Machine-generated text creation is a relief for copywriters and content managers in e-commerce. Anyone involved in writing texts benefits from it. There are two key technologies: generative AI and rule-based data-to-text systems. Most have already had their first experiences with generative AI, most notably GPT. But how do these two technologies differ in terms of […]
The Case
As a news and press agency, APA creates reports about all kinds of elections, from European elections to very local Austrian municipality elections. APA needs to report the results very quickly and accurately to their customers, the publishers.
The Challenge
Initially, APA only had a lot of election data for many Austrian municipalities, but no texts at all. As APA wanted to take election analyses one step further and to hyper localize their reports, it was impossible to write texts for the existing data manually, and in a very short amount of time. Also, to ensure journalistic ethics and accuracy, it is very important that the election analyses are reliable, of high quality, and punctual. That means that all texts have to be available as soon as possible, right after the election results are announced, as there is no time latitude. In terms of automation, there was another obstacle: Due to the circumstances of the project, the further development and quality management was difficult and not feasible in the way it’s usually done in e-commerce businesses. That’s because elections are complex, unpredictable and difficult to scale, as they only take place every now and then.
Their goals:
- Create scalable texts for many different elections
- Create high quality texts very quickly
- Being able to turn all available election data into texts and graphics
- Hyper localize election reports
The Solution
The APA team knew they needed to find a way to automate their election analyses. So they decided to use AX Semantics.
Direct API
As speediness, reliability and high quality was of essence in creating the election texts, APA and AX Semantics decided on the solution of using a direct API. Which means that the published election results of each municipality are sent directly and immediately to the API, which serves as an interface to AX Semantics. Then, out of this data, the texts are created with the AX Semantics software, and instantly contributed to the publishers.Election texts at scale
The texts created with the AX Semantics software are scalable and statistics as well as raw data are easily transformed into ready-made election texts for consumers. Therefore, APA can now report about many different elections, using the initial set up of rules. Furthermore, APA can generate texts on topics that would have been too expensive and time consuming if written manually. It is now possible to report on elections on a hyper localized level without any manpower or extreme expenditure of time needed whatsoever.Creating high quality texts consistently
Due to the automation, the election texts are consistently of high quality, as the AX Semantics software is solely based on the rules and statements initially set by APA. Therefore, APA can deliver the election texts to the publishers without needing to worry about meeting the standards of journalistic accuracy and ethics. Also, due to the time saved by the content automation, APA has been able to expand their portfolio and sell not only data and graphics to various publishers, but also continuously produce high quality texts with interpretations of these data.
Election analyses of APA generated with the AX Semantics software, published on the website of kurier.at.
The Results
As of today, APA has automated thousands of election texts on European, national, regional and even hyperlocal level. The agency creates one text for every municipality in every single election. Therefore, the demand quantity of content varies according to the type of election from about 69 to 2,200 texts. Due to the automation, hyper localization is not an issue anymore and can be carried out in a flash. In summary, APA was able to generate more than 11,000 unique elections reports within one year with AX Semantics.In addition, due to the success of content automation, APA was able to set up new projects, like the “Wiener Dateng’schichten: LostPetsVienna”: Data of found pets in Vienna are transformed into texts and are automatically published on the Twitter feed of APA. Also planned is the production of texts for “Österreich Werbung” on different touristic topics, like detailed descriptions of skiing areas and so on.
