "email": "[email protected]"
        }
    ],
    "year": "",
    "venue": null,
    "identifiers": {},
    "abstract": "Mapping user locations to countries can be useful for many applications such as dialect identification, author profiling, recommendation systems, etc. Twitter allows users to declare their locations as free text, and these userdeclared locations are often noisy and hard to decipher automatically. In this paper, we present the largest manually labeled dataset for mapping user locations on Arabic Twitter to their corresponding countries. We build effective machine learning models that can automate this mapping with significantly better efficiency compared to libraries such as geopy. We also show that our dataset is more effective than data extracted from GeoNames geographical database in this task as the latter covers only locations written in formal ways.",
    "pdf_parse": {
        "paper_id": "2021",
        "_pdf_hash": "",
        "abstract": [
            {
                "text": "Mapping user locations to countries can be useful for many applications such as dialect identification, author profiling, recommendation systems, etc. Twitter allows users to declare their locations as free text, and these userdeclared locations are often noisy and hard to decipher automatically. In this paper, we present the largest manually labeled dataset for mapping user locations on Arabic Twitter to their corresponding countries. We build effective machine learning models that can automate this mapping with significantly better efficiency compared to libraries such as geopy. We also show that our dataset is more effective than data extracted from GeoNames geographical database in this task as the latter covers only locations written in formal ways.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Abstract",
                "sec_num": null
            }
        ],
        "body_text": [
            {
                "text": "Twitter is one of the most popular social media platforms in the Arab region. Spoken across more than 20 countries, Arabic is one of the most dominant languages on Twitter. Arabic Twittersphere presents us with an audience of diverse demographics. Identifying countries of the users can help in various NLP tasks such as dialect identification, author profiling, and recommendation systems. Identifying geolocation of Twitter users can also help in event detection (Agarwal et al., 2012) or disaster management (Earle et al., 2012; Carley et al., 2016) .",
                "cite_spans": [
                    {
                        "start": 465,
                        "end": 487,
                        "text": "(Agarwal et al., 2012)",
                        "ref_id": "BIBREF7"
                    },
                    {
                        "start": 511,
                        "end": 531,
                        "text": "(Earle et al., 2012;",
                        "ref_id": "BIBREF13"
                    },
                    {
                        "start": 532,
                        "end": 552,
                        "text": "Carley et al., 2016)",
                        "ref_id": "BIBREF10"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Introduction",
                "sec_num": "1"
            },
            {
                "text": "Deducing a Twitter user's location from geotagged tweets is difficult because less than 1% of tweets are geotagged (Cheng et al., 2010; Hecht et al., 2011) . Although Twitter provides an option to users to declare their location in their profile, this is often noisy. Users can choose to specify their location at the level of countries, regions, cities or towns. Many of these names are written in informal way and sometimes in mixed languages. Some of these names also contain emojis and special symbols. This makes it difficult to automatically infer the country of many users. This complication often prompts researchers to manually annotate user profiles for their countries. Many works therefore, (e.g., (Bouamor et al., 2019; Charfi et al., 2019) ), manually annotate user profiles for their locations. Being expensive, manual annotation often limits size of datasets. This is evident in the datasets by Bouamor et al. (2019) and Charfi et al. (2019) since they both contain around 3K manually annotated users.",
                "cite_spans": [
                    {
                        "start": 115,
                        "end": 135,
                        "text": "(Cheng et al., 2010;",
                        "ref_id": "BIBREF12"
                    },
                    {
                        "start": 136,
                        "end": 155,
                        "text": "Hecht et al., 2011)",
                        "ref_id": "BIBREF19"
                    },
                    {
                        "start": 710,
                        "end": 732,
                        "text": "(Bouamor et al., 2019;",
                        "ref_id": "BIBREF9"
                    },
                    {
                        "start": 733,
                        "end": 753,
                        "text": "Charfi et al., 2019)",
                        "ref_id": "BIBREF11"
                    },
                    {
                        "start": 911,
                        "end": 932,
                        "text": "Bouamor et al. (2019)",
                        "ref_id": "BIBREF9"
                    },
                    {
                        "start": 937,
                        "end": 957,
                        "text": "Charfi et al. (2019)",
                        "ref_id": "BIBREF11"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Introduction",
                "sec_num": "1"
            },
            {
                "text": "Related works for Arabic primarily focus on dialect identification (e.g., (Bouamor et al., 2019; Zaidan and Callison-Burch, 2011 ), many of which involve manual annotation of dialects for sentences. Our focus in this work significantly differs from dialect identification since our purpose is to provide a dataset that can be used to map Twitter user locations to countries, which in turn, can aid in NLP tasks such as dialect identification or event detection. To our knowledge, there has been very few works for Arabic that map noisy user locations to countries, with work by Mubarak and Darwish (2014) being one of the most notable ones and the closest to our work. Mubarak and Darwish (2014) map the most common 10K locations to Arab countries in order to build a multidialectal Arabic corpus. We refer to this dataset as Top10KLoc. Our work extends Top10KLoc by increasing unique user locations from most common 10K to random 28K user locations obtained from 160K locations that are self-declared by users 1 .",
                "cite_spans": [
                    {
                        "start": 74,
                        "end": 96,
                        "text": "(Bouamor et al., 2019;",
                        "ref_id": "BIBREF9"
                    },
                    {
                        "start": 97,
                        "end": 128,
                        "text": "Zaidan and Callison-Burch, 2011",
                        "ref_id": "BIBREF31"
                    },
                    {
                        "start": 578,
                        "end": 604,
                        "text": "Mubarak and Darwish (2014)",
                        "ref_id": "BIBREF26"
                    },
                    {
                        "start": 669,
                        "end": 695,
                        "text": "Mubarak and Darwish (2014)",
                        "ref_id": "BIBREF26"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Introduction",
                "sec_num": "1"
            },
            {
                "text": "The contributions of this paper are summarized below:",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Introduction",
                "sec_num": "1"
            },
            {
                "text": "\u2022 We present UL2C; the largest dataset for mapping user locations on Arabic Twitter to countries which contains more than 28K unique locations.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Introduction",
                "sec_num": "1"
            },
            {
                "text": "\u2022 We perform analysis of the data collected, identifying key characteristics.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Introduction",
                "sec_num": "1"
            },
            {
                "text": "\u2022 We show that by using machine learning models trained on our dataset, we can achieve significantly better results compared to existing libraries (e.g. geopy package) or resources (e.g. GeoNames or Top10KLoc 2 datasets). Models trained on our dataset achieve macroaverage F1 score of 88.1, which outperforms similar models trained on other datasets with at least 26% relative improvement.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Introduction",
                "sec_num": "1"
            },
            {
                "text": "\u2022 We provide a web interface that users can use to map user locations to countries.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Introduction",
                "sec_num": "1"
            },
            {
                "text": "There has been a number of works that focus on identifying locations of Twitter users. Krishnamurthy et al. (2008) Despite Arabic being one of the most popular languages on Twitter, there has been very few works aimed at mapping user location in the Arab region to countries. The related field of dialect identification has received significant attention recently. Some works identify country-level dialects of Arabic tweets (e.g., and some focus on dialects at user-level (Bouamor et al., 2019) . While Abdelali et al. (2020) automatically labeled 500K+ tweets for their country-level dialects, Bouamor et al. 2019 As discussed earlier, the closest work to ours that targets converting user locations directly to countries is by Mubarak and Darwish (2014) . The authors collect 10K top user locations from 92M tweets. To map these locations to countries, the authors first use GeoNames dataset and then manually revise them. Further, the authors show that after filtering out non-dialectal tweets, the countries obtained from the user locations can be strong indicator of dialects for the remaining tweets.",
                "cite_spans": [
                    {
                        "start": 87,
                        "end": 114,
                        "text": "Krishnamurthy et al. (2008)",
                        "ref_id": "BIBREF21"
                    },
                    {
                        "start": 473,
                        "end": 495,
                        "text": "(Bouamor et al., 2019)",
                        "ref_id": "BIBREF9"
                    },
                    {
                        "start": 730,
                        "end": 756,
                        "text": "Mubarak and Darwish (2014)",
                        "ref_id": "BIBREF26"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Related Work",
                "sec_num": "2"
            },
            {
                "text": "We used twarc search API 3 to collect Arabic tweets. During the years 2018, 2019 and 2020, we collected 88M tweets with language filter set to Arabic (lang:ar). From unique users who authored those tweets, we randomly selected 160K unique users for which we obtained 28K unique user locations.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Data Collection",
                "sec_num": "3"
            },
            {
                "text": "The user locations are information provided by Twitter users, and they can be real locations (e.g., country and city names written in formal, informal ways or nicknames), landmarks, or unreal locations, and can be written in any language. In our data collection, we found that 62% of users pro-vided non-empty locations. Around 60% of the non-empty locations are written in Arabic and 40% are written in other languages, mainly in English.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Data Collection",
                "sec_num": "3"
            },
            {
                "text": "We used geopy 4 to map user locations to countries. geopy is a Python package that locates the coordinates of addresses, cities, countries, and landmarks across the globe using third-party geocoders and other data sources. It includes geocoder classes for the OpenStreetMap Nominatim, Google Geocoding API (V3), and many other geocoding services. Figure 1 shows the output from geopy for an arbitrary location. In our study, we focus on mapping user locations to countries and we use ISO 3166-1 alpha-2 for country codes 5 . We observed that geopy has difficulties in identifying many locations when they are short, have special characters, or unreal locations in addition to many correct locations. Table 1 shows some examples for these errors.",
                "cite_spans": [],
                "ref_spans": [
                    {
                        "start": 347,
                        "end": 355,
                        "text": "Figure 1",
                        "ref_id": null
                    },
                    {
                        "start": 700,
                        "end": 707,
                        "text": "Table 1",
                        "ref_id": "TABREF2"
                    }
                ],
                "eq_spans": [],
                "section": "Data annotation",
                "sec_num": "4"
            },
            {
                "text": "We used geopy output as an initial mapping of user locations to countries then all unique user locations were revised manually by an Arabic native speaker.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Data annotation",
                "sec_num": "4"
            },
            {
                "text": "In addition to mapping clear locations to countries, the annotator was asked to consider any clues in user location string that indicate belonging to a specific country. Some common examples and special annotation cases are shown in Table 2 .",
                "cite_spans": [],
                "ref_spans": [
                    {
                        "start": 233,
                        "end": 240,
                        "text": "Table 2",
                        "ref_id": "TABREF4"
                    }
                ],
                "eq_spans": [],
                "section": "Data annotation",
                "sec_num": "4"
            },
            {
                "text": "We randomly selected 500 unique user locations and checked annotation quality. We agreed with the manual annotations in 98% of the cases which indicates that annotation quality is very high.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Data annotation",
                "sec_num": "4"
            },
            {
                "text": "User locations and their country mapping (UL2C dataset) can be downloaded from this link:",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Data annotation",
                "sec_num": "4"
            },
            {
                "text": "https://alt.qcri.org/resources/ UL2C-UserLocationsToCountries.tsv",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Data annotation",
                "sec_num": "4"
            },
            {
                "text": "GeoNames geographical database covers all countries and contains over 11M placenames whereof 4.8M populated places and 13M alternate names. Users can manually edit, correct and add new names using a user friendly interface. Dataset can be downloaded from: Figure 2 shows some information about Damascus, the capital of Syria, and its alternate names written in tens of languages as obtained from GeoNames. We extracted Arabic and English names of all places with population of 10K or more 6 . The figure shows also an example of the excluded locations that we anticipate users will not use to describe their locations. We ended up with having a list of 66K English location names (ASCII name field) and a shorter list of 13K Arabic names for some of them.",
                "cite_spans": [],
                "ref_spans": [
                    {
                        "start": 256,
                        "end": 264,
                        "text": "Figure 2",
                        "ref_id": "FIGREF1"
                    }
                ],
                "eq_spans": [],
                "section": "GeoNames Dataset",
                "sec_num": "4.1"
            },
            {
                "text": "In the experiments section, we will examine the efficiency of using GeoNames to identify countries of Twitter users.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "GeoNames Dataset",
                "sec_num": "4.1"
            },
            {
                "text": "By assigning countries to unique user locations, we could map locations of \u2248 90K users to countries which represent 56% of the 160K users in our dataset. Many of user locations were either empty (38%) or cannot be mapped to a specific country (6%). Distribution of user countries is shown in Figure 3 . Although there are 22 Arab countries, in our collection we didn't find locations from two countries, namely Djibouti and Comoros. We observe that users from Saudi Arabia (SA) represent more than half of Arab Twitter users. Around 70% of Twitter users are from Gulf region (countries: SA, KW, OM, AE, QA and BH), 4% of users are from Levant region (JO, PS, LB and SY), 3% of users are from Maghreb region (DZ, LY, MA and TN), and users from other regions (EG, YE and IQ) are in between. It's worth mentioning that country distribution obtained from UL2C and Top10KLoc datasets are very similar which indicates that most probably any random big collection of tweets will have similar country distribution.",
                "cite_spans": [],
                "ref_spans": [
                    {
                        "start": 292,
                        "end": 300,
                        "text": "Figure 3",
                        "ref_id": "FIGREF2"
                    }
                ],
                "eq_spans": [],
                "section": "Data Analysis",
                "sec_num": "5"
            },
            {
                "text": "From the geographical map of all Twitter users shown in Figure 4 , we can see that Arabic tweets come from almost all world countries. The top 5 countries outside the Arab World where Arabic tweets come from are: US (United States), GB (United Kingdom), TR (Turkey), DE (Germany) and FR (France) in order. This can give an estimation about countries that Arabs live in outside Arab countries. We notice also that tweets from US are more than those from individual Maghreb countries like DZ and TN. Figure 5 shows the most common words used in user locations for 4 countries, namely SA (Saudi Arabia), EG (Egypt), SY (Syria) and DZ (Algeria) which represent major regions in the Arab World (Gulf, Levant, Nile Basin and Maghreb regions respectively). We can see country and major city names are written in bigger font in different languages. For example, while majority of names are written in Arabic in SA (Gulf region), they are written in Arabic, English and French in DZ (Maghreb region). Arabic and English names are widely used in EG and SY. This gives an indication about the popularity of language usage across different regions in the Arab World.",
                "cite_spans": [],
                "ref_spans": [
                    {
                        "start": 56,
                        "end": 64,
                        "text": "Figure 4",
                        "ref_id": "FIGREF3"
                    },
                    {
                        "start": 498,
                        "end": 506,
                        "text": "Figure 5",
                        "ref_id": "FIGREF4"
                    }
                ],
                "eq_spans": [],
                "section": "Data Analysis",
                "sec_num": "5"
            },
            {
                "text": "In this section, we compare effectiveness of mapping user locations to countries with classifiers trained on different datasets, namely, UL2C, Top10KLoc, GeoNames and combination of these datasets. In our experiments, we merge all countries that are outside Arab region to \"UNK/OTHER\" class, yielding a total of 21 classes (20 Arab countries + Unknown/Other).",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Experiments and Results",
                "sec_num": "6"
            },
            {
                "text": "In order to reduce noise in user-declared locations, we perform the following preprocessing steps: \u2022 Convert English and Arabic decorated letters (e.g. some Farsi letters) to original letters using the mapping list shared by (Mubarak and Abdelali, 2016) .For example, we map \"\u03b1, \u03b2\" to \"a, B\" and \" \" to \" \" in order.",
                "cite_spans": [
                    {
                        "start": 225,
                        "end": 253,
                        "text": "(Mubarak and Abdelali, 2016)",
                        "ref_id": "BIBREF24"
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Preprocessing text",
                "sec_num": "6.1"
            },
            {
                "text": "In summary, a location like \"\u03b1\u03b2HA1 :) *, \" is converted to \"abha \" (city in SA) after decorated letters mapping and applying other normalization steps.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Preprocessing text",
                "sec_num": "6.1"
            },
            {
                "text": "Word n-gram Since our input text is name of a location and therefore, typically short, we limit range of word n-gram to [1] [2] . We use term frequency-inverse term document frequency (tf-idf) for weighting the n-gram vectors.",
                "cite_spans": [
                    {
                        "start": 120,
                        "end": 123,
                        "text": "[1]",
                        "ref_id": null
                    },
                    {
                        "start": 124,
                        "end": 127,
                        "text": "[2]",
                        "ref_id": null
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Features",
                "sec_num": "6.2"
            },
            {
                "text": "Character n-gram We experimented with character n-grams ranging from [2-3] to [2-5], beyond which, we did not see any further improvement. Similar to word n-gram, we used tf-idf weighting.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Features",
                "sec_num": "6.2"
            },
            {
                "text": "Features Acc P R F1 -geopy (baseline) - (Hassan et al., 2020b,a) , spam detection , adult content detection , and dialect identification (Abdelali et al., 2020), SVMs have shown promising results. We used LibSVM implementation with default parameters by scikit-learn 7 for training.",
                "cite_spans": [
                    {
                        "start": 40,
                        "end": 64,
                        "text": "(Hassan et al., 2020b,a)",
                        "ref_id": null
                    }
                ],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Data Model",
                "sec_num": null
            },
            {
                "text": "geopy baseline Serving as our baseline model, geopy achieves F1 score of 69.2 when evaluated on UL2C dataset.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Experiment results",
                "sec_num": "6.4"
            },
            {
                "text": "GeoNames We trained several classifiers on GeoNames dataset and evaluated on our UL2C dataset. The classifiers yielded very poor results 7 http://scikit-learn.org/ initially. We noticed that many of the locations were outside Arab region which likely caused the poor performance (omitted here for conciseness).",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Experiment results",
                "sec_num": "6.4"
            },
            {
                "text": "To address this problem, we rebalanced the dataset to have equal number of locations from within and outside the Arab region. This yielded a total of 8,632 unique instances (4,316 from the Arab region and 4,316 from outside the Arab region). The results are summarized in Table 3 . We can see that even after rebalancing the data, with maximum F1 score of 50.1, the classifiers trained on GeoNames are outperformed by all others.",
                "cite_spans": [],
                "ref_spans": [
                    {
                        "start": 272,
                        "end": 279,
                        "text": "Table 3",
                        "ref_id": "TABREF5"
                    }
                ],
                "eq_spans": [],
                "section": "Experiment results",
                "sec_num": "6.4"
            },
            {
                "text": "We trained similar set of classifiers on Top10KLoc dataset and evaluated on our dataset. The models were seen to outperform geopy by a small margin and GeoNames by a large margin with maximum F1 score of 70.5 (see Table 3 ).",
                "cite_spans": [],
                "ref_spans": [
                    {
                        "start": 214,
                        "end": 221,
                        "text": "Table 3",
                        "ref_id": "TABREF5"
                    }
                ],
                "eq_spans": [],
                "section": "Top10KLoc",
                "sec_num": null
            },
            {
                "text": "We performed 5-fold crossvalidation with same set of classifiers on the dataset presented in this paper. The results were seen to improve by a significant margin since the best results were obtained by SVM trained with character [2-5] n-grams, a relative improvement of 26% in F1 score from previous best (70.5) when trained on Top10KLoc. UL2C + GeoNames In this set of experiments, we modified the cross-validation setting by adding GeoNames dataset with each of the 5 folds of our dataset during training. We can see from Table 3 that adding our dataset to GeoNames dataset offsets the lower performance when using GeoNames alone and with F1 score of 88.2, improves the results from using UL2C dataset alone by a small margin.",
                "cite_spans": [],
                "ref_spans": [
                    {
                        "start": 524,
                        "end": 531,
                        "text": "Table 3",
                        "ref_id": "TABREF5"
                    }
                ],
                "eq_spans": [],
                "section": "UL2C (Our dataset)",
                "sec_num": null
            },
            {
                "text": "Lastly, we modified the cross-validation setting by adding Top10KLoc to each of the folds during training. From Table 3 , we see a similar trend where the lower performance of using Top10KLoc only is offset by use of UL2C. However, there is no significant improvement when additional data is used compared to when using only UL2C.",
                "cite_spans": [],
                "ref_spans": [
                    {
                        "start": 112,
                        "end": 119,
                        "text": "Table 3",
                        "ref_id": "TABREF5"
                    }
                ],
                "eq_spans": [],
                "section": "UL2C + Top10KLoc",
                "sec_num": null
            },
            {
                "text": "We build an interface for users to map user locations to countries. The web interface is added to Arabic Social media Analytics and unDerstanding (ASAD) at https://asad. qcri.org. Figure 6 shows sample outputs from the website.",
                "cite_spans": [],
                "ref_spans": [
                    {
                        "start": 180,
                        "end": 188,
                        "text": "Figure 6",
                        "ref_id": "FIGREF5"
                    }
                ],
                "eq_spans": [],
                "section": "Interface",
                "sec_num": "7"
            },
            {
                "text": "The user can type user location to be mapped to countries. The user can also test random samples from UL2C dataset to see their mapping. This allows the user to easily understand the functionalities of the interface. The user is then shown probabilities of the location belonging to different countries. To help the users visualize distribution of possible countries related to the location, we dis-play a heatmap of the probabilities. We also allow the user to upload a file consisting user locations. This allows users to map many user locations at the same time. We impose a restriction on file size in order to limit abuse of our system. The user is then able to download a file containing predictions and probabilities of the user locations belonging to different countries.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Design",
                "sec_num": "7.1"
            },
            {
                "text": "We use Bootstrap 8 for our responsive frontend design. We use Flask 9 , a python-based web development framework, for backend development and javascript for communication between backend and frontend. To visualize the heatmap, we use Leaflet.js 10 and Heatmap.js 11 with Open-StreetMap 12 map server.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Implementation",
                "sec_num": "7.2"
            },
            {
                "text": "In this paper, we have presented a large manually annotated and publicly available dataset of Twitter user locations from the Arab region, mapped to their respective countries. We analyzed different characteristics of our data such as country distribution and top locations. We built machine learning models that can use the data to map user locations to countries more effectively compared to existing resources such as geopy Python package or GeoNames geographical database. Lastly, we provide a web interface to access this service easily.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "Conclusion",
                "sec_num": "8"
            },
            {
                "text": "About 2.2K locations from Top10KLoc appear in our dataset.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "",
                "sec_num": null
            },
            {
                "text": "Mubarak and Darwish (2014) didn't build an automatic classifier for mapping locations to countries.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "",
                "sec_num": null
            },
            {
                "text": "https://github.com/DocNow/twarc",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "",
                "sec_num": null
            },
            {
                "text": "https://pypi.org/project/geopy/ 5 https://en.wikipedia.org/wiki/List_ of_ISO_3166_country_codes https://www.geonames.org/.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "",
                "sec_num": null
            },
            {
                "text": "We want to exclude river names, mountain names, lakes, etc.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "",
                "sec_num": null
            }
        ],
        "back_matter": [],
        "bib_entries": {
            "BIBREF4": {
                "ref_id": "b4",
                "title": "UL2C + Top10KLoc SVM Char",
                "authors": [],
                "year": null,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "UL2C + Top10KLoc SVM Char [2-3] 89.4 94.5 78.7 85.2 Char [2-4] 90.4 95.0 81.5 87.3",
                "links": null
            },
            "BIBREF5": {
                "ref_id": "b5",
                "title": "Arabic dialect identification in the wild",
                "authors": [
                    {
                        "first": "Hamdy",
                        "middle": [],
                        "last": "References Ahmed Abdelali",
                        "suffix": ""
                    },
                    {
                        "first": "Younes",
                        "middle": [],
                        "last": "Mubarak",
                        "suffix": ""
                    },
                    {
                        "first": "Sabit",
                        "middle": [],
                        "last": "Samih",
                        "suffix": ""
                    },
                    {
                        "first": "Kareem",
                        "middle": [],
                        "last": "Hassan",
                        "suffix": ""
                    },
                    {
                        "first": "",
                        "middle": [],
                        "last": "Darwish",
                        "suffix": ""
                    }
                ],
                "year": 2020,
                "venue": "ArXiv",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "References Ahmed Abdelali, Hamdy Mubarak, Younes Samih, Sabit Hassan, and Kareem Darwish. 2020. Ara- bic dialect identification in the wild. ArXiv, abs/2005.06557.",
                "links": null
            },
            "BIBREF6": {
                "ref_id": "b6",
                "title": "You tweet what you speak: A city-level dataset of Arabic dialects",
                "authors": [
                    {
                        "first": "Muhammad",
                        "middle": [],
                        "last": "Abdul-Mageed",
                        "suffix": ""
                    },
                    {
                        "first": "Hassan",
                        "middle": [],
                        "last": "Alhuzali",
                        "suffix": ""
                    },
                    {
                        "first": "Mohamed",
                        "middle": [],
                        "last": "Elaraby",
                        "suffix": ""
                    }
                ],
                "year": 2018,
                "venue": "Proceedings of the Eleventh International Conference on Language Resources and Evaluation (LREC-2018)",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Muhammad Abdul-Mageed, Hassan Alhuzali, and Mo- hamed Elaraby. 2018. You tweet what you speak: A city-level dataset of Arabic dialects. In Proceed- ings of the Eleventh International Conference on Language Resources and Evaluation (LREC-2018), Miyazaki, Japan. European Languages Resources Association (ELRA).",
                "links": null
            },
            "BIBREF7": {
                "ref_id": "b7",
                "title": "Catching the long-tail: Extracting local news events from twitter",
                "authors": [
                    {
                        "first": "Puneet",
                        "middle": [],
                        "last": "Agarwal",
                        "suffix": ""
                    },
                    {
                        "first": "R",
                        "middle": [],
                        "last": "Vaithiyanathan",
                        "suffix": ""
                    },
                    {
                        "first": "Saurabh",
                        "middle": [],
                        "last": "Sharma",
                        "suffix": ""
                    },
                    {
                        "first": "G",
                        "middle": [],
                        "last": "Shroff",
                        "suffix": ""
                    }
                ],
                "year": 2012,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Puneet Agarwal, R. Vaithiyanathan, Saurabh Sharma, and G. Shroff. 2012. Catching the long-tail: Extract- ing local news events from twitter. In ICWSM. 8 https://getbootstrap.com/ 9 https://flask.palletsprojects.com/en/1.1.x/ 10 https://leafletjs.com/ 11 https://www.patrick-wied.at/static/heatmapjs/ 12 https://www.openstreetmap.org/",
                "links": null
            },
            "BIBREF8": {
                "ref_id": "b8",
                "title": "Exploring twitter as a source of an arabic dialect corpus",
                "authors": [
                    {
                        "first": "Areej",
                        "middle": [],
                        "last": "Alshutayri",
                        "suffix": ""
                    },
                    {
                        "first": "Erik",
                        "middle": [],
                        "last": "Atwell",
                        "suffix": ""
                    }
                ],
                "year": 2017,
                "venue": "International Journal of Computational Linguistics (IJCL)",
                "volume": "8",
                "issue": "2",
                "pages": "37--44",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Areej Alshutayri and Erik Atwell. 2017. Exploring twitter as a source of an arabic dialect corpus. In- ternational Journal of Computational Linguistics (IJCL), 8(2):37-44. This is an open access article under the terms of the Creative Commons Attribu- tion License (CC-BY).",
                "links": null
            },
            "BIBREF9": {
                "ref_id": "b9",
                "title": "The MADAR shared task on Arabic finegrained dialect identification",
                "authors": [
                    {
                        "first": "Houda",
                        "middle": [],
                        "last": "Bouamor",
                        "suffix": ""
                    },
                    {
                        "first": "Sabit",
                        "middle": [],
                        "last": "Hassan",
                        "suffix": ""
                    },
                    {
                        "first": "Nizar",
                        "middle": [],
                        "last": "Habash",
                        "suffix": ""
                    }
                ],
                "year": 2019,
                "venue": "Proceedings of the Fourth Arabic Natural Language Processing Workshop",
                "volume": "",
                "issue": "",
                "pages": "199--207",
                "other_ids": {
                    "DOI": [
                        "10.18653/v1/W19-4622"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Houda Bouamor, Sabit Hassan, and Nizar Habash. 2019. The MADAR shared task on Arabic fine- grained dialect identification. In Proceedings of the Fourth Arabic Natural Language Processing Work- shop, pages 199-207, Florence, Italy. Association for Computational Linguistics.",
                "links": null
            },
            "BIBREF10": {
                "ref_id": "b10",
                "title": "Crowd sourcing disaster management: The complex nature of twitter usage in padang indonesia",
                "authors": [
                    {
                        "first": "Kathleen",
                        "middle": [
                            "M"
                        ],
                        "last": "Carley",
                        "suffix": ""
                    },
                    {
                        "first": "Momin",
                        "middle": [],
                        "last": "Malik",
                        "suffix": ""
                    },
                    {
                        "first": "Peter",
                        "middle": [
                            "M"
                        ],
                        "last": "Landwehr",
                        "suffix": ""
                    },
                    {
                        "first": "J\u00fcrgen",
                        "middle": [],
                        "last": "Pfeffer",
                        "suffix": ""
                    },
                    {
                        "first": "Michael",
                        "middle": [],
                        "last": "Kowalchuck",
                        "suffix": ""
                    }
                ],
                "year": 2016,
                "venue": "Safety Science",
                "volume": "90",
                "issue": "",
                "pages": "48--61",
                "other_ids": {
                    "DOI": [
                        "10.1016/j.ssci.2016.04.002"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Kathleen M. Carley, Momin Malik, Peter M. Landwehr, J\u00fcrgen Pfeffer, and Michael Kowalchuck. 2016. Crowd sourcing disaster management: The com- plex nature of twitter usage in padang indonesia. Safety Science, 90:48 -61. Building Community Resilience to Global Hazards: A Sociotechnical Ap- proach.",
                "links": null
            },
            "BIBREF11": {
                "ref_id": "b11",
                "title": "A fine-grained annotated multi-dialectal Arabic corpus",
                "authors": [
                    {
                        "first": "Anis",
                        "middle": [],
                        "last": "Charfi",
                        "suffix": ""
                    },
                    {
                        "first": "Wajdi",
                        "middle": [],
                        "last": "Zaghouani",
                        "suffix": ""
                    },
                    {
                        "first": "Esraa",
                        "middle": [],
                        "last": "Syed Hassan Mehdi",
                        "suffix": ""
                    },
                    {
                        "first": "",
                        "middle": [],
                        "last": "Mohamed",
                        "suffix": ""
                    }
                ],
                "year": 2019,
                "venue": "Proceedings of the International Conference on Recent Advances in Natural Language Processing",
                "volume": "",
                "issue": "",
                "pages": "198--204",
                "other_ids": {
                    "DOI": [
                        "10.26615/978-954-452-056-4_023"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Anis Charfi, Wajdi Zaghouani, Syed Hassan Mehdi, and Esraa Mohamed. 2019. A fine-grained anno- tated multi-dialectal Arabic corpus. In Proceed- ings of the International Conference on Recent Ad- vances in Natural Language Processing (RANLP 2019), pages 198-204, Varna, Bulgaria. INCOMA Ltd.",
                "links": null
            },
            "BIBREF12": {
                "ref_id": "b12",
                "title": "You are where you tweet: a content-based approach to geo-locating twitter users",
                "authors": [
                    {
                        "first": "Zhiyuan",
                        "middle": [],
                        "last": "Cheng",
                        "suffix": ""
                    },
                    {
                        "first": "James",
                        "middle": [],
                        "last": "Caverlee",
                        "suffix": ""
                    },
                    {
                        "first": "Kyumin",
                        "middle": [],
                        "last": "Lee",
                        "suffix": ""
                    }
                ],
                "year": 2010,
                "venue": "CIKM '10",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Zhiyuan Cheng, James Caverlee, and Kyumin Lee. 2010. You are where you tweet: a content-based ap- proach to geo-locating twitter users. In CIKM '10.",
                "links": null
            },
            "BIBREF13": {
                "ref_id": "b13",
                "title": "Twitter earthquake detection: Earthquake monitoring in a social world",
                "authors": [
                    {
                        "first": "Paul",
                        "middle": [],
                        "last": "Earle",
                        "suffix": ""
                    },
                    {
                        "first": "Daniel",
                        "middle": [],
                        "last": "Bowden",
                        "suffix": ""
                    },
                    {
                        "first": "Michelle",
                        "middle": [],
                        "last": "Guy",
                        "suffix": ""
                    }
                ],
                "year": 2012,
                "venue": "Annals of geophysics = Annali di geofisica",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {
                    "DOI": [
                        "10.4401/ag-5364"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Paul Earle, Daniel Bowden, and Michelle Guy. 2012. Twitter earthquake detection: Earthquake monitor- ing in a social world. Annals of geophysics = Annali di geofisica, 54.",
                "links": null
            },
            "BIBREF14": {
                "ref_id": "b14",
                "title": "Arabic dialect identification in the context of bivalency and code-switching",
                "authors": [
                    {
                        "first": "Mahmoud",
                        "middle": [],
                        "last": "El-Haj",
                        "suffix": ""
                    },
                    {
                        "first": "Paul",
                        "middle": [],
                        "last": "Rayson",
                        "suffix": ""
                    },
                    {
                        "first": "Mariam",
                        "middle": [],
                        "last": "Aboelezz",
                        "suffix": ""
                    }
                ],
                "year": 2018,
                "venue": "Proceedings of the Eleventh International Conference on Language Resources and Evaluation (LREC-2018)",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Mahmoud El-Haj, Paul Rayson, and Mariam Aboelezz. 2018. Arabic dialect identification in the context of bivalency and code-switching. In Proceedings of the Eleventh International Conference on Language Resources and Evaluation (LREC-2018), Miyazaki, Japan. European Languages Resources Association (ELRA).",
                "links": null
            },
            "BIBREF15": {
                "ref_id": "b15",
                "title": "Asad: Arabic social media analytics and understanding",
                "authors": [
                    {
                        "first": "Sabit",
                        "middle": [],
                        "last": "Hassan",
                        "suffix": ""
                    },
                    {
                        "first": "Hamdy",
                        "middle": [],
                        "last": "Mubarak",
                        "suffix": ""
                    },
                    {
                        "first": "Ahmed",
                        "middle": [],
                        "last": "Abdelali",
                        "suffix": ""
                    },
                    {
                        "first": "Kareem",
                        "middle": [],
                        "last": "Darwish",
                        "suffix": ""
                    }
                ],
                "year": 2021,
                "venue": "Proceedings of the Software Demonstrations of the 16th Conference of the European Chapter of the Association for Computational Linguistics",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Sabit Hassan, Hamdy Mubarak, Ahmed Abdelali, and Kareem Darwish. 2021. Asad: Arabic social media analytics and understanding. In Proceedings of the Software Demonstrations of the 16th Conference of the European Chapter of the Association for Compu- tational Linguistics.",
                "links": null
            },
            "BIBREF16": {
                "ref_id": "b16",
                "title": "ALT at SemEval-2020 task 12: Arabic and English offensive language identification in social media",
                "authors": [
                    {
                        "first": "Sabit",
                        "middle": [],
                        "last": "Hassan",
                        "suffix": ""
                    },
                    {
                        "first": "Younes",
                        "middle": [],
                        "last": "Samih",
                        "suffix": ""
                    },
                    {
                        "first": "Hamdy",
                        "middle": [],
                        "last": "Mubarak",
                        "suffix": ""
                    },
                    {
                        "first": "Ahmed",
                        "middle": [],
                        "last": "Abdelali",
                        "suffix": ""
                    }
                ],
                "year": 2020,
                "venue": "Proceedings of the Fourteenth Workshop on Semantic Evaluation",
                "volume": "",
                "issue": "",
                "pages": "1891--1897",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Sabit Hassan, Younes Samih, Hamdy Mubarak, and Ahmed Abdelali. 2020a. ALT at SemEval-2020 task 12: Arabic and English offensive language iden- tification in social media. In Proceedings of the Fourteenth Workshop on Semantic Evaluation, pages 1891-1897, Barcelona (online). International Com- mittee for Computational Linguistics.",
                "links": null
            },
            "BIBREF17": {
                "ref_id": "b17",
                "title": "Ammar Rashed, and Shammur Absar Chowdhury. 2020b. ALT submission for OSACT shared task on offensive language detection",
                "authors": [
                    {
                        "first": "Sabit",
                        "middle": [],
                        "last": "Hassan",
                        "suffix": ""
                    },
                    {
                        "first": "Younes",
                        "middle": [],
                        "last": "Samih",
                        "suffix": ""
                    },
                    {
                        "first": "Hamdy",
                        "middle": [],
                        "last": "Mubarak",
                        "suffix": ""
                    },
                    {
                        "first": "Ahmed",
                        "middle": [],
                        "last": "Abdelali",
                        "suffix": ""
                    }
                ],
                "year": null,
                "venue": "Proceedings of the 4th Workshop on Open-Source Arabic Corpora and Processing Tools, with a Shared Task on Offensive Language Detection",
                "volume": "",
                "issue": "",
                "pages": "61--65",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Sabit Hassan, Younes Samih, Hamdy Mubarak, Ahmed Abdelali, Ammar Rashed, and Shammur Absar Chowdhury. 2020b. ALT submission for OSACT shared task on offensive language detection. In Pro- ceedings of the 4th Workshop on Open-Source Ara- bic Corpora and Processing Tools, with a Shared Task on Offensive Language Detection, pages 61-65,",
                "links": null
            },
            "BIBREF18": {
                "ref_id": "b18",
                "title": "European Language Resource Association",
                "authors": [
                    {
                        "first": "France",
                        "middle": [],
                        "last": "Marseille",
                        "suffix": ""
                    }
                ],
                "year": null,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Marseille, France. European Language Resource As- sociation.",
                "links": null
            },
            "BIBREF19": {
                "ref_id": "b19",
                "title": "Tweets from justin bieber's heart: The dynamics of the location field in user profiles",
                "authors": [
                    {
                        "first": "Brent",
                        "middle": [],
                        "last": "Hecht",
                        "suffix": ""
                    },
                    {
                        "first": "Lichan",
                        "middle": [],
                        "last": "Hong",
                        "suffix": ""
                    },
                    {
                        "first": "Bongwon",
                        "middle": [],
                        "last": "Suh",
                        "suffix": ""
                    },
                    {
                        "first": "Ed",
                        "middle": [
                            "H"
                        ],
                        "last": "Chi",
                        "suffix": ""
                    }
                ],
                "year": 2011,
                "venue": "Proceedings of the SIGCHI Conference on Human Factors in Computing Systems, CHI '11",
                "volume": "",
                "issue": "",
                "pages": "237--246",
                "other_ids": {
                    "DOI": [
                        "10.1145/1978942.1978976"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Brent Hecht, Lichan Hong, Bongwon Suh, and Ed H. Chi. 2011. Tweets from justin bieber's heart: The dynamics of the location field in user profiles. In Proceedings of the SIGCHI Conference on Hu- man Factors in Computing Systems, CHI '11, page 237-246, New York, NY, USA. Association for Computing Machinery.",
                "links": null
            },
            "BIBREF20": {
                "ref_id": "b20",
                "title": "A hierarchical location prediction neural network for twitter user geolocation",
                "authors": [
                    {
                        "first": "Binxuan",
                        "middle": [],
                        "last": "Huang",
                        "suffix": ""
                    },
                    {
                        "first": "Kathleen",
                        "middle": [],
                        "last": "Carley",
                        "suffix": ""
                    }
                ],
                "year": 2019,
                "venue": "Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP)",
                "volume": "",
                "issue": "",
                "pages": "4732--4742",
                "other_ids": {
                    "DOI": [
                        "10.18653/v1/D19-1480"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Binxuan Huang and Kathleen Carley. 2019. A hierar- chical location prediction neural network for twitter user geolocation. In Proceedings of the 2019 Con- ference on Empirical Methods in Natural Language Processing and the 9th International Joint Confer- ence on Natural Language Processing (EMNLP- IJCNLP), pages 4732-4742, Hong Kong, China. As- sociation for Computational Linguistics.",
                "links": null
            },
            "BIBREF21": {
                "ref_id": "b21",
                "title": "A few chirps about twitter",
                "authors": [
                    {
                        "first": "Balachander",
                        "middle": [],
                        "last": "Krishnamurthy",
                        "suffix": ""
                    },
                    {
                        "first": "Phillipa",
                        "middle": [],
                        "last": "Gill",
                        "suffix": ""
                    },
                    {
                        "first": "Martin",
                        "middle": [],
                        "last": "Arlitt",
                        "suffix": ""
                    }
                ],
                "year": 2008,
                "venue": "",
                "volume": "",
                "issue": "",
                "pages": "19--24",
                "other_ids": {
                    "DOI": [
                        "10.1145/1397735.1397741"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Balachander Krishnamurthy, Phillipa Gill, and Martin Arlitt. 2008. A few chirps about twitter. pages 19- 24.",
                "links": null
            },
            "BIBREF22": {
                "ref_id": "b22",
                "title": "Home location identification of twitter users",
                "authors": [
                    {
                        "first": "Jalal",
                        "middle": [],
                        "last": "Mahmud",
                        "suffix": ""
                    },
                    {
                        "first": "Jeffrey",
                        "middle": [],
                        "last": "Nichols",
                        "suffix": ""
                    },
                    {
                        "first": "Clemens",
                        "middle": [],
                        "last": "Drews",
                        "suffix": ""
                    }
                ],
                "year": 2014,
                "venue": "ACM Transactions on Intelligent Systems and Technology",
                "volume": "5",
                "issue": "3",
                "pages": "1--21",
                "other_ids": {
                    "DOI": [
                        "10.1145/2528548"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Jalal Mahmud, Jeffrey Nichols, and Clemens Drews. 2014. Home location identification of twitter users. ACM Transactions on Intelligent Systems and Tech- nology, 5(3):1-21.",
                "links": null
            },
            "BIBREF23": {
                "ref_id": "b23",
                "title": "Unifying text, metadata, and user network representations with a neural network for geolocation prediction",
                "authors": [
                    {
                        "first": "Yasuhide",
                        "middle": [],
                        "last": "Miura",
                        "suffix": ""
                    },
                    {
                        "first": "Motoki",
                        "middle": [],
                        "last": "Taniguchi",
                        "suffix": ""
                    },
                    {
                        "first": "Tomoki",
                        "middle": [],
                        "last": "Taniguchi",
                        "suffix": ""
                    },
                    {
                        "first": "Tomoko",
                        "middle": [],
                        "last": "Ohkuma",
                        "suffix": ""
                    }
                ],
                "year": 2017,
                "venue": "Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics",
                "volume": "1",
                "issue": "",
                "pages": "1260--1272",
                "other_ids": {
                    "DOI": [
                        "10.18653/v1/P17-1116"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Yasuhide Miura, Motoki Taniguchi, Tomoki Taniguchi, and Tomoko Ohkuma. 2017. Unifying text, meta- data, and user network representations with a neural network for geolocation prediction. In Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 1260-1272, Vancouver, Canada. Association for Computational Linguistics.",
                "links": null
            },
            "BIBREF24": {
                "ref_id": "b24",
                "title": "Arabic to english person name transliteration using twitter",
                "authors": [
                    {
                        "first": "Hamdy",
                        "middle": [],
                        "last": "Mubarak",
                        "suffix": ""
                    },
                    {
                        "first": "Ahmed",
                        "middle": [],
                        "last": "Abdelali",
                        "suffix": ""
                    }
                ],
                "year": 2016,
                "venue": "Proceedings of the Tenth International Conference on Language Resources and Evaluation (LREC'16)",
                "volume": "",
                "issue": "",
                "pages": "351--355",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Hamdy Mubarak and Ahmed Abdelali. 2016. Arabic to english person name transliteration using twitter. In Proceedings of the Tenth International Conference on Language Resources and Evaluation (LREC'16), pages 351-355.",
                "links": null
            },
            "BIBREF25": {
                "ref_id": "b25",
                "title": "Spam detection on arabic twitter",
                "authors": [
                    {
                        "first": "Hamdy",
                        "middle": [],
                        "last": "Mubarak",
                        "suffix": ""
                    },
                    {
                        "first": "Ahmed",
                        "middle": [],
                        "last": "Abdelali",
                        "suffix": ""
                    },
                    {
                        "first": "Sabit",
                        "middle": [],
                        "last": "Hassan",
                        "suffix": ""
                    },
                    {
                        "first": "Kareem",
                        "middle": [],
                        "last": "Darwish",
                        "suffix": ""
                    }
                ],
                "year": 2020,
                "venue": "Social Informatics",
                "volume": "",
                "issue": "",
                "pages": "237--251",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Hamdy Mubarak, Ahmed Abdelali, Sabit Hassan, and Kareem Darwish. 2020. Spam detection on ara- bic twitter. In Social Informatics, pages 237-251, Cham. Springer International Publishing.",
                "links": null
            },
            "BIBREF26": {
                "ref_id": "b26",
                "title": "Using twitter to collect a multi-dialectal corpus of Arabic",
                "authors": [
                    {
                        "first": "Hamdy",
                        "middle": [],
                        "last": "Mubarak",
                        "suffix": ""
                    },
                    {
                        "first": "Kareem",
                        "middle": [],
                        "last": "Darwish",
                        "suffix": ""
                    }
                ],
                "year": 2014,
                "venue": "Proceedings of the EMNLP 2014 Workshop on Arabic Natural Language Processing (ANLP)",
                "volume": "",
                "issue": "",
                "pages": "1--7",
                "other_ids": {
                    "DOI": [
                        "10.3115/v1/W14-3601"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Hamdy Mubarak and Kareem Darwish. 2014. Using twitter to collect a multi-dialectal corpus of Arabic. In Proceedings of the EMNLP 2014 Workshop on Arabic Natural Language Processing (ANLP), pages 1-7, Doha, Qatar. Association for Computational Linguistics.",
                "links": null
            },
            "BIBREF27": {
                "ref_id": "b27",
                "title": "Adult content detection on arabic twitter: Analysis and experiments",
                "authors": [
                    {
                        "first": "Hamdy",
                        "middle": [],
                        "last": "Mubarak",
                        "suffix": ""
                    },
                    {
                        "first": "Sabit",
                        "middle": [],
                        "last": "Hassan",
                        "suffix": ""
                    },
                    {
                        "first": "Ahmed",
                        "middle": [],
                        "last": "Abdelali",
                        "suffix": ""
                    }
                ],
                "year": 2021,
                "venue": "Proceedings of the Sixth Arabic Natural Language Processing Workshop",
                "volume": "",
                "issue": "",
                "pages": "",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Hamdy Mubarak, Sabit Hassan, and Ahmed Abdelali. 2021. Adult content detection on arabic twitter: Analysis and experiments. In Proceedings of the Sixth Arabic Natural Language Processing Work- shop.",
                "links": null
            },
            "BIBREF28": {
                "ref_id": "b28",
                "title": "Twitter user geolocation using a unified text and network prediction model",
                "authors": [
                    {
                        "first": "Afshin",
                        "middle": [],
                        "last": "Rahimi",
                        "suffix": ""
                    },
                    {
                        "first": "Trevor",
                        "middle": [],
                        "last": "Cohn",
                        "suffix": ""
                    },
                    {
                        "first": "Timothy",
                        "middle": [],
                        "last": "Baldwin",
                        "suffix": ""
                    }
                ],
                "year": 2015,
                "venue": "Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics and the 7th International Joint Conference on Natural Language Processing",
                "volume": "2",
                "issue": "",
                "pages": "630--636",
                "other_ids": {
                    "DOI": [
                        "10.3115/v1/P15-2104"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Afshin Rahimi, Trevor Cohn, and Timothy Baldwin. 2015. Twitter user geolocation using a unified text and network prediction model. In 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), pages 630-636, Beijing, China. Association for Computational Linguistics.",
                "links": null
            },
            "BIBREF29": {
                "ref_id": "b29",
                "title": "Semi-supervised user geolocation via graph convolutional networks",
                "authors": [
                    {
                        "first": "Afshin",
                        "middle": [],
                        "last": "Rahimi",
                        "suffix": ""
                    },
                    {
                        "first": "Trevor",
                        "middle": [],
                        "last": "Cohn",
                        "suffix": ""
                    },
                    {
                        "first": "Timothy",
                        "middle": [],
                        "last": "Baldwin",
                        "suffix": ""
                    }
                ],
                "year": 2018,
                "venue": "Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics",
                "volume": "1",
                "issue": "",
                "pages": "2009--2019",
                "other_ids": {
                    "DOI": [
                        "10.18653/v1/P18-1187"
                    ]
                },
                "num": null,
                "urls": [],
                "raw_text": "Afshin Rahimi, Trevor Cohn, and Timothy Baldwin. 2018. Semi-supervised user geolocation via graph convolutional networks. In Proceedings of the 56th Annual Meeting of the Association for Computa- tional Linguistics (Volume 1: Long Papers), pages 2009-2019, Melbourne, Australia. Association for Computational Linguistics.",
                "links": null
            },
            "BIBREF30": {
                "ref_id": "b30",
                "title": "Fine-grained Arabic dialect identification",
                "authors": [
                    {
                        "first": "Mohammad",
                        "middle": [],
                        "last": "Salameh",
                        "suffix": ""
                    },
                    {
                        "first": "Houda",
                        "middle": [],
                        "last": "Bouamor",
                        "suffix": ""
                    },
                    {
                        "first": "Nizar",
                        "middle": [],
                        "last": "Habash",
                        "suffix": ""
                    }
                ],
                "year": 2018,
                "venue": "Proceedings of the 27th International Conference on Computational Linguistics",
                "volume": "",
                "issue": "",
                "pages": "1332--1344",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Mohammad Salameh, Houda Bouamor, and Nizar Habash. 2018. Fine-grained Arabic dialect identi- fication. In Proceedings of the 27th International Conference on Computational Linguistics, pages 1332-1344, Santa Fe, New Mexico, USA. Associ- ation for Computational Linguistics.",
                "links": null
            },
            "BIBREF31": {
                "ref_id": "b31",
                "title": "The Arabic online commentary dataset: an annotated dataset of informal Arabic with high dialectal content",
                "authors": [
                    {
                        "first": "Omar",
                        "middle": [
                            "F"
                        ],
                        "last": "Zaidan",
                        "suffix": ""
                    },
                    {
                        "first": "Chris",
                        "middle": [],
                        "last": "Callison-Burch",
                        "suffix": ""
                    }
                ],
                "year": 2011,
                "venue": "Proceedings of the 49th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies",
                "volume": "",
                "issue": "",
                "pages": "37--41",
                "other_ids": {},
                "num": null,
                "urls": [],
                "raw_text": "Omar F. Zaidan and Chris Callison-Burch. 2011. The Arabic online commentary dataset: an annotated dataset of informal Arabic with high dialectal con- tent. In Proceedings of the 49th Annual Meeting of the Association for Computational Linguistics: Hu- man Language Technologies, pages 37-41, Portland, Oregon, USA. Association for Computational Lin- guistics.",
                "links": null
            }
        },
        "ref_entries": {
            "FIGREF0": {
                "uris": null,
                "text": "manually labeled about 3K users for their countries of origin. Some works targeted region level classification of Arabic dialects. Zaidan and Callison-Burch (2011) collected a 52M word dataset from newspapers and annotated them for dialects of 5 Arab regions, namely, Maghrebi, Egyptian, Levantine, Gulf, and Iraqi. Alshutayri and Atwell (2017) and El-Haj et al. (2018) annotate texts from the five regions. Some other works target city level classification of Arabic dialects. Salameh et al. (2018) present a parallel corpus of dialects of 25 cities and Abdul-Mageed et al. (2018) construct a dataset representing dialects of 29 cities from 10 Arab countries. Charfi et al. (2019) presents a corpus of 5M tweets of about 3K users from 17 different Arab countries.",
                "num": null,
                "type_str": "figure"
            },
            "FIGREF1": {
                "uris": null,
                "text": "GeoNames examples: First place is included and second place is excluded from our dataset",
                "num": null,
                "type_str": "figure"
            },
            "FIGREF2": {
                "uris": null,
                "text": "Country distribution of Twitter users",
                "num": null,
                "type_str": "figure"
            },
            "FIGREF3": {
                "uris": null,
                "text": "Country map of Twitter users",
                "num": null,
                "type_str": "figure"
            },
            "FIGREF4": {
                "uris": null,
                "text": "Top locations in 4 countries: SA, EG, SY and DZ in order \u2022 Remove all Arabic diacritics \u2022 Remove all punctuation marks and emojis \u2022 Convert all Latin letters to lowercase \u2022 Normalize different shapes of Arabic Hamza, final Alif Maqsoura and Taa Marbouta letters to plain Alif, Yaa and Haa letters respectively",
                "num": null,
                "type_str": "figure"
            },
            "FIGREF5": {
                "uris": null,
                "text": "Online interface for mapping user locations to countries (https://asad.qcri.org)",
                "num": null,
                "type_str": "figure"
            },
            "TABREF2": {
                "text": "",
                "html": null,
                "type_str": "table",
                "content": "<table/>",
                "num": null
            },
            "TABREF4": {
                "text": "Annotation clues and special cases",
                "html": null,
                "type_str": "table",
                "content": "<table/>",
                "num": null
            },
            "TABREF5": {
                "text": "Evaluation of different datasets",
                "html": null,
                "type_str": "table",
                "content": "<table><tr><td>6.3 Classification Models</td></tr><tr><td>geopy baseline The geopy library acts as our</td></tr><tr><td>baseline. We call library with Twitter user loca-</td></tr><tr><td>tions and extract countries they are mapped to by</td></tr><tr><td>the library.</td></tr><tr><td>Support Vector Machines (SVMs) SVMs have</td></tr><tr><td>traditionally been used for many classification</td></tr><tr><td>tasks. Even in recent Arabic text classification tasks</td></tr><tr><td>such as offensive language identification</td></tr></table>",
                "num": null
            }
        }
    }
}