From pen and paper to the digital age, journalism has always been shaped by technology. While computer-assisted forms of reporting have thrived over the last decades, they did not affect journalists’ autonomy. This is now brought into question by recent developments in newsroom automation that we discussed last month during a workshop held at FuJo.
News automation is now employed at every stage of newsmaking, from the collection to the distribution of news, while involving at the same time low to high levels of autonomy[i]. News gathering with low levels of autonomy would see journalists going on the ground with none or few computational means while news production with high levels of autonomy could translate into the auto generation of news texts through algorithms, a practise known as automated journalism.
Over the years, a growing number of news organisations have in fact turned to automated journalism, prompting media practitioners to reexamine their own skills and to ponder on their relation to the technology. If on one hand journalists would fear to be replaced by this new development, they could also align with the technology while creating templates for automated news or using them as background materials for their stories[ii].
At the same time, even these positive outcomes need to be investigated critically[iii]. For instance, if automated journalism takes on routine tasks such as traffic and weather reports so that journalists can focus instead on investigative or in-depth reporting, it could also displace a category of workers or prevent early-career reporters from breaking into journalism. This decision will ultimately rest with newsroom management.
Global trend
In recent years, much of the spotlight has been on United States with news organisations the likes of the Associated Press and The Washington Post championing their own use of automated journalism. That being said, news automation is a global phenomenon. In China for instance, commercial and state-owned media have devised their own automation strategies, although little is known about this in Western countries[iv].
In Europe, much of the attention has been on English and Nordic outlets, even though the deployment of automation strategies in other European countries would make for an interesting case study. It would be especially worthwhile to look at Europe’s largest media market[v], Germany, where 7% of newspapers have reported experimenting with news automation[vi] and where at least five automated journalism providers compete in the same market[vii].
Leveraging NLG
Much remains to be known about the precise reach of automated journalism providers as some of them signed non-disclosure agreements with their media clients[viii]. One interesting area of investigation, though, is the relationship they entertain with journalism[ix]. While they are selling content to media clients they also seem to be keeping their distance from journalism, seeing these services as a benchmark to evaluate the products they sell among others to financial and government clients.
Any organisation dealing for instance with multiple languages can indeed benefit from natural language generation (NLG), the technology behind automated journalism. At the European level, while the information flow is usually well taken care of in languages such as English, French and German, there is a lack of available resources for less represented tongues such as Finnish or Irish. To address this, the EMBEDDIA project is experimenting with advanced NLG techniques to embed the overall meaning of a text into another language[x], a process distinct from a regular translation.
Classical vs. advanced methods
One of the aspects that EMBEDDIA is considering includes the use of a mixed framework for NLG production. While traditional schemes involve rule-based systems that build on pre-written story templates, advanced machine learning techniques require algorithms to feed on a large amount of journalistic texts to produce a similar output. However, classical approaches to NLG are generally believed to be too rigid and advanced methods too unpredictable. To solve this, a rule-based structure could sparingly incorporate advanced techniques at various levels, thus creating a mixed approach to NLG production[xi].
Finally, it also worth noting that news automation is not limited to text generation. Another example includes the auto production of graphs and visualizations when sufficient levels of abstraction are in place[xii]. As a result, there seems to be an emerging market to provide news outlets with automated graphs, what could possibly be combined with automated texts.
In conclusion, while newsroom automation is no panacea to all the hardships experienced nowadays in journalism, it is becoming ever more ubiquitous. As a result, we should investigate its impacts not only on readers and practitioners, but also on society as a whole. If journalists lose their autonomy to machines, what will become of electoral or financial results once their coverage are automated? Now more than ever is the time to reflect on this.
This project has received funding from the European Union’s Horizon 2020
research and innovation programme under the Marie Skłodowska-Curie grant
agreement No 765140 (JOLT) and the Research and Innovation grant agreement No
825153 (EMBEDDIA).
[i] Cools, Hannes. 2019. ‘From low-level to high-level automation in the newsroom: towards a multi-level typology of computational journalism.’ Presentation delivered at News automation at work. Dublin: November 4-5, 2019.
[ii] Dierickx, Laurence. 2019. ‘News automation within the uses and professional practices.’ Presentation delivered at News automation at work. Dublin: November 4-5, 2019.
[iii] Danzon-Chambaud, Samuel. 2019. ‘Displace, alleviate, transform and enhance: a critical typology of how automated journalism impacts the work of media practitioners.’ Presentation delivered at News automation at work. Dublin: November 4-5, 2019.
[iv] Tuulonen, Hanna. 2019. ‘News automation in China: The impact on media practices and content in China and in Europe.’ Presentation delivered at News automation at work. Dublin: November 4-5, 2019.
[v] Körner, Theresa. 2019. ‘Automated Journalism.’ Presentation delivered at News automation at work. Dublin: November 4-5, 2019.
[vi] Lindén, Carl-Gustav and Hanna Tuulonen (eds.). 2019. News automation: the rewards, risks and realities of ‘machine journalism.’ Frankfurt, DE: WAN-IFRA.
[vii] Dörr, Konstantin. 2016. ‘Mapping the field of Algorithmic Journalism.’ Digital Journalism 4 (6): 700–722, doi: 10.1080/21670811.2015.1096748.
[viii] Dörr, Konstantin. Ibid.
[ix] Sirén-Heikel, Stefanie, Martin Kjellman and Carl-Gustav Lindén. 2019. ‘Software providers of news automation as peripheral actors in journalism.’ Presentation delivered at News automation at work. Dublin: November 4-5, 2019.
[x] Toivonen, Hannu and Carl-Gustav Linden. 2019. ‘EMBEDDIA: cross-lingual embeddings for less-represented languages in European news media.’ Presentation delivered at News automation at work. Dublin: November 4-5, 2019.
[xi] Leo Leppänen. 2019. ‘Automated news text production: where we are and where we might be heading.’ Presentation delivered at News automation at work. Dublin: November 4-5, 2019.
[xii] Sheehan, Shane. 2019. ‘Information visualisation.’ Presentation delivered at News automation at work. Dublin: November 4-5, 2019.
