I am a news automation developer with Torstar in Toronto, Ontario, Canada. I write software to find, clean and automatically process structured and unstructured data into written narrative stories. Torstar has some of the most talented journalists out there. Having them write rote stories is a waste of their time and their talent. You wouldn't ask Picasso to paint a fence. I write the software that paints the fences to free them up to paint the masterpieces.
Before my current gig, I was the head of The Canadian Press Digital Data Desk. I lead a team of journalists, developers and data librarians to innovate new news products and automate news production. We created tools to automate COVID coverage, write sports results stories, cover elections and turn serialized structured data into hundreds of local-focused stories.
My recent work has focused on developing tools and using AI and machine learning to automate news production and assist research in the newsroom. My ideas for automation lead to the creation of a new innovation-based bureau at the Canadian Press and won funding from the Google News Initiative.
In my first year leading the team, we successfully automated thousands of stories, saved thousands of hours of work and enabled reporters in the newsroom to focus on reporting complex stories while leaving the rote work for machines. We integrated AI and machine learning into our processes to increase the value of the content we created and to provide added context to data to enable better reporting.
Prior to leading the team, I worked as a hybrid journalist / designer developer. I created interactive graphics and tools for The Canadian Press Multimedia Desk. Prior to that, I was the special editor for data journalism and data visualization at the Edmonton Journal. I am also an independent iOS and Android app developer.
I have built tools and templates for quickly creating online charts and graphics. I used Node.js to write web scrapers to pull in data and software to automatically write stories using that data. Coupled with APIs I helped create and develop, I have also built Twitter bots that update readers with live election data.
I have also taught data analysis and data literacy skills to colleagues at work and to students at MacEwan University.
My graphics have won and been nominated for awards ranging from the Online News Association’s Online Journalism Awards to the Canadian Online Publishing Awards. I was also named as the inaugural ONA MJ Bear fellow as one of the world’s top digital journalists under 30 years old.
My work is featured on several platforms, ranging from Twitter to mobile apps to websites and in print.
I recently spoke at Data Driven 2021 on automating daily COVID-19 stories and charts. I also had a paper accepted at the Computation + Journalism Symposium 2021 on Using Automation in Election Story Research.
You can read more about my work here:
How The Canadian Press is using bots to build the future of election coverage - JSource
Reporter: How “robojournalism” can save local news - Ryerson Journalism Review
Automation at The Canadian Press: Doing More with Less - Sourcefabric
COVID-19 with automated journalism: leveraging innovation in times of crisis - Columbia Journalism School’s Tow Center for Digital
Journalism (Samuel Danzon-Chambaud, School of Communications, Dublin City University)
Included on this page is a short sampling of the types of work I have created. This includes automation, tools and graphics. This should give you a good idea of my style and ideas.
I am an advanced Javascript developer. I also have expert-level knowledge of Photoshop, Illustrator and InDesign.
In addition to my design work, I am skilled in data journalism. That includes Excel, MySql and NoSQL databases and OpenRefine for data analysis.
I understand what it means to work on incredibly tight deadlines with public results where accuracy is paramount.
And most importantly, I know what it takes to survive in today’s media landscape. I’m always learning and improving and helping others to do the same.
Early in the pandemic, the newest COVID numbers were coming in at irregular intervals from jurisdictions across the country. It took a lot of time and effort on the part of numerous reporters to compile the data and turn it into something readable that would get updated throughout the day. We quickly realized we needed a system that would keep the stories in an identical format and eliminate the possibility for human error. The turnaround time from idea to the first working version of the tool took less than 24 hours, and it has continued to evolve over the course of the pandemic as more comprehensive data became available.
In its current state, the tool now generates stories in both French and English, with detailed regional breakdowns of several relevant metrics (from the number of new cases, to rates of infection and the number of vaccines distributed), along with several accompanying charts for both print and online -- and it does it all within seconds.
We created software to generate riding previews to go out to the newswire a few weeks before the British Columbia election in 2020. We automated a story for each of the 87 electoral districts, which included information from the previous election such as the person who won, the results for each candidate and the voter turnout.
The software was also used on election ight. When an electoral district was called, a story with the results would generate automatically. The story software connected to the CMS, andposted the stories automatically. This tool freed up time for CP reporters and allowed them to cover more in-depth election night stories. A challenge of automating the election results in 2020 was accounting for an influx of vote-by-mail ballots in the results calculations. For many electoral districts, the vote-by-mail ballots did not prevent a winner from being called on election night. Any electoral district that didn’t have a winner called on election night had a brief published with the results from a certain time at night with an additional sentence included about the vote-by-mail ballots.
I created a Slack-based bot to help reporters with research for the 2019 Canadian Federal election. It uses a custom built API and combines the 2015 election results with all the data from the 2016 Canadian Census. It will give summaries and deep dives into data about ridings. It will also compare two ridings and write a 700 word summary about a riding. The link above is to a J-Source piece I wrote about the tool.
The sports desk used to write one or two sentence summaries of CHL games based on box score, but it was impossible to cover all games that happened in one night. The CHL story writer was created to allow CP sports reporters to write full game stories on more CHL games than before. The CHL bot does full stories without the quotes. A reporter simply takes the boxscore URL and puts it in the respective field in the tool. The tool then connects to a D3 API, takes the URL information, and connects to an unsecure API that is run by the CHL. For each game it pulls in the preview information and the box score and converts it into a story. There are tick boxes at the bottom of the CHL bot web page which contain elements of the stories that can be written. The story will be written with or without certain information depending on if a user checks or unchecks the boxes. Another feature of the CHL bot is that it generates variations of sentences so not every story is exactly the same.
The Canadian Press Business Desk writes quickhit stories of companies quarterly earnings based on press releases. The reporters usually have an hour from the time the press release comes out, to the time the story has to go out to the wire. Then, they write another story about the company once more information comes out in the earnings calls or from sources. To save reporters time for writing the quick-hit stories, we created a tool that would automatically generate these stories. The business reporters now put in the key metrics in the respective boxes in the form, and a story will be automatically written based on the information provided.
Work is ongoing on the machine learning component. Soon the tool will not need human intervention, or very little intervention, to publish. Once a press release is published, the software will recogonize the release, read the story, pull out the important information and automatically write the story and then publishe it.
The Labour Force Survey is a monthly release of data about changes in The Canadian job market - from the unemployment rate to an estimate of the number of jobs lost/gained in a particular sector. Canadian Press reporters cover this release every month - usually a quickhit, several write-thrus and a series of lists.
The tool generates a story with national employment statistics, a list of the provincial unemployment rates, and a list of the unemployment rates in large Canadian cities. It also writes a series of briefs, which writes short, quickhit style briefs for multiple levels of geography. In total, the tool will generate more than 130 stories: quick hits for Canada, each province, more than 30 major cities and all 65 economic regions, as well as a series for lists by geography and major industries.
A collection of charts that are quick to deploy and use csv files for the data. The idea was to give our reporters a quick and easy way to visualize data.
A tool that allows night deskers at The Canadian Press to build their own locator maps in the browser. It means they can make maps without needing an Adobe Illustrator license or the knowledge to use Illustrator. It is no longer being actively maintianed by The Canadian Press.
A tool for creating social media quote cards that reporters can use when they tweet links to their stories.
The 2017 Ontario provincial election map I made. Uses Mapboxgl for the base and has a d3js SVG overlay. The overlay is loaded as topojson. Data is read from an API I helped create. Includes poll-by-poll level results and equal area results. Used online on the Globe and Mail, Toronto Star, Global News and more. It was also used as the big board map on CityTV and Global News.
(Note: The data source is no longer being maintained by The Canadian Press but the map is still visible without data)
The French version of the 2016 Canadian federal election map I made. Uses Mapbox (leaflet) for the base and has a d3js SVG overlay. Data is read from three JSON files.
(Note: The data source and the Mapbox backend are no longer being maintained by The Canadian Press but the map is still visible without data or a base layer)
The R. James Travers Foreign Corresponding Fellowship partly funded this series. It commemorates Jim Travers's career and aims to enable significant foreign reporting projects by Canadian journalists to give Canadians first-hand, in-depth coverage of stories beyond the country's borders. Travers spent six years reporting from Africa and the Middle East and deeply believed in the power of international reporting. I did the layout and graphics.
A series of stories on Canada's 150th birthday. I did the layout and graphics/games for all the stories. No longer being maintained by The Canadian Press.
A charted look at the history of the Edmonton Oilers. (No longer being maintianed by the Edmonton Journal)
A playable version of an old Indigenous game that is regaining popularity at powwows in Canada. No longer being maintained by The Canadian Press.
Explaining a Patrick Chan quad jump for the 2018 Olympics.
Comparing the 2016 Olympics 100m gold medal race results to the world record. No longer being maintained by The Canadian Press.
A minute-by-minute breakdown of a day in court. Used to illustrate why there are so many delays in trials.
An animated Mark Buerhle baseball card. Hovering flips the card over and shows his pitch profile. No longer being maintained by The Canadian Press.
A tool to outline which team would advance between TFC and the Montreal Impact based on the away goals rule. No longer being maintained by The Canadian Press.
Comparing Mike Babcock’s salary to other big-time coaching contracts. No longer being maintained by The Canadian Press.
A look at potential trades made to imitate Tindr. No longer being maintained by The Canadian Press.
Email me at lucas@lucastimmons.com
or call me at 416-937-5346
I hope to hear from you!