• Oksana Biens | Nathalie Ung
  • Published: 20 January 2020


There is a plethora of data that exists today, and even more, created daily. By 2020, it’s estimated that for every person on earth, 1.7 MB of data will be created every second1. With this exponential growth in volume, we have excessive information to process and analyze.

Data visualization is the process of transforming data into a graphical form to understand patterns, trends and insights that would otherwise be difficult to establish. It makes complex data more understandable, accessible and usable.

Data visualization is not a new topic, the market is flooded with visualization tools like Tableau, Qlikview and Power BI. Recently the industry saw an increase in the use of natural language-based technologies in conjunction with data visualization. These innovative interfaces unlock the value of data for everyone, eliminating the need for technical skills to interact with tools and reducing time-to-insight.



Several types of natural language-based technologies exist today like Natural Language Processing, Speech Recognition andNatural Language Generation. While the first two are used as an input interface – to request what you would like to see – the latter is used as an output interface to express in words what appears in data visualization. When used together, these technologies offer an enhanced user experience. Let’s take a closer look at each of the technologies.

Natural Language Processing (NLP) and Speech Recognition (SR)

In today’s data-driven world, customized and impactful visual presentations are key to effective communication. However, to represent information in a way that can be quickly and easily understood it requires lengthy data exploration and analysis, as well as design skills. Data visualization tools with Natural Language Processing and Speech Recognition capabilities are gaining in popularity as they are intuitive and user-friendly. These tools allow translating human speech and writing patterns into a database query language to extract data, then filter and categorize results, and present them in the form of a chart or diagram.

NLP and SR can be used to optimize and improve the customerservice aspect in digital banking. New generation banks do not have physical offices and mostly interact with their customers over the phone and via online chat, with robots handling user requests or rooting them to appropriate human experts. As a result, digital banks heavily rely on text and speech recognition techniques to analyze client expectations, general sentiment and behavior, to prevent customer attrition, build brand loyalty, and offer customized products and services.

While NLP and SR analytics help identify and classify the prevailing reasons why clients contact customer service as well as their emotions, dynamic dashboards offer the most effective way to visualize the findings. For customer service management, data visualization tools are essential to track market trends, monitor alerts in real-time and drive decision making.

With all the exciting possibilities offered by NLP and SP, it should be noted that these technologies still have certain limitations. For example, the case of sentiment analysis mentioned above, detecting irony or sarcasm requires a deep analysis of the context. While the tone of a voice can provide some insights about the person’s feelings in case of voice messages, textual comments are difficult to evaluate automatically, sometimes even for a human.

Natural Language Generation (NLG)

NLG enables the automatic generation of textual content from a dataset. Adding the NLG component to a data visualization tool will take the user experience to the next level, by providing dynamic and impactful written narratives for charts and diagrams. A dashboard with a story is always better than a dashboard without one. Moreover, NLG functionality is machine-based, hence it allows for a significant reduction in human involvement and inaccuracies associated with human errors, such as visualization misinterpretation and biases.


Saving time is not the only advantage of automation; reliable explanations of findings can be tailored to the audience, thus improving the communication efficiency. The NLG-driven solution can output an insightful narrative for any population, from C-suite executives to analysts, taking into account specific business focus and level of data expertise. As a result, the audience can easily understand the message and act rapidly on insights.

With the ability to explain key findings concisely, NLG transforms the image of data analytics, making it understandable to everyone in the organization and thus easing the adoption of analytics. Nevertheless, the NLG technology has still a long way to go. For example, most NLG-based visualization systems typically require users to read the textual representation of findings and mentally map them to the visualization, which can be tricky if the finding is complex or the data density on the diagram is high. The technology could be enhanced by highlighting a finding in the visualization for easier reading2.




Regulatory compliance is an aspect that all industries, including financial services, must handle. Regulatory oversight is costly and time-consuming due to continuous monitoring and repetitive reporting tasks. Automated solutions providing relevant and personalized information in conversational form may represent a way to free up time and resources dedicated to regulatory reporting.

As part of the Anti-Money Laundering/Counter-Terrorist Financing (AML/CTF) regulatory framework, financial institutions are required to issue Suspicious Activity Reports (SARs). Compliance executives are tasked to determine if transactions performed by a customer are in line with his/her profile and nature of the business, as well as to evaluate the risk level related to tansactions or involved parties. Monitoring transactions involves processing a large volume of data, including customer personal information, account information, transaction data, location data etc. Leveraging on NLP allows to easily process and analyze data from internal and external sources, uncovering patterns and detecting abnormal behaviors. Visual representations such as
comparative charts come in very handy for monitoring trends and raising alerts.

Traditionally, investigators examine alerts to determine whether they should be escalated for further review and finally issue Suspicious Activity Reports that are submitted to the regulator. In a digitalized environment, NLG tools automatically generate reports which can be dynamically updated in case new information is added to the analysis.

This example shows how the combination of NLP and NLG components enables financial institutions to augment efficiency in the identification of suspicious or fraudulent activities, and automatically produce high-quality reports, helping them meet the regulatory requirements, while also reducing costs through process automation.



New technologies currently being adopted by an increasing number of businesses, expand data visualization possibilities and provide a whole new level of user experience. With the help of natural language-based interfaces, users can tell better stories about their data by transforming queries into visualizations, and visualizations into reports.

However, even with the advances in this field, the technology still has some limitations and leaves space for potential misinterpretations. It is important to note that the key to correct data representation and interpretation is ensuring that the underlying data is exhaustive and of good quality, otherwise the insights generated from this data will be incorrect and will lead to inappropriate decisions.


Julien Blanchet, Partner
T: +33 1 77 72 64 55

Jerome Lecussan, Partner
T: +33 1 77 72 64 40