Texting, without Texting

After running the Textmodel Creator, the automatically generated text models can be adjusted in structure. You can add additional messages and templates to the text model according to your preferences. There are two ways to change the content of the dynamic templates created automatically by the Textmodel Creator.  

On the one hand, you can influence the configuration of the text output during the process, e.g. by editing the prompts.

However, if you want to retain maximum control over the generated content and adjust it afterwards, the dynamic template must be converted into static text first. For this, start the Message Composer via the three-dot menu next to the message that contains the dynamic template, and let it guide you through the process to get fully editable templates.

In short: Dynamic templates are generated using the GPT model and offer varying creative text contents for each query; static templates contain predictable, fixed text variants.

The Text Model Creator generates messages with dynamic templates. Dynamic templates are text components that are based on OpenAI's GPT model and generate new, unique and creative text with each new query. The texts obtained in this way change with each API call - they are therefore dynamic. To gain increased control over the output content and to know exactly what text the textengine.io is publishing on the target page, you can convert these dynamic templates into static templates using the Message Composer. A dynamically generated text content is "fixed" so that a variant from a pool of predefined texts appears with each retrieval. The static templates also adapt to your data. You can also enrich them with synonyms and alternative phrases in a few clicks to bring variety to your content.  

Converting to static templates with the Message Composer has the following advantages:

• It enables you to maintain control over the output text. Static templates are predictable and, once adapted to the corporate language, can be safely replicated without risk to the brand.
• If you have no clear idea for the creation of logical rules of your text to cover different use cases, the Message Composer can help you by suggesting a meaningful subdivision and logical rule operators. For example, if your dataset contains trousers in various fabrics, a suggestion will be made to divide these according to temperature or weather. For example, automatically suggested rules will generate sunny weather texts for light fabrics, while solid fabrics will be described as warm and weather resistant. This way, you don't have to go through the long list of data values yourself and develop your own rules, but automatically receive a suggestion for a sensible division. Even if you want to make adjustments here, you get a much easier insight into possible rule operators.

Certainly. There are various rationales why a mixed model makes sense for your solution:

• If there are sections in your content strategy or desired text structure for which no data or only unstructured data are available, dynamic templates help to generate coherent text content adapted to the individual data situation.
• Also, for text sections where data gaps such as general information (e.g. travel destinations) can be automated by LLMs, enhancing the overall text and reducing text effort, it is worthwhile to use dynamic templates.
• For search engine optimised texts, it is also beneficial to use AI-based dynamic texts with appropriate prompts.

However, as soon as it becomes necessary to have the textual output controlled by a human instance, static templates should be created. This makes the text output predictable. Static templates also adapt to the given data fields. Here, minute adjustments can contribute to variety in the text: variations to descriptive adjectives, synonyms, as well as entire text modules can be added with a few clicks.

It is also important for sensitive content that formulations be checked by a human instance before scaling. Static templates are your choice here, as they reliably deliver the correct, clearly defined content in advance.

The Message Composer draws its strength from AI-driven analysis of text parts to recognise a logical structure. With the underlying data, a sensible splitting and subsequent application of rules take place. It is important here that the initial message contains only a single dynamic template. Depending on the length and context of meaning, the Message Composer can generate from it several messages with a static template each. The Message Composer assists you in the restructuring process and allows you to make ongoing adjustments. Messages resulting from this remain fully editable even after using the Message Composer.

The Message Composer only works when applied to a message with a single dynamic template. Also, after successfully running the Message Composer, the original message is overwritten. Our tip: simply copy the message before the start of the transformation. Thus, you retain the original dynamic variant and a static alternative with rules.

If you have never worked with textengine.io before or do not yet have a fixed structure for your target text in mind, the Text Model Creator assists in creating it for you. With a few clicks, you can generate your first text model and get an idea of what a product description could look like based on your data, for example.

The Text Model Creator provides you with variable, dynamic templates based on the GPT model. Already during the prompting, care can be taken to maintain the desired tonality in text creation. This reduces subsequent editing effort by directly receiving suitable dynamic texts.

One of the advantages of the Message Composer is that it can make the generated dynamic texts editable afterwards, so the target content can be viewed, edited and further adapted to your own corporate language by a human. This gives you the opportunity to quickly create legally compliant text modules.

Large Scale Content Automation combines generative AI and machine learning technologies and the deterministic approach of NLG processes to help businesses scale. It brings in the advantages that come with the creativity and uniqueness of large language models, as well as the deterministic predictability of a rule-based text model.

Our offering distinguishes itself based on the principle of Large Scale Content Automation through the synergy of AI-generated creative content and template-based text structures. Here, all the advantages of AI-based texts such as creativity and uniqueness meet a legally secure framework that creates logical text structures based on your data. By enabling constant human control ("human-in-the-loop") and continuous customization of your text structure, text models can be created in a very short time. These text models contain various wordings and can be translated into all common European and Asian languages.

OpenAI's GPT model is used to create the dynamic texts. GPT or Generative Pre-trained Transformer is a family of autoregressive language models that generate text suggestions based on a prompt. There are several generations of this language model in the GPT series, developed by OpenAI. They can be used in various applications, including natural language processing, text generation, and machine answering.

We use the current OpenAI services via MS Azure, which allows us to transfer data within the EU and comply with the General Data Protection Regulation (GDPR).

All currently implemented functions in textengine.io that use a GPT component can be enabled or disabled. Disabling the GPT features has no effect on the general use of textengine.io and the ability to create, edit, or improve text models. If you do not want support from LLMs in text generation, contact us. We can adjust the settings for you.

No. Data that you enter into textengine.io is never passed on to large language models and used for training purposes.