We are living in a knowledge society where knowledge is usually just a click away. But what happens when the amount of information becomes too great? The relevant information gets lost! As a result, trying to find the information we need is time-consuming and exhausting. In the end, we may even end up with the wrong information. Automatic knowledge base analysis can help.
Companies are faced with the task of managing information overload and making knowledge available in the right places. However, the editorial effort required often exceeds the staff available resources. The result is outdated, incorrect and redundant information in the company’s knowledge base. This increases costs, for example because searches take longer and errors occur more frequently.
It is therefore advantageous for companies to increase the quality of their knowledge base. On the one hand, this reduces follow-up costs. On the other hand, using knowledge effectively can become an important competitive advantage.
The Question is: Where to Start?
Most of the time, those responsible for the knowledge base are themselves not aware of where the problems lie. It is rarely clear which information is important and up-to-date and which is outdated, incorrect or redundant.
Manual evaluation is scarcely an efficient use of time. Let’s say it takes 10 minutes to get an overview of one page. For a knowledge base with 1,000 articles, this already results in a workload of more than 166 hours.
Clearly, manual analysis will lead to very high costs – if it is even possible at all. In order to get a handle on the current situation quickly and inexpensively, it is a good idea to analyse the contents of the knowledge base automatically.
Even though the content is presented as text in human language, Natural Language Processing means that computers can evaluate this text. This allows the content of a knowledge base to be analysed quickly and cost-effectively in an automated manner.
Developing the Model – How to Get Meaningful Results
The development of a model that estimates the quality of information automatically proceeds in 5 steps.
- In the first step, you need to identify all relevant data and make it available. Most knowledge bases provide the option of exporting the individual articles as XML or HTML files.
- Next, you should prepare the data and investigate which key metrics can be calculated from the data. For example, comprehensibility can be determined by using the FRE score or the content can be checked for outdated terms. Natural Language Processing can also be used to identify duplicates within the knowledge base. Based on the findings, you can calculate the key metrics.
- In the third step, you can evaluate the key metrics calculated in step two. What specific FRE score does my content need to achieve to be rated as “good”? In this way, values can be given for each key metric, which lead to plus or minus points in the evaluation.
- Then you can translate the findings into a Python script that calculates a score for each article in the knowledge base.
- Finally, apply the script to the knowledge base under investigation and derive further steps from the results.
Knowledge Base Analysis: What Does the Result Tell Me?
Once you have analysed the company’s knowledge base with the help of the model, it is time to draw some conclusions. The model generates a quality score for each article. For the entire knowledge base of a large company, the distribution could look like this, for example:
The majority of the articles are in the score range from -5 to 4. Some articles also achieve very good or very poor scores. In addition to this complete overview, the analysis provides details on how the rating of an article is arrived at.
These results provide a high level of transparency and a good basis for planning knowledge management projects. They allow the current quality of the knowledge base to be identified. It also becomes clear where there is potential for improvement. The time that will be required for a project can also be determined more precisely.
The model can furthermore be used to evaluate the success of actions.
Companies are often faced with the challenge of measuring and evaluating the effectiveness of knowledge management actions. The model can be applied to the knowledge base first before the project starts and then after the project ends. The results can then be compared. In this way, the actual improvement in quality can be determined.
Conclusion
Due to the continuing flood of information, the topic of computer-aided analysis of information will become even more relevant in the future and will provide a wide variety of use cases.
Whenever you
- want to gain knowledge from texts,
- need very good documentation,
- want to optimise your business knowledge
Information Management can help. IM makes it possible to establish a company-wide knowledge management system. This is the basis for successful customer service, cloud migrations or IT operations, for example.