Understanding Data—Modalities and Preprocessing

As data continues to grow and technology advances, it is natural to consider the various ways it can be represented. However, upon closer examination, we can narrow these representations down to just three: numerical, textual, and visual. This is not a new concept; even ancient civilizations like the Egyptians and Sumerians stored their data through symbolic or pictorial representations, as well as scripts and numbers. While the methods of storage have changed over time, these three modalities of data representation have remained consistent. This reinforces the importance of studying and understanding each modality. By doing so, we can extract valuable insights and meaning from the data and use it to drive progress and innovation. In recent years, machine learning has made significant strides, with over 100 papers published daily on Arvix alone. The number of algorithms developed is impossible to keep track of, but they all revolve around the same fundamental data types discussed earlier. Looking back at history, it’s unlikely that we’ll see a new data representation emerge anytime soon. As practitioners working with data, it is crucial that we have a strong foundation in data modalities. In this chapter, we will delve into these modalities in detail, discussing their characteristics and how they can be utilized to extract valuable insights from data. By understanding the different modalities, we can effectively apply the appropriate techniques and algorithms to the right types of data, leading to more effective and efficient data analysis.