Introduction
Literature review
Young adults’ declining automobile orientation
Socio-demographic features (i.e. income, gender, car access) | Built environment (i.e. population density, public transport access) | Values and attitudes (i.e. environment awareness, attitude towards cars) | ICT (i.e. smart phone and social media use) | Graduated Driving License |
---|---|---|---|---|
McDonald and Trowbridge (2009) Raimond and Milthorpe (2010) Williams (2011) Licaj et al. (2012) Kuhnimhof et al. (2012) Sivak and Schoettle (2012) Delbosc and Currie (2013) Delbosc and Currie (2014a) Delbosc and Currie (2014b) Le Vine et al. (2014b) Le Vine and Polak (2014) Schoettle and Sivak (2014) Tefft et al. (2014) Brown and Handy (2015) Ciari and Axhausen (2015) Curry et al. (2015) Baradaran et al. (2016) Hjorthol (2016) Delbosc and Nakanishi (2017) Thigpen and Handy (2018) Rérat (2018) Bayart et al. (2020) Vaca et al. (2020) | McDonald and Trowbridge (2009) Raimond and Milthorpe (2010) Williams (2011) Licaj et al. (2012) Sivak and Schoettle (2012) Delbosc and Currie (2013) Delbosc and Currie (2014b) Le Vine et al. (2014b) Le Vine and Polak (2014) Schoettle and Sivak (2014) Brown and Handy (2015) Baradaran et al. (2016) Hjorthol (2016) Thigpen and Handy (2018) Rérat (2018) Bayart et al. (2020) Vaca et al. (2020) | Williams (2011) Delbosc and Currie (2013) Delbosc and Currie (2014c) Le Vine et al. (2014a) Schoettle and Sivak (2014) Brown and Handy (2015) Fylan and Caveney (2018) Thigpen and Handy (2018) | Sivak and Schoettle (2012) Delbosc and Currie (2013) Delbosc and Currie (2014c) Delbosc and Currie (2014b) Le Vine et al. (2014b) Schoettle and Sivak (2014) Brown and Handy (2015) Thigpen and Handy (2018) | Raimond and Milthorpe (2010) Sivak and Schoettle (2011) Delbosc and Currie (2013) Tefft et al. (2014) Thigpen and Handy (2018) |
Symbolic/affective motives for car use
Sentiment analysis
Study | Objective | Data source | Geography | Algorithm/software used | Key findings |
---|---|---|---|---|---|
Qi et al. (2020) | A framework to extract and analyze public opinions on transport service | Twitter | Miami-Dade country, US | AFINN | After introducing sentiment variables, the prediction accuracy can rise 18.6% and reach ~ 90% |
Kinra et al. (2020) | Public opinion about the adoption of autonomous vehicles | Twitter | Denmark | SentiStrengh | Text analytics can be used as a complement to surveys Safety, labor participation and congestion are the most important concerns |
Mondschein et al. (2020) | Customer sentiment towards parking | Yelp | Phoenix, US | Lexicon-based algorithm | Sentiment about parking is in general negative Parking sentiment is part of the overall perception of customers toward a business Districts with more parking spaces per business tend to have more positive parking sentiment Parking is viewed more positively when shared parking facilities are provided |
El-Diraby et al. (2019) | Customers’ satisfaction on transit service | Twitter | Vancouver, Canada | SentiStrength | Sentiment is in general negative Sentiments toward disruption, especially those related to public safety incidents, showed lower levels of negative sentiment The sentiment of the sub-network of the most influential players closely matched the topics and sentiment of the full network |
Mendez et al. (2019) | An approach to capture user satisfaction with public transport | Twitter | Santiago, Chile | SentiStrength | The amount of bus stops and bus services covered by the proposed approach is larger than survey data The proposed approach is effective in diagnosing problems in a timely manner |
Pratt et al. (2019) | Public opinion regarding ridesharing service | Twitter | US | Aylien | The number of negative tweets outweighs the number of positive ones about the service characteristics (like routing and travel time) Most tweets about other passengers feature “humor” about other passengers |
Rahim Taleqani et al. (2019) | Public opinion regarding dockless bikesharing | Twitter | Multiple countries (primarily US) | Logistic regression, support vector machines, and naïve Bayes | The dockless bikesharing system receives more positive sentiments than negative ones The mostly mentioned sub-topics relevant to dockless bikesharing are electric scooters, private e-hailing companies, and blockage of sidewalks |
Haghighi et al. (2018) | A framework to analyze public opinion on transit service quality | Twitter | Utah, US | Rsentiment proposed by (Bose et al. 2017) | The number of negative tweets is greater on weekends than weekdays Most negative tweets are related to transit routes with higher ridership There is potential to use social media data to analyze transit service quality |
Kulkarni et al. (2018) | A system that can analyze public opinion on transport | Twitter | California, US | Valence Aware Dictionary and sEntiment Reasoner (VADER) | The quality of the system depends on the size of the dataset, the number of topics that are specific to the topic modelling algorithm, and the positive/negative thresholds of the sentiment analysis algorithm |
Pai and Liu (2018) | Predict vehicle sales by sentiment analysis | Twitter | US | SentiStrength | Both social media sentiment and stock values have predictive power to forecast monthly total vehicle sales |
Ali et al. (2017) | Using sentiment analysis to monitor transportation activities | Twitter | Not mentioned | SentiWordNet | The proposed approach can determine real-time traffic congestion mapping |
Baj-Rogowska (2017) | Public opinion regarding Uber (ridehailing) | Facebook | Not mentioned | ProSuite | Sentiment analysis reflected events that affected the company’s reputation |
Wijnhoven and Plant (2017) | Predict car sales by sentiment analysis | Coosto, Twitter, Facebook, LinkedIn, YouTube, Google + , Hyves, Instagram and Pinterest | The Netherlands | Coosto | Social media sentiments have little predictive power towards car sales, while Google Trends data and social mention volume have significant predictive power |
Effendy et al. (2016) | Public opinion regarding public transport | Twitter | Indonesian | Support vector machines | The accuracy of sentiment analysis using support vector machine can reach 78% |
Fraedrich and Lenz (2016) | Public interest in autonomous driving | Online comments on newspaper articles | Germany and US | Qualitative content analysis | Response to autonomous driving in different countries and different types of media is different Sentiment towards autonomous driving is generally positive, however the authors report finding some negative sentiments |
Giancristofaro and Panangadan (2016) | Public opinion of the California Department of Transportation | Instagram | California, US | Support vector machines, naïve Bayes, and random forests | The precision of sentiment analysis can be improved if images and texts are combined |
Hao et al. (2016) | Public opinion towards the I-710 Corridor Project | Twitter | California, US | API based on a naïve Bayes classifier | There are increasing twitter users participating the I-710 Corridor Project over time The number of comments from personal twitter accounts is positively correlated to the number of tweets from the organization account Twitter users are more likely to sent positive comments in the morning and negative comments in the afternoon towards I-710 Project Twitter users are more positive towards “Freeway Tunnel”, “Light Rail Transit” and improving the existing infrastructure, and more negative towards “Rapid Bus Transit” |
Das et al. (2015) | Users’ sentiment toward bikesharing | Twitter | Washington DC, US | Lexicon-based algorithm | Most people view the bikesharing system positively |
Luong and Houston (2015) | Public opinion on light rail service | Twitter | Los Angeles, US | Lexicon-based algorithm | Commuters mainly retweeted from other individuals or transit agencies, while schools and firms did not have strong retweet connections The Red Line was associated with the most positive tweets whereas the Blue Line has the most negative sentiments (of Los Angeles’ heavy rail lines) |
Schweitzer (2014) | Public opinion regarding public transit | Twitter | US | Lexicon-based algorithm | Public transit receives the most negative comments Transit companies that respond to other social media users receive statistically more positive sentiments |
Collins et al. (2013) | Transit riders’ satisfaction | Twitter | Chicago, US | SentiStrength | Riders express negative sentiments more than positive sentiments when an event occurs When an accident occurs, unusually high total social media engagement also occurs |
Cultural evolution in popular music
Data
Songs
Automobile reference tokens
-
Cars (general)
-
Car brands
-
Car parts (see listing of observed car-part reference in “Appendix 1”)
-
Car passenger travel
-
Driving
-
Stationary cars (as opposed to driving)
-
Taxi/hitching a ride
-
Traffic conditions
Results
Popular music trends, from 1950 to 2010s
Frequency of automobile references
-
Fig. 7 panel (a) shows the change of tokens associated with cars (general) and driving. In general, tokens associated with these two criteria are consistently high in all these years, and there are no major time trends.
-
Fig. 7 panel (b) shows the change of tokens associated with car parts and car brand. There is an increasing trend over time in the frequency of tokens associated with these two criteria.
-
Fig. 7 panel (d) shows the change of tokens associated with taxi/hitching a ride and car passenger travel. They are mentioned at the lowest frequency among all groups, and no clear trend over time is observed.
Sentiment towards cars over time
Alchemy | Joy Watson | Human analyst 1a | Human analyst 2 | |
---|---|---|---|---|
Alchemy | 1.0 | 0.44 | 0.16 | 0.13 |
Joy Watson | 1.0 | 0.13 | 0.09 | |
Human analyst 1 | 1.0 | 0.60 | ||
Human analyst 2 | 1.0 |
Alchemy | Joy Watson | Human analyst 1 | Human analyst 2 | |||||
---|---|---|---|---|---|---|---|---|
Automobile references | All lyrics | Automobile references | All lyrics | Automobile references | All lyrics | Automobile references | All lyrics | |
Pearson’s correlation between sentiment and year | − 0.09 (p < 0.01) | − 0.20 (p < 0.01) | − 0.16 (p < 0.01) | − 0.22 (p < 0.01) | 0.12 (p < 0.01) | − 0.04 (p = 0.28) | 0.17 (p < 0.01) | 0.03 (p = 0.46) |
Regression analysis
-
year each song was published,
-
the gender of the artist (1 = female, or the percentage of members that are female in the case of multi-member group artists),
-
the classification of the token (as presented in “Automobile reference tokens” section), and
-
the genre of the songs as independent variables.
Alchemy | Joy Watson | Human analyst 1 | Human analyst 2 | |||||
---|---|---|---|---|---|---|---|---|
Coefficients | p value | Coefficients | p value | Coefficients | p value | Coefficients | p value | |
Constant | 7.92 | 0.01 | 6.42 | < 0.01 | 0.55 | < 0.01 | 0.43 | < 0.01 |
Year | − 0.04 | < 0.01 | − 0.03 | < 0.01 | * | * | ||
Percentage of female artists | * | * | − 0.31 | < 0.01 | * | |||
Genre: Pop | 0 (fixed) | 0 (fixed) | 0 (fixed) | 0 (fixed) | ||||
Genre: Hip-Hop/Rap | * | − 0.05 | 0.04 | * | 0.29 | < 0.01 | ||
Genre: R&B/Soul | * | * | − 0.23 | 0.02 | * | |||
Genre: Rock | − 0.09 | 0.13 | − 0.07 | 0.04 | − 0.57 | < 0.01 | − 0.28 | < 0.01 |
Genre: Country | * | * | * | * | ||||
Genre: Other | * | * | * | * | ||||
Criteria: Cars | * | * | * | * | ||||
Criteria: Traffic | * | * | − 0.49 | < 0.01 | − 0.70 | < 0.01 | ||
Criteria: Driving | * | * | 0.14 | 0.07 | 0.09 | 0.13 | ||
Criteria: Passenger | * | * | * | * | ||||
Criteria: Stationary in Vehicle | * | * | * | * | ||||
Criteria: Car Parts | * | * | * | 0.12 | 0.06 | |||
Criteria: Brand | * | * | 0.21 | < 0.01 | 0.19 | < 0.01 | ||
Criteria: Taxi/Hitching a Ride | * | − 0.11 | 0.02 | * | * | |||
Adjusted \(r^{2}\) | 0.01 | 0.05 | 0.10 | 0.15 | ||||
Model significance | 0.03 | < 0.01 | < 0.01 | < 0.01 |