PaperlayerView exact revision on arXiv

Source-linked AI summary

Tweets on the road

Maxime Lenormand, Antònia Tugores, Pere Colet, José J. Ramasco

arXiv:1405.5070v1physics.soc-phcs.CYcs.SI

TL;DR

The paper asks how geo-located tweets can characterize use of European transport networks. It identifies tweets associated with highways and railways, then finds substantial cross-country variation in coverage and transport-modality patterns, with penetration rates explaining only part of the differences.

  • Problem

    The paper examines how geo-located Twitter data can characterize transport-network use and mobility patterns across Europe.

    Source: §I · p. 1

  • Method

    The authors identify geo-located tweets associated with highway and railway segments across 39 European countries and compare network coverage and transport-use patterns.

    Source: §I · p. 1, A. Datasets · p. 1

  • Results

    Twitter penetration explains only part of the cross-country coverage differences, while transport-modality patterns are heterogeneous across Europe.

    Source: §I · p. 1, §1 · p. 7

  • Takeaways & Limitations

    Geo-located tweets can be used to examine transport use, with road-tweet counts positively correlated with highway AADT in France and the UK.

    Source: §I · p. 1

    The analysis covers 39 European countries and tweets from September 2012 to November 2013.

    Source: §I · p. 1

Abstract

from arXiv · show

The pervasiveness of mobile devices, which is increasing daily, is generating a vast amount of geo-located data allowing us to gain further insights into human behaviors. In particular, this new technology enables users to communicate through mobile social media applications, such as Twitter, anytime and anywhere. Thus, geo-located tweets offer the possibility to carry out in-depth studies on human mobility. In this paper, we study the use of Twitter in transportation by identifying tweets posted from roads and rails in Europe between September 2012 and November 2013. We compute the percentage of highway and railway segments covered by tweets in 39 countries. The coverages are very different from country to country and their variability can be partially explained by differences in Twitter penetration rates. Still, some of these differences might be related to cultural factors regarding mobility habits and interacting socially online. Analyzing particular road sectors, our results show a positive correlation between the number of tweets on the road and the Average Annual Daily Traffic on highways in France and in the UK. Transport modality can be studied with these data as well, for which we discover very heterogeneous usage patterns across the continent.

I. INTRODUCTION

The paper examines how geo-located Twitter activity appears on European transport networks and whether it can illuminate transport use. It analyzes a database of more than five million tweets across 39 countries, alongside highway and railway network data.

I. INTRODUCTION

  • The study identifies tweets posted on roads and railways across 39 European countries between September 2012 and November 2013.It explores transport-network use and compares the share of highway and railway segments covered by tweets country by country.Source: §I · p. 1
  • Tweets cover different proportions of transport networks across countries, and Twitter penetration rates do not fully explain those differences.The paper suggests cultural diversity may also be related to country-level variation.Source: §I · p. 1

A. Datasets

  • The dataset contains 5,219,539 geo-located tweets from 1,477,263 users, with western Europe better represented than eastern Europe.Tweets were gathered through the general data streaming with the Twitter API.Source: A. Datasets · p. 1
  • The analyzed networks include main highways, but not rural roads, and main railway lines; they were extracted from OpenStreetMap.The networks total 274,365 km of highways and 451,475 km of railways.Source: A. Datasets · p. 1

B. Identify the tweets on the road

Tweets are assigned to nearby highway or railway segments, and a segment counts as covered when at least one tweet is assigned to it.

B. Identify the tweets on the road

  • Tweets within 20 meters of a highway or railway are assigned to the closest segment, which counts as covered if it has at least one tweet.Coverage is calculated as the percentage of road or rail segments meeting that condition.Source: B. Identify the tweets on the road · p. 4

C. Ethics statement

The analysis uses public online data that were anonymized and aggregated, and the authors state that it followed local data protection laws.

C. Ethics statement

  • The public Twitter and OpenStreetMap data were anonymized and aggregated before analysis, which the authors say complied with local data protection laws.Source: C. Ethics statement · p. 4

III. RESULTS

Geo-located Twitter penetration varies across European countries, with lower rates in central Europe and no clear relationship to GDP per capita in this regional sample.

1. Penetration rate

  • Geo-located Twitter penetration is not uniform across Europe and is lower in central European countries.Penetration is defined as Twitter users divided by country inhabitants.Source: §1 · p. 4
  • No clear correlation appears between European countries’ Twitter penetration rates and GDP per capita.The passage notes that this does not conflict with a positive worldwide correlation reported in earlier studies, because countries from different continents cluster together.Source: §1 · p. 4
  • Figure 4 presents degree distributions for five sampled countries and box plots of fitted exponents and R² values across 39 countries.Source: Figure 4 · p. 5

2. Social network

The paper builds country-level Twitter interaction networks from user replies and finds broadly similar network structure across the sampled countries. The networks are sparse, with similar degree distributions and clustering values among the most populated countries.

2. Social network

  • Country networks connect users who exchanged replies, with each user assigned to the country where most of their geo-located tweets were posted.Replies are messages designed to answer another user's tweet and can relate to more intense social relations.Source: §2 · p. 5
  • Across 39 countries, fitted degree distributions are very similar, while maximum degree varies with network size.Five randomly selected countries show similar slopes that can be fitted with a power-law distribution.Source: §2 · p. 5
  • The networks are sparse, and clustering for the most populated countries ranges from 0.02 to 0.04.The authors attribute sparsity to retaining only users with geo-located tweets and connections formed by replies.Source: §2 · p. 5

1. Dependence between coverage and penetration rate

Tweet coverage varies across European road and rail networks, and the paper examines how this variation relates to geo-located Twitter penetration. Coverage is higher for highways than railways overall, differs by country, and is not fully explained by penetration rates.

1. Dependence between coverage and penetration rate

  • 39% of European highway kilometers and 24% of railway kilometers are covered by tweets.The paper suggests the higher highway coverage may relate to lower annual passenger-kilometers on rail.Source: §1 · p. 5
  • Western European countries generally have better coverage than eastern European countries, except Turkey and, to a lesser extent, Russia.Figure 5 maps tweet locations and covered road and rail segments, with covered segments shown in red.Source: §1 · p. 5, Figure 5 · p. 6
  • The UK and Netherlands lead the country rankings: highway coverage is 97% and 89%, and railway coverage is 80% and 78%, respectively.Moldavia has the lowest reported coverage, at 2.5% for highways and 1% for railways.Source: §1 · p. 6
  • Figure 6 ranks the top 20 countries separately by highway and railway coverage.The UK and Netherlands are the top two in both rankings.Source: §1 · p. 6, Figure 6 · p. 7
  • Coverage of both networks is generally positively correlated with geo-located Twitter penetration, but penetration differences do not fully explain country coverage differences.The reported mean absolute errors are around 7.5% for the first two relationships and around 6.5% for the relationship between highway and railway coverage.Source: §1 · pp. 6–7

2. Differences across European countries

Twitter coverage of European transport networks varies substantially across countries and is only partly explained by geolocated-Twitter penetration rates. The section relates these differences to country-specific transport and usage patterns, including rail travel and road traffic.

2. Differences across European countries

  • Spain’s highway coverage is 77%, compared with 55% in France, despite similar highway infrastructure, mobile-data fares, and accessibility.The authors suggest cultural differences in Twitter use in transport as a possible factor.Source: §2 · p. 7
  • UK has more than twice Ireland’s proportion of transport tweets, while its highway and railway coverages are each one and a half times higher.Source: §2 · p. 7
  • With similar penetration rates, the Netherlands has three times Turkey’s railway coverage despite both having high highway coverage; Belgium also has three times Norway’s highway coverage despite Norway’s higher transport-tweet proportion.The authors suggest traveler economic differences as one possible explanation for the Netherlands–Turkey contrast, and low-traffic northern roads as a possible factor in Norway.Source: §2 · p. 7
  • Figure 7’s log-log regression slopes are about 0.6 for highway coverage versus penetration, 0.8 for railway coverage versus penetration, and 1 for railway versus highway coverage.Source: Figure 7 · p. 8
  • Countries in North and Central Europe have a higher proportion of road than rail tweets, and countries with lower rail-tweet proportions also have lower rail passenger-kilometers overall.Source: §2 · pp. 7, 9
  • With about 40% highway coverage in both countries, Switzerland has about 47% railway coverage versus Estonia’s 11%; railway-coverage differences matched rail passenger-kilometer differences in 10 of 13 country pairs.The pairs were selected for highway coverage above 20% and differences below five percentage points.Source: §2 · p. 9, Table I · p. 9

C. Average Annual Daily Traffic

Highway tweet counts are positively correlated with average annual daily traffic in France and the United Kingdom, although the correlations are modest. The authors also find that high-traffic segments with few tweets tend to be shorter than other segments.

C. Average Annual Daily Traffic

  • The comparison covers 1,974 French and 877 UK highway segments; tweet counts were associated with segments using a 20-meter distance threshold.Average segment lengths were 5.8 km in France and 3.6 km in the UK.Source: C. Average Annual Daily Traffic · pp. 9–10
  • Pearson correlations between AADT and road-tweet counts are around 0.5 in France and 0.3 in the UK.Many highway segments had high AADT but very few tweets.Source: C. Average Annual Daily Traffic · p. 10
  • High-AADT segments with very few tweets were significantly shorter on average than the other segments: 5.2 versus 11.3 km in France and 2.5 versus 5.9 km in the UK.The comparison used a Welch two-sample t-test.Source: C. Average Annual Daily Traffic · p. 10
  • The authors suggest shorter time available to tweet on shorter segments as one possible explanation, while segment type and congestion may also influence the results.Source: C. Average Annual Daily Traffic · p. 10

IV. DISCUSSION

The discussion reports that the study investigated Twitter use in European transport networks.

IV. DISCUSSION

  • The discussion frames the work as an investigation of Twitter use in European transport networks.Source: §IV · p. 10

APPENDIX

The appendix provides country-level maps of geo-located tweets on roads and railways for seven European countries.

APPENDIX

Related papers

Loading 1405.5070v1…