Meet the Innovator interrogating words to help improve transport services

Trawling through thousands of social media mentions, many hours of telephone recordings and stacks of survey responses to identify trends in what people are saying about a company would take a considerable effort for even a dedicated and well-resourced team.

Staff may be excellent at acting upon individual complaints, requests or suggestions; but getting a broader picture of how a service is perceived – and spotting patterns among the many comments – would represent a mighty challenge.

Step forward Ruth Erdal, the Innovation Manager for technology firm Wordnerds who, along with her husband Steve and his co-founders Pete and Angela Daykin, have developed a ‘linguistic model’ to glean insights from text using artificial intelligence. It combs numerous sources of written feedback received from customers to better understand exactly what is being said, and how a service might be improved.

The company received £30,000 two years ago from the Transport Research & Innovation Grants (TRIG) programme – delivered by Connected Places Catapult on behalf of the Department for Transport – to refine the product; and has since gone on to work with National Highways to understand how drivers could be made to feel more at ease when passing through roadworks.

Keeping the messaging simple

Ruth Erdal was born in Sheffield, grew up in a suburb on the edge of the Peak District, and read American and Canadian Studies at the University of Birmingham. Year three of her course was spent in the city of Toronto where she immersed herself in the local architecture, photography and culture.

Ruth enrolled in a Masters in publishing at Oxford Brookes University, inspired by her photos of buildings in Toronto, with the idea of creating beautiful coffee table books. During the course she met her future husband and business partner Steve, and later secured a job with Osprey Publishing in Oxford, producing designs for illustrated military history books.

After four years the couple decided to go travelling and ended up moving north to Tyneside, but found only limited employment opportunities in publishing. Ruth took a marketing job at Northumbria University, her husband started at Newcastle University and the pair later founded an agency called Kiss Copy (which stands for ‘Keep It Simple, Stupid’).

They analysed the language used in university prospectuses, to try and help academic institutions to better communicate what was on offer to people from different socio-economic and cultural backgrounds; using compelling language that was easy to understand.

Steve co-founded Wordnerds in 2017 with creative agency owners Angela and Pete, who were asked by Nissan to identify trends in issues reported by drivers. They developed a linguistic model to interrogate the data and understand the nature of comments received. Ruth joined two years later to focus on developing the customers’ experiences of the product, before exploring new business opportunities for using the model in different settings such as public transport and infrastructure.

Exploring sentiments in text

Software developed by Wordnerds explores the relationships between different words in a passage of text, and a series of ‘rules’ identify whether a comment is worth exploring further.

The company analyses customer sentiment on social media on behalf of several large retail clients, who also collect data from online surveys posted on the bottom of till receipts or transcripts of telephone conversations with call centre staff.

“Surveys receive thousands of responses every day and it is far too onerous to read every single comment,” says Ruth. “We are here to help companies work out their customer priorities, what is causing the most grief, and what small things they can fix. It’s really important that changes can be made in response to feedback.”

Wordnerds also works with several housing associations to help ensure they are listening sufficiently to the needs of tenants; and is becoming more active in both rail and road.

One new project with Transport for Wales will involve creating a behavioural science model that aims to understand what motivates people to choose different modes of travel.

The program not only shines a light on common themes that people are talking about right now, but allows clients to delve deep into customer feedback on a particular subject to head off any emerging issues before they risk becoming a larger problem.

Eliminating bias from the process

Ruth first heard about Connected Places Catapult in 2021 after she applied for TRIG funding to develop techniques to eliminate ‘bias’ in artificial intelligence models that interrogate customer communication data associated with public transport.

Large language processing models risk demonstrating certain assumptions based on the name of whoever makes a comment, which can make it more difficult to accurately reflect the sentiments of certain groups of passengers.

Wordnerds was already involved in several rail projects at the time, so focusing efforts on removing bias from responses was intended to improve confidence among train company clients that they can treat everyone equally.

Ruth says the Catapult offered the company “a lot of support alongside the money” and commissioned a videographer to produce testimonials of its technology. “They have been so supportive, and made sure we have opportunities to share what we’ve done.”

Shortly after completing the TRIG project the company was approached by National Highways to take part in an innovation competition focused on developing new ways of using data to improve the customer experience in roadworks. “Our software brings all of their data from across the organisation into one place,” Ruth explains.

Wordnerds is currently analysing customer feedback from emails, social media comments and surveys on four major road projects in England; looking specifically at how popular or otherwise newly introduced measures are proving to be, including new dynamically responsive diversion signs. “We will be able to show whether people are more or less pleased with the signs, in order to show whether they are worth repeating in future.”

More general causes for concern on the roads which the software can be used to track customer sentiments include potholes, litter and a perceived lack of activity on roadwork sites. “There’s hundreds of factors we can look at, from the very granular to the big picture,” says Ruth.

Ruth says that working closely on developing an innovative technology “has been amazing; especially working with National Highways. Roads was a sector I previously knew nothing about, but I am learning constantly and it is fascinating to have a little window into their world.”

And would Ruth describe herself as a ‘word nerd’? “Yes, I would say so. My idea of fun over lunch is doing the crossword.”

Find out more about the Transport Research and Innovation Grants programme and keep your eyes peeled for future opportunities.