Speech mining — a revolution in process optimization?

The birth of a concept

Traditional process mining focuses on analyzing the digital traces that employees leave behind in the company's IT systems, where processes are considered purely technical. But what happens when the focus is not on digital traces but on the spoken word? This is where an innovative concept is being developed that combines speech analytics and audio mining — speech mining.

You've never heard of speech mining? No wonder, because, strictly speaking, it doesn't even exist yet! Rather, it is a vision of how AI could revolutionize process modeling in the future. But right from the start: What exactly does speech mining actually mean.

Speech analytics, also known as interaction analysis, is a technology that uses artificial intelligence to understand, process, and analyze human language. Audio mining describes the systematic extraction of information from audio sources. It is about gaining important information by listening to spoken language. The combination of speech analytics and audio mining creates a new concept: speech mining — which combines human interaction and efficient process design.

In contrast to classic process mining, speech mining focuses on the spoken word. Speech mining is therefore a modern method that can be used to represent the current situation or the current state of a thing or process. Human interaction is used in real time to describe processes, generate new content and create initial models. The approach is based on extracting process-relevant information from spoken communication.

Practice-oriented speech mining

To put theory into practice, the following example provides a vivid insight into the application of speech mining:

1. Speech-to-text recording: Use speech-to-text technology to record your spoken instructions during the assembly process.
2. Transformation into assembly instructions: Use an appropriate prompt to transform the spoken word in ChatGPT into assembly instructions that meet your company's needs.
3. Tabular structure for the presentation: We can recommend a tabular structure with columns No., Description and Picture for the assembly instructions. As a result, clear and easy-to-understand steps for the assembly processes are recorded.
4. Creating images of the assembly process: Take pictures of the assembly process to provide visual support for the assembly instructions.
5. Automatic assignment of images to assembly steps: Use ChatGPT4 to automatically assign the images to the appropriate assembly steps. This makes the assembly instructions an ideal specification and working aid that makes the assembly process easier.

Same use case, different setting: Why not also record structured conversations about a process using speech-to-text? This approach enables assembly instructions to be created collaboratively by several people describing various steps in detail. In contrast to traditional methods, where one person describes the entire process, in this setting, different participants can describe different parts of the process. One more stage of development: Imagine how cool it would be if a group of people talked loosely about a process without sticking to a fixed structure, and then you let ChatGPT create a coherent and meaningful sequence from all these steps.

The company as a living 'digital twin'

Imagine being able to iteratively design a dynamic process landscape for your company, including process models, work instructions, and goals, through leadership impulses. This evolving “digital twin” is initially based on generic input data, but is becoming more and more specific as a result of continuous adjustments to current leadership impulses. This approach focuses on reducing generic management documentation and creating a flexible structure that reacts to change and is driven by leadership impulses. You can also record conversations about processes in your organization and thus keep your digital twin up to date with discussions and decisions. It also makes sense to integrate classic process mining into the digital twin. On this basis, a living system is created, because it is possible to ask the Digital Twin questions at any time, such as: “Can I buy the pencil myself or do I need approval for it?” and the correct answer is always delivered. In the background, however, a clearly structured process landscape that can be read by humans remains.

A promising future

Although the topic is still in its infancy, it is already apparent today that it can be used successfully. The possibilities are varied and the momentum that speech mining brings to the world of process modeling is impressive. The question of how this field will develop remains exciting. But one thing is certain: The great potential of speech mining will decisively shape the way we understand, document and communicate processes. It remains to be seen what exciting developments the future will bring in this area.

‍