Spring til indhold

Qvest has specialized in helping organizations speeding up the process and quality of strategy implementation. You get deep insight and data from across the organization to benchmark how the implementation is progressing.

Their basic philosophy is that to reach important goals, we must ask each other questions. Communicating goals and plans do not bring the plans into action, a culture of curiosity and collaboration is needed. Hence the need for asking questions throughout the entire organization and use these questions as qualitative input to analysis. 

A method and a digital tool were developed to support the change process based on asking questions, making it possible to go from strategic initiatives to involving and mobilizing the entire organization.

Saving time performing analysis

The analysis performed by Qvest based on questions across an entire organization entails a large amount of data. To help structure these qualitative data with NLP (natural language processing) and clustering, Ambolt was brought onboard.

The AI solution provided by Ambolt provide sophisticated help with initial sorting of the data, making it possible for Qvest to spend less time on this part and go directly to the analysis, carried out by Qvest consultants. This way AI supports, but certainly do not replace natural intelligence. We keep “human in the loop” and use data for good.

Text is represented as a point in 3-dimensional space

The animation above shows each piece of text represented as a point in 3-dimensional space. A text is converted to a high-dimensional vector of numbers (a feature vector) using NLP transformer models. The vectors are reduced to 3 dimensions for visualization purposes by using a dimensionality reduction method called UMAP (more dimensions are used when measuring distances to not lose valuable information).

The distance between each data point represents the semantic similarities between the texts i.e. points that are close to each other represents texts that are similar in meaning. The color of the points represents the cluster assignments found by using the HDBSCAN algorithm.

Ambolt automate the process by

  1. Analyzing the text and identify questions with similarities (nearest neighbors in the questions)
  2. Cluster the questions, to form an overview of different themes in the texts (dataset) and act as suggestions for questions to go in the same category.

The combination of these two processes makes it easy to initiate clustering of data and then move forward with categorizing.

Reach out to Ambolt

Curious About the Value of Your Data?

Reach out to Ambolt and learn more

Contact