partner news

OPTIMUM pilot project supports research on travel-related stress

August 3, 2016

Twitter is geared to real-time use more than any other form of social media. Certainly not a medium for expressing nuance, ‘tweets’ do deliver the punch of immediacy. The OPTIMUM project, through one of its three pilot studies — “Proactive improvement of transport systems' quality and efficiency” — has funded a study to analyse short text messages in order to determine levels of stress that commuters experience while travelling. If it proves possible to interpret such messages with reliable accuracy, transport authorities will be able to respond quickly to incidences of high stress in order improve traffic management and quality of service. 

Mike Theiwall is the author of the new study “TensiStrength: Stress and relaxation magnitude detection for social media texts”. “TensiStrength,” the study begins, “is a system to detect the strength of stress and relaxation expressed in social media text messages. It uses a lexical approach and a set of rules to detect direct and indirect expressions of stress and relaxation, particularly in the context of transportation.”

Intelligent transport systems (ITS) already use traffic sensors, road monitoring cameras, mobile phone GPS signals and number plate recognition technology to harness information, but a wealth of text information available to computing systems can now be mined to make further improvements on the predictive power of ITS and other systems.  

As outlined in the description of the OPTIMUM pilot study: “Since changes in traffic conditions and relevant incidents (such as accidents) can occur unexpectedly at any time, it is necessary to inform travellers and to suggest alternative routes. The integration of various real-time traffic data sources will provide the required information to realise traffic-state-aware routing that can guide travellers along routes to their destinations” — and, hopefully, with a minimum of stress, we can now add.

According to the study, “TensiStrength is an adaptation of the sentiment strength detection software SentiStrength.” The tasks of the two programs are related but not equivalent. TensiStrength borrows its emotion terms from SensiStrength, but includes manually added stress terms and indicators “derived from a range of academic and non-academic sources that describe stress in general, or stressors associated with travel.” Examples of commute-related stressors include: heavy traffic, frequent braking, traffic jams, congestion, slow average speeds, transport signals, and unpredictability of journey time.

TensiStrength research involved assigning scores on a 1 to 5 scaling system to more than 3,000 stress-related tweets. The tweets were collected "using the keywords from a variety of sources over a one-month period in July 2015 and then randomly sampled”, the study explains.

While the results of the study have not produced a consensus, they do show that “TensiStrength is able to detect expressions of stress and relaxation in tweets with a reasonable level of accuracy compared to human coders, more accurately than a similar sentiment analysis program … TensiStrength can therefore be used as an off-the-shelf solution for stress and relaxation detection.”

Magnitudes of stress and relaxation can be measured on a personalised level to determine the emotional state of travellers. A user under stress is less likely to select a mode of travel with which he or she is unfamiliar, while a relaxed user may be more accepting of less habitually familiar modal options. The OPTIMUM platform can provide proactive recommendations to match an individual user’s behavioural profile. On a system level, spikes of high-stress tweets within a given area of a transport network may result in driving behaviours that increase the probability of accident and also contribute to the build-up of traffic volume — through, for example, erratic lane changes. Such information will be beneficial for generating more accurate forecasts and other improvements that OPTIMUM hopes to introduce to its event-detection mechanisms.

Of course, much work remains to be done in this new field of research, but the initial results are encouraging from OPTIMUM’s ITS-related vantage point.

Click here to find more scientific papers published by OPTIMUM