combining learning theory with modern data technology

Modern companies are more frequently leveraging technology to make data-driven decisions — ranging from cloud solutions to access data, machine learning techniques to predict the future, and BI tools to understand data. While there’s no shortage of avenues to engage company data, individual analysts still need to interpret the information in order to translate data to tangible decisions for the business.

“The danger inherent in a visual medium is that the power of the image will overwhelm the substantive content that it represents”
-Ryan Womach, Data Visualization and Literacy [1]

Relying on users to interpret data presents a potential risk for modern data teams as individual user analysis varies greatly based on each individual’s data literacy level. That data analysis process can be broken down into 3 distinct tasks. [2]

  1. Locating Tasks — Require user to find the relevant pieces of information
  2. Integrating Tasks — Require reader to aggregate several pieces of information
  3. Generating Tasks — Require user to process information and make document-based inferences based on personal knowledge

Modern BI visualization tools have alleviated a majority of these user tensions — locating tasks are largely made trivial through visual and spatial design and integrating tasks can be accomplished through clever design. However, traditional data visualization does little in translating visual information to actionable decisions for the business, and consequently, the user must arrive at those decisions themselves.

In order to remedy this issue, Slalom borrows from the field of learning science, specifically the use of computer-based scaffolding [3] combined with visual/text multimodal learning to make the data more accessible and interpretable to the end user. Education theory suggests “students can learn more effectively, …particularly when words and illustrations are presented together” [4] — Slalom adopts this same theory in dashboard design by pairing traditional visualization platforms with Natural Language Generation (NLG) technology.

Figure 1: Natural Language providing contextual analysis to supporting data

In Figure 1 above, natural language is largely repeating the content of the dashboard in a different medium. By pairing text with visualizations, less data-savvy users can interact with the content in a medium they’re comfortable with (text), which in turn provides the scaffolding as they learn a new medium (BI technology), ultimately increasing user adoption.

Figure 2: User data literacy required for content ingestion

Increased user adoption affords organizations to reach the next form of data maturity; once users become confident in their ability to consume dashboards, they’re able to branch out and become creators thereby creating moving organizations to a self-service data culture.

Figure 3: NLG dashboard being used to drive the same actions

Narrative pairing also creates less room for interpretation — the business activity can be clearly outlined in the text with supporting data as opposed to users interpreting the data and making potentially inconsistent decisions. For example, in figure 3 above natural language is used to create a custom profile for a customer and provide the corresponding analyst with a clear action plan based on the data.

Note that NLG is largely different than Natural Language Process (NLP) — another common term in the modern data stack. NLP largely deals with converting text/voice into data points for analysis, while NLG does the exact opposite.

Figure 4: The role of NLP & NLG in modern speaker assistants

Slalom’s NLG capabilities allow for the narratives to be systematically generated based on the data present so the corresponding text does not need to be written out, but rather generated based on user selections.

The two US market leaders in this space are Automated Insights and Narrative Science. While both technologies’ NLG solution looks similar, their process is vastly different. Automated Insights allows for full narrative customization through a logic editor, which allows users to craft narratives in verbiage that is consistent with their organization’s language. Conversely, Narrative Science takes a more analytical approach where it leverages machine learning to automatically convert data into text. Other companies in that space include Phrasetech, Arria, AX Semantics, and Yseop.

Disclaimer: Slalom is a system integration partner of Automated Insights

Slalom is a modern consulting firm focused on strategy, technology, and business transformation. Slalom NYC is leveraging technologies such as NLG as part of their Innovation Accelerator program that drives organizations towards their specific business outcomes using a modern technology stack. @slalomnyc

Saurabh is a consultant on the Data Visualization and Discovery team for Slalom New York.

References

1. Womack, Ryan. “Data Visualization and Information Literacy.” IASSIST Quarterly, 2014.

2. Jeremy Boy, Ronald A. Rensink, Enrico Bertini, Jean-Daniel Fekete. A Principled Way of Assessing

Visualization Literacy. IEEE Transactions on Visualization and Computer Graphics, Institute of Electrical and Electronics Engineers, 2014, 20 (12), pp.10. ff10.1109/TVCG.2014.2346984ff. ffhal-01027582f

3. Sharma, Priya, and Michael J. Hannafin. “Scaffolding in Technology-Enhanced Learning Environments.” Interactive Learning Environments, vol. 15, no. 1, 2007, pp. 27–46., doi:10.1080/10494820600996972.

4. Stokes, Suzanne. “Visual Literacy in Teaching and Learning: A Literature Perspective.” Electronic Journal for the Integration of Technology in Education.