TIBCO Spotfire: Configuring Narratives


Documentation for Arria NLG’s Add-in for TIBCO Spotfire Analyst


 

This documentation is organized as follows:

Overview of configuring narratives

Configuring and generating narratives

Add Arria’s custom visual into a sample Spotfire analysis
Configure a narrative to describe the visuals

Basics of configuring a narrative

Reference for configuring narratives

Other documentation for Arria’s add-in to TIBCO Spotfire Analyst 

 

Overview of configuring a narrative

The core insights delivered by Arria’s add-in are determined by the dimensions and measures you select in the Spotfire analysis. You configure your narrative by telling the add-in wizard what you want. This happens in three steps:

 

Configuring and generating narratives

You’ll begin by downloading a sample Spotfire analysis. Next, you’ll add Arria’s custom visualization to your analysis. After that, you’ll provide some information about your data and the kind of things you would like to say about it in your narrative. Then, with the click of a button, you’ll generate a narrative from your data.

Prerequisites

 

Add Arria’s custom visual into a sample Spotfire Analysis

1. Download the Configuring Narratives sample Spotfire analysis (.DXP file).

2. In Spotfire, open the analysis.

3. In the Authoring bar, click Visualizations > Add visualization…

4. Select ArriaNLG at the bottom of the Visualization types… list.

5. Move and resize the Arria add-in to suit your screen.

 

Configure a narrative to describe the visuals

Now you can configure a narrative to describe the same data selection illustrated in the bar charts.

1. In the ArriaNLG add-in, click on the sliders icon to begin selecting data for your narrative.

2. From Dimensions, select Segment and Product.

3. From Measures, select Sales (NOT Sale Price or Gross Sales) and Profit.

4. Click the Configure Narrative tab.

5. You can now configure your narrative. You do this in three pages of the Arria add-in wizard:

  • Tell me about your data. You select the dimensions and measures you want to be covered in the narrative, and then characterize them.
  • Tell me what type of analysis you want. You select the kind of analysis you want to perform (e.g. descriptive statistics, trends, variance) and then provide the necessary parameters.
  • Tell me how to say it. You preview the generated text to decide on how much narrative you want – something brief, or something more comprehensive.

Note: To move between these pages, click Next or Previous at the top of the wizard:

6. On the Tell me about your data page, in the Sales and Profit fields, change the Entity Type to Currency, and in the Unit field enter USD (the default unit type).

7. On the Tell me what type of analysis you want page, uncheck the Variance field.

8. On the Tell me how to say it page, select Important things. You’ll see a preview of the configured narrative.

9. Click Generate Text to see the full narrative.

10. Refer to the Basics and Reference sections for more information on configuring each field of the wizard.

 

Basics of configuring a narrative

 

  • Scrolling in the wizard. Note the scroll bar at the right side of the wizard. Use this to make sure you’re seeing all the dimensions and measures for your selected data, and all the characteristics that can be customized.
  • Prioritizing of data fields. Use the grab bars at the top left of each dimension and measure to move them up or down in the wizard. The higher the appearance in the wizard, the higher the priority for description in the narrative generated.
  • Copy and paste. If you want to copy the narrative generated to paste elsewhere, you can hover in the top-right corner of the wizard to find the copy icon. Click this icon to copy the narrative to your clipboard.

 

 

Reference for configuring narratives

 

UI element
Description
Where is this?
Alias
This can be an alternative name of the field in Spotfire, for example, the field “COGS” could have the alias “Cost of Goods and Services”.
Note: The alias can be different from the field name, but it doesn’t have to be.
Step 1. Tell me about your data
(All dimensions and measures)
Comparison
For time-based variance, specify how you would like measures to be compared (e.g. Latest vs. previous, Latest vs earliest).
Step 2. Tell me what type of analysis you want
(Variance > Time-based variance)
Correlations
This type of analysis looks at the strength of a relationship between measures.
Step 2. Tell me what type of analysis you want
Coverage
For variance (time-based and target-based), specify an amount from 0 to 100. The amount you specify is the number of drivers or offsets contributing to the percent-coverage of the variance. For example: when the entities are ranked in terms of contribution, if the coverage is 80%, the system selects drivers or offsets that contribute to 80% of the variance; the system leaves out the bottom 20%.
Step 2. Tell me what type of analysis you want
(Variance > Time-based variance)
Dimension
A dimension is an attribute or a column in your data that contains categorical values. Examples of dimensions include Product, Country, and Date. In your analysis, if you want to analyze numerical attributes such as Sales or Profit, how do you want to analyze them? By Country? By Product? By Date? Those are dimensions.
Step 1. Tell me about your data
Distributional analysis
This type of analysis describes how the data is distributed through the range of dimensions (e.g. Products, Countries, Segments). Selecting this option gives you more insights into the skew of the data, if it exists.
Step 2. Tell me what type of analysis you want (Descriptive statistics)
Entity selection
For variance analysis, choose between Top 1 and Top 5 for how many entities you want mentioned in the narrative. For example, “The top-three performing regions were…”. What you are doing here is choosing which algorithm will determine the contributing drivers and offsets.
Step 2. Tell me what type of analysis you want (Variance)
 
Entity type
For characterizing dimensions and measures, the choices of known entity types include DateTime, Location, and Currency. If none of the known types apply, leave this selection blank.
Step 1. Tell me about your data
(All dimensions and measures)
Important things
Selecting this option gives you a narrative of the filtered facts based on a ranking algorithm.
Step 3. Tell me how to say it
Include distributional analysis
For descriptive statistics, this is the on/off switch for distributional analysis. If you want distributional analysis included in your narrative, check this box.
Step 2. Tell me what type of analysis you want (Descriptive statistics)
Let us decide
See Ranking.
Step 2. Tell me what type of analysis you want (Descriptive statistics)
Measure
A measure is an attribute or a column that contains numerical values, such as Sales, Profit, or Temperature.
Step 1. Tell me about your data
Measures to compare
For correlations analysis, specify the measures you want to consider in the analysis. If you have more than two, click +Add measures to add more.
Step 2. Tell me what type of analysis you want (Correlations)
 
Period
For time-based variance, select a time period. Options include Month, Quarter, and Year.
Step 2. Tell me what type of analysis you want (Variance)
Ranking
For descriptive statistics, do you want the system to rank statistical insights based on the significance of the facts or insights it finds? Or based on the selected order of the dimensions in the order you have them placed in Step 1 Tell me about your data?
Step 2. Tell me what type of analysis you want (Descriptive statistics)
Use selected attribute order
Priority is set by how you order your fields in 1. Tell me about your data. (In Step 1, use the drag handles at the top left corner of the fields in the widget to move them up and down in priority order.)
Step 2. Tell me what type of analysis you want (Descriptive statistics > Ranking) 
Let us decide
Let Arria’s algorithms decide priority.
Step 2. Tell me what type of analysis you want (Descriptive statistics > Ranking)
Selected attribute order
See Ranking.
Step 2. Tell me what type of analysis you want (Descriptive statistics > Ranking) 
Summary
Selecting this option gives you a narrative of the filtered facts based on a ranking algorithm, in a more concise way than the Important Things option.
Step 3. Tell me how to say it
Target-based variance
This is for when you want to report on the variance of two measures, such as Actual Sales vs Target Sales or Actual Expenditure vs Budgeted Expenditure, for a single time period.
Step 2. Tell me what type of analysis you want (Variance)
Tell me everything
Selecting this option gives you a narrative of every fact that Arria’s Insight API extracts from your data.
Step 3. Tell me how to say it
Time-based variance
This is for when you want to report on the variance of a measure in two time periods, such as in the current month vs the previous month or the current quarter vs the same quarter of the previous year.
Step 2. Tell me what type of analysis you want (Variance)
Unit
NLG Studio supports 10 ISO codes, and in narratives, these are converted to symbols. Anything other than the 10 ISO codes supported by Studio are created as a unit and appear in the narrative after the corresponding numerical value. Note: You must enter an entity type in order to use Unit.
Step 1. Tell me about your data
(All measures)
Use selected attribute order
See Selected attribute order.