What is a widget?
Stratifyd’s data visualizations are called widgets, and they are housed in tabs. A widget is visual representing data analytics results. They are similar to charts typically featured in spreadsheet applications, but widgets have many more options for data visualization. Widgets can be simple visuals utilizing one or two data fields, or complex visuals leveraging three or more data fields.
You can filter widgets and create custom calculated fields within them. Stratifyd has over 5,000 visualizations that can be created in the widget editor. We pre-populate unstructured data tabs with four standard widgets: Number of Records, Average Sentiment, Sentiment Distribution, and Semantic Topics.
To create a widget, simply click the icon in the bottom right corner of a dashboard tab. This will bring up the Widget Editor.
To edit a widget, click the editing icon on the top right corner of a widget, then select Edit.
How to Use the Widget Editor
There are five main modules that comprise the widget editor:
- Data Fields Tab
- Data Dimensions Header
- Visualization Options Tab
- Widget Options Footer
- Widget Visualizer
With over 5,000 visuals possible, the best way to learn how to use the widget editor is to play around with its options. We will cover the basics of each module listed above and teach you the fundamentals of the most frequently used widgets for unstructured and structured data.
Data Fields Tab
This is where your datasets and data fields are shown. Each widget can handle one dataset unless data fusion between two datasets is initiated. Let’s go over how this tab works within the widget editor:
Change the dataset by clicking on the dataset name above the Search bar. This will bring up a Select Data prompt where you can choose which data set to use for the widget.
+New Equation allows you to create a calculated field using the existing data fields by performing mathematical and statistical computations to structured data fields.
- Stratifyd Analysis are data fields calculated by Stratifyd during its machine learning of the documents.
The taxonomy widget is a Stratifyd Analyses dimension when a taxonomy is applied to a dataset.
- Calculated Fields are structured data fields that are calculated by computational means.
Original Fields are the original data fields extracted by a data connector.
You can select any data field and it will transfer to the Data Dimensions header.
When you select a data field, it populates on the Data Dimension header. Stratifyd will attempt to use the best possible fit for the size and color dimensions based on the data dimensions selected if a visual requires these dimensions to be filled. The visual options automatically pre-populate with the available visuals that selected data dimensions can apply to. Each data dimension is color coded:
- Yellow = Unstructured, Textual Data
- Blue = Structured, Numerical Data
- Green = Temporal Data
- Dark Blue = Sequential
A data dimension may have a drop-down arrow allowing for customization.
For example, drop-down options for temporal data allow the user to change the date intervals. Numerical data can be aggregated and formatted using their drop-down menus if available. Textual data can also be aggregated or changed to sequential data.
How you format your data dimensions is dependent on the visual option selected from the Visualization Options Tab.
Stratifyd has twenty-four standard visualization options to choose from. The widget editor will automatically detect the data dimensions chosen and present you with the visualization options that can present those dimensions. Some visuals are commonly used while others are rarely used except in specific analytical cases. Let’s go over some common visualization options, their requirements, and when you should consider using each.
Requires Stratifyd Sentiment Analysis as a data dimension. The visual displays the total average sentiment score for the dataset analyzed on a gauge with negative, neutral, and positive sections.
Bar, Area, Stream, Singe Line
Requires at least one numerical dimension to display an area graph. The area chart is included in the Semantic Topics pre-populated widget with the Number of Records plotted against the Temporal data to show number of records over time with split areas for each topic.
The bar chart visual requires a numerical dimension and is best paired with temporal or categorical data as its x-axis. It is used to show changes in a numerical dimension over time or compare categories – such as genders or contributors – against each other. You can apply a linear or quadratic regression line on the bar chart and change the style from bar to area, line, or stream.
Requires at least one numerical dimension. This visual can plot multiple numerical data dimensions against a common x-axis dimension like Temporal data. For example, the prices of two or more stocks over time could be plotted and differentiated with a legend and color coding. You can normalize the Y-axis to compare the relationships between numerical dimensions that differ widely in range.
Requires at least one numerical dimension. The treemap displays hierarchical data in rectangles, with smaller rectangles within representing sub-branches. The colors delineate separate dimensions while the sizes of each rectangle are proportional to a numerical dimension such as Number of Records. Treemaps are generated using tiling algorithms that attempt to optimize the aspect ratio and order of data.
Why use a treemap?
Big Data benefits from treemap visuals due to their economical size; a treemap can display thousands of records on your screen. Treemaps allow you to easily see patterns that are more difficult to identify otherwise, especially when the size and color of the rectangles are correlated to the treemap structure.
The pie chart needs at least one numerical dimension. Pie chart slices display the numerical proportion of data, usually with percentages relative to the whole. Large data dimensions do not work well with pie charts because the slices are typically too thin to recognize even with color coding and arrows. You would use a pie chart to visualize the breakout of categorical or ordinal data by number of records or proportion of records.
This visual requires a numerical dimension and is typically used to visualize conversion. Funnels can visualize how any set of steps impact conversion.
Bubble charts require three data dimensions, the third of which must be numerical and act as the size dimension.
Scatterplots require one numerical dimension and at least one other dimension. Scatterplots display how correlated two dimensions are to each other with one dimension as the x axis and another as the y axis. A third dimension can determine color or size on a scatterplot. If you have three dimensions, then bubble charts and scatterplots can be used interchangeably. A trendline can be applied to the scatterplot through linear regression and its equation indicates how correlated the two dimensions are. A correlation number represented as r2 can range from -1 to +1, with a more positive number indicating a direct correlation and a more negative number indicating an inverse correlation.
Waffles require one numerical dimension in order to compare quantities based off of areas similar to how the pie chart compares proportions with slices. Each square of a waffle visualization is scaled to represent a defined amount. Colors define the categorical data being measured. The squares are shaded in color segments depending on the amount belonging to each category. We recommend the waffle over the pie chart in most scenarios because it illustrates the same “at-a-glance” statement and allows for interactive analysis.
Our simplest visualization requires one numerical dimension total and displays a numerical total, average, mean, median, maximum, minimum, etc. The most common uses for this visual is to display the total Number of Records and average star ratings.
The word cloud requires a Stratifyd Key N-Grams dimension, either unigrams or bigrams, and a size dimension like Number of Records or Proportion of Records. A word cloud consists of the top N-grams sized by number of appearances within the corpus. This visual is useful to see the most frequently used N-grams. The word cloud that comes attached to the pre-prepared Stratifyd Semantic Topics widget can be sized by count or pointwise-mutual-information score, a statistical measure of word importance.
A geographical dimension is required to generate the geomap widget. Stratifyd will automatically generate a heatmap based on sentiment, size, or another metric of your choosing to overlay on the geographical map. You can view global, regional, country, and city specific locations and filter by sentiment, size, and other measurements incorporated into the map. This widget is useful to visualize where certain trends are occurring compared to others.
You can create lists in Stratifyd with multiple columns and rows. Lists can be filtered and sorted by columns and are useful to see the most frequently occurring datapoints or to create a “Top 10” list, for example.
Used to show flow and proportions between two separate dimensions, parallel sets are a colorful way of visualizing data. Unlike Sankey Diagrams, parallel sets do not have arrows that indicate the direction of flow.
The taxonomy tree displays your taxonomy in a hierarchical, faceted visual. Each node represents a label, and several narrower term labels may branch off from each node. The taxonomy tree is best visualized when adding a size dimension, such as number of records, to show the relative amount of data sorted into each node.
Certain elements to a widget can be changed via the components tab located one icon to the right of the visualizations icon on the right-hand panel. The widget and text components toggle the editing toolbars for each directly above the widget. These toolbars provide options for altering the entire widget or text placed within a widget from the Add a New Component toolbox.
You can also edit the margins of the widget so that the visual is surrounded by more or less border space in the dashboard. Add new components - such as labels, text boxes, expression binding, images, videos, web pages, and raw HTML - by dragging and dropping their icons into the widget.
Colors for any dimension in a widget can be changed under the color icon. Depending on the type of data selected, you can change the color by single color, range palettes, or gradients to make your data stand out.