DataScribe

Natural Language Generation from 2D Data

• Get insights on the trends and anomalies in data as a part of a narration.

Dependencies

• Python 3.1 or above
• matplotlib
• numpy

Instructions:

Refer to 'sample_narration.py' for full implementation of module.

Accessing the library:

Import the library from a python script running on the same directory from datascribe import *

Importing data from .CSV

Create a processed data object from the Process_CSV module data = Process_CSV('filename.csv')

Setting Attributes:

• Set the labels_row attribute to True if the first row in the CSV contains the column labels and False otherwise.

      ```data.labels_row = False/True```
  

• If there is a column with only the labels to each data object, specify it using .variable_col

  • Example: If the first column contains labels, Set data.variable_col = 0

• Set X and Y axes using the .set_axis with parameters:

  1. "x" or "y" (to define the axis being set)
  2. columnID (The column index which the axis refers to)
  3. DataType (Integer/Float/Datetime)
  4. Additional formatting if the data is of type Datetime.
  • Example:
    • If X axis is column:1 of type datetime: data.set_axis("x",1,'datetime','%b %d %Y')
    • If Y axis is column:2 of type float: data.set_axis("y",2,'float', None)

Analysing Data:

After setting the necessary attributes, execute the following for analysing the data: Narration_Data = data.transform()

Accessing the Narrations:

• The output of .transform() is a list with each element referring to each label if there are more than one labels
• To access the narrations in a graphical interface, execute: GUI(Narration_Data[0]) for viewing the narration of the first label;
  • Change the index for viewing subsequent narrations of other labels