Automation is considered one of the key issues for making companies viable for the future. The use of artificial intelligence has made rapid progress in this area for some years now, creating cross-sector economies of scale that would have been unthinkable with conventional technologies.
Even more automation in natural language generation with the help of AI
The same applies to the automated creation of texts (natural language generation, also known as automatic text generation). Here, artificial intelligence initially ensures grammatically correct texts and enables a high degree of process automation – through AI, the NLG software can, for example, learn a new language or the independent generation of a wide range of formulations.
And this is exactly where the current NLG release from Retresco comes in: with the release of the new text variant suggestions feature, Retresco is the first NLG provider in the world to provide a software function that uses artificial intelligence to independently submit formulation suggestions in the form of complete sentences. This feature is currently available for German-language projects on the platform. It allows the entire process of machine text creation to be considerably shortened. In addition, the text variance can be significantly increased without user intervention.
What added value does the feature provide in practice?
In the future, users of textengine.io (the NLG software from Retresco) will be able to enter a complete sentence into the user interface, click on ‘text variant suggestions’, and high-quality formulation alternatives, which are created completely automatically, will be suggested. These suggestions can be adopted either partially or completely from the form proposed. In addition, the texts can be revised and modified at any time. The result: significantly more text variance, significantly more efficient processes in the creation of texts, a better user experience and greater flexibility in creative processes.
For whom is this interesting? In principle, the feature is relevant wherever statements need to be paraphrased. This means that it can be used in those areas where a large number of texts with many variations are required. This is of great importance, for example, in e-commerce, where numerous and above all versatile product descriptions are required, which should simultaneously be created quickly while also possessing such high quality that they bring the product as close as possible to the customer through good descriptions. A further advantage of texts with many variants is search engine optimisation.
Text variant suggestions – a decisive step towards end-to-end text generation
With the current feature release, Retresco can further expand its technological pioneering activities in the field of AI-based language technologies and take a decisive step towards end-to-end text generation. The goal of the end-to-end approach is that an NLG system complete, on its own, the process of text generation – from the data on which the created text is based to the finished text.
The long-term goal of this is that, in the future, this system can be easily fed with data or sample sentences, learn from them and abstract them independently, so that new formulations and texts are created without human intervention. The advantage is obvious: new use cases could be developed within a very short time, allowing a considerable increase in the degree of automation and efficiency – and all this without any initial setup by human hands. In this respect, the new Retresco text variant suggestions feature represents a decisive step towards end-to-end text generation and thus into the future of natural language generation: Retresco’s NLG software now independently generates several suggestions of alternative formulations – without the system having to be explicitly manually trained by the user.
With this new feature, Retresco is setting new standards in the development and application of NLG. The pioneering work is to be further promoted and continued: numerous other releases in the direction of end-to-end generation will follow in the coming months and years, enabling the machine creation of text with an even higher degree of automation, efficiency and variety.