Browse Prior Art Database

An Widely Applied Approach for Automatically Detect and Sort Likert Scale Values

IP.com Disclosure Number: IPCOM000236769D
Publication Date: 2014-May-15
Document File: 7 page(s) / 108K

Publishing Venue

The IP.com Prior Art Database

Abstract

Likert scale is a very commonly data scale type, which should be ordinal as a variable in analytics systems. But many times, system has to process likert scale data in plain text, which doesn't contain the order information. So system has to treat it as a nominal variable, which will impact the accuracy of the analytic result, or detect the likert scale based on plain text. This arcticle invent a general way to detect various likert scales.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 39% of the total text.

Page 01 of 7

An Widely Applied Approach for Automatically Detect and Sort Likert Scale Values

Likert scale is a psychometric scale commonly used in various questionnaires. It is the most widely used approach to scaling responses in survey research. For example, "Strongly Agree, Agree, Neither Agree nor Disagree, Disagree, Strongly Disagree", "Very Good, Good, Average, Poor, Very Poor". The scale commonly has 3 levels or 5 levels. As we see from the example, the level in order represents a response from the most positive point to the most negative point.

Likert scale values commonly appear in various user data, e.g. survey result. Below table shows an example of the data (e.g. CSV/Excel file) that contain Likert scale.

1

A very common form of data is preference data; that is data that expresses how much something is preferred. These occur very commonly in surveys, as the answers to questions like "how strongly do you agree with ....?" or "how likely are you to buy ...?" and so on. Often, such a response is coded as a number, where, for example, 1 means very positive and 5 very negative. But it is also quite common for the data to be recorded as text, in which case it poses a challenge for analysis as the data appears to be simply categorical -- a set of nominal values, whereas in fact it is really an ordered set -- ordinal data. Thus the ability to take a set of text value, detect that they are in fact ordinal, and assign correct order to them is a critical need for the analysis of such data.


Page 02 of 7

If the data is used to build report or run analysis, without manual interfering, the reporting system or analysis system can only recognize the values as nominal categories (no meaningful order, e.g. pear, apple, orange), rather than ordinal categories (with meaningful order, e.g. January, February, March, … ). That makes a large difference in reporting and analytic systems.

Use a visualization reporting tool as an example, see the graphs below:

2


Page 03 of 7

Recognizing the values as nominal categories, the system commonly sort the values in alphebetical order. See the horizontal axis of the graph, the axis ticks are in alphebetical order, not the order in meaning.

The following graph shows values in meaningful order:

Comparing the two graphs, it is difficult for user to find the trends in the first graph, but very easy in the second graph. No doute, the second graph is better than the first one.

See the following Likert scale examples:

Very Poor, Poor, Average, Good, Very Good

Very Poor, Poor, Neutral, Good, Very Good

3


Page 04 of 7

Very Bad, Bad, Average, Good, Very Good

Worst, Worse, OK, Better, Best

They represent similar meaning but are different in wording. Also, besides the 4, we could list many other wordings that represent the similar meanings.

Therefore, there are so many different ways to represents Likert scales that a system cannot list them all. Also the Likert scale pattern is not so obvious that a system co...