iiRDS for a Better Support Experience

Original text written by Harald Stadlbauer (NINEFEB) and Helmut Nagy (Semantic Web Company) for tcworld magazine, July 2024.

In this article, we present a proof of concept based on a use case from a multinational corporation. The aim was to enable intelligent information retrieval for detailed questions and problems. This information is based on intelligent content created from structured content enriched by a knowledge graph. The proof of concept looks quite promising and has already led to further developments.

A Use Case

The Machinery Directive 2006/42/EG as the basis for technical documentation was originally aimed at technical documentation in a handbook structure. Even when written in DITA, documentation that is available only as a handbook or PDF makes search and retrieval difficult for service technicians. As an example, let us take a look at Karl, a support engineer. The hydraulic system of a bolter miner is not working properly, so Karl is called on site. He is new to the product. The manual has about 1200 pages.

Karl now needs to disconnect the level sensor, find the root cause using the troubleshooting information, and fix the problem, all the while staying alert for safety issues. To do this, he needs to quickly find:

Information about the problem

• • Where to find the level indicator, page 260
  • Where to find the level display on the machine, page 1084
  • Where to find the sensor to be checked, page 261

What to do

• • Short description of how to fill the machine’s fluid, page 536
  • What kind of fluid to use, page 759

What to watch out for

• • Hazard statements for working on the hydraulic systems, page 186

In an ideal scenario, Karl is presented with a tailored information package including the right answer to his problem (ideally only one answer) to enable him to solve the issue fast and directly in front of the machine. Our proof of concept addressed the hydraulics part, a system with a high risk of injury due to its high-pressurized hydraulic oil. In this scenario, it is vital to point out dangerous situations to the service technician and raise awareness through warnings.

Step 1: From structured content to a first content graph

So, how can we achieve our ideal scenario? The basis is structured content. This is not always a given. In our case, Tridion Docs had been used to create the documentation, providing us with structured content based on DITA. (We consider DITA to be the ideal standard, but our approach is not tied to any specific standard). We also integrated e-learning data (slides and links to videos), which were unstructured.

Based on the DITA ontology, the provided DITA data for our proof of concept was transformed into a graph, see Figure 1.

Figure 1: DITA topics as RDF graph

The granularity of the graph and what is represented in it can be tailored to the use case. This means not all data from the CCMS has to be included in the content graph, and the content graph structure can be adapted to the use case. We followed an iterative approach, starting small and growing the graph based on our needs and use cases. From our experience, this approach allows for early results, which can then be built upon. Graphs are suitable for agile use case development, as they are easy to extend.

After this, we transformed the XML schema into a graph structure that opens up new ways to interact with content.

Step 2: From terminology to smart tagging

Step 2 was to enrich the content graph based on a developed taxonomy. In addition to the existing terminology database of the company, we used industry taxonomies from WAND (a provider of industry vertical taxonomies) and the taxonomy included in the iiRDS standard.

These taxonomies were used to automatically tag the content and thus provide high-quality metadata to each DITA element. Semantic tagging goes beyond the traditional metadata approach. Just like the ontologies used, the taxonomies are part of the intelligent content graph we created. Providing a semantic footprint for each content element allows recommending similar content based on it. Ideally, this tagging is already done when content is created in the CCMS. If this is not the case, this step can be added downstream when content is delivered.

We now have the enriched content graph as a basis for what we wanted to achieve for our use case (see Figure 2).

Figure 2: Enrich the content graph via tagging

Step 3: iiRDS for targeted intelligent search

The Intelligent Information Request and Delivery Standard (iiRDS) is a great match for DITA. Topic types for DITA and iiRDS are identical, making it easy to extend our enriched content graph using iiRDS ontology. This is the last step in our journey to intelligent content.

Using iiRDS, we can connect to product data and bring context into the content structure, allowing us to identify exactly the information needed for our use case. Troubleshooting information is created automatically: For each troubleshooting topic, the appropriate technical input, the respective process information, and the related warning messages are dynamically generated. This enables customized results, switching from a document-centric approach to a user-centric search process, providing support engineers with the right information to get the job done (see Figure 3).

Figure 3: Intelligent content structure

Excellent work! In this tutorial, you built your first Semantic AI search application using the PoolParty Application Development Framework (ADF). You learned how to create and test a new configuration using the Search Application UI as well as how to integrate the ADF functionalities into your custom application using the provided APIs.

Conclusion

Our proof of concept shows how the combination of an RDF view of DITA metadata, a company-specific taxonomy, and iiRDS can create a dynamic intelligent content graph to support the focused retrieval of information for service technicians. Figure 4 shows the proof of concept setup and high-level process with all systems involved.

Figure 4: Taxonomies coming together

Stay tuned! Next use cases are already underway, investigating the potential for DITA and iiRDS in multimedia-based e-learning to support the performance of service technicians in the age of forthcoming skills shortages. In addition, we will explore how generative AI can bring additional value to use cases.

Blog reposted with permission from tcworld.