Text Analysis
Innehållsförteckning
General information about Text Analysis
Machine learning technology makes it possible for our tool to understand text data.
Using open-text questions is one of the best ways to gather quality feedback from your customers and employees. These responses are a goldmine for meaningful and actionable insights, but to analyze them manually takes time and the process is impossible to scale.
This is where Netigate’s text insight engine comes into play. Intelligent, swift, and scalable, it helps you to quickly extract information from text responses and get to the heart of what your customers and employees really think and care about. The text is analyzed in real time so there’s no need to wait for the survey to be completed, you can start the analysis immediately.
Sentiment Analysis
The Text Analysis engine extracts one sentiment per answer and gives you the polarity in that answer (positive, negative, neutral or mixed). Mixed sentiment is when a respondent is both positive and negative in the same response.
Why Text Analysis?
People express their feelings more and more when giving feedback, about companies, products or services. Automatically processing and analyzing feedback from respondents allows companies to better understand how people are feeling and then act on those insights.
Problem
Processing and reading text answers written by respondents in a survey and extracting the sentiment is a time-consuming process. The human factor comes into play when processing the data manually. People have different opinions about what is positive and negative.
Use cases
- Understand what respondents are feeling when they answer a question.
- Handle large volumes of data and scale up or down quickly.
- Shorten time-to-action based on insights with fast processing and analysis. Improve data quality with consistent use of criteria.
Keyword extraction
Key phrase extraction is about quickly identifying the main concepts in an answer. The keyword can be one word or one phrase containing several words. The insight engine reads the entire answer and uses an algorithm (mathematical model) to determine what would be a useful keyword.
Keyword analysis extracts the most important and used words from open text answers. This helps summarize the content through a visualization of what respondents are talking about. Identifying the most important words and phrases from many open text answers can provide valuable insights into what matters the most to your respondents.
Use cases
- Get an overview of what your respondents are talking about the most.
- Highlight keywords across multiple answers to see patterns in the conversation.
- See the frequency of keywords.
Lemmatization
Lemmatization is the process of reducing multiple variants of a word to its unique lemma — the word form you would find in a dictionary. For instance, the word “universities” is found under “university”, while the word “universe” is found under universe, leaving no room for misinterpretation.
Highlights
- Netigate supports Text Analysis in English, German, Norwegian and Swedish.
- The Text Analysis engine reads the entire text answer when it performs the analysis. The question is NOT included in the analysis, just the answer.
- Analysis is performed on all text box questions within a survey.
- Text Analysis must be activated on a survey before responses are received for them to be processed.
- Activating Text Analysis retroactively on older surveys will not process the text answers.
- Using Text Analysis is always associated with a cost.
The Text Analysis setting is found under survey settings and is per default OFF. If you want the engine to perform Text Analysis on the survey, you need to turn it ON. Answers will NOT BE processed retroactively; only new answers will be processed from the time Text Analysis is activated.
Get started with Text Analysis
Once Text Analysis is activated on your account it can be used on a survey. Text Analysis must be enabled in the survey’s settings, to be able to process text answers and show the result in the report. Text Analysis will process all text answers from all Textbox questions within a survey.
Important! Make sure that Text analysis is enabled before the survey receives text answers from respondents. Text answers cannot be processed by Text analysis retroactively.
How to activate Text Analysis on a survey
- Create a new or edit an existing survey.
- Click on Settings and scroll down to the Text Analysis section in the settings modal
- Click the option Enable Text Analysis for this survey.
- Click Save changes and close the modal.
Text Analysis is now active for the survey and will start to process text answers once the respondents start responding to the survey. The processed text answers will be shown in the report.
Report boxes and options
When Text Analysis is active on a survey, new box visualizations become available in the report. These box visualizations are shown for each dash that is a Textbox question.
This is where to find all the different Text Analysis boxes and can add them to the report.
Sentiment Chart
The Sentiment Chart shows the distribution of how the respondents were feeling when they answered a question. The feelings are shown as polarity, positive, neutral, negative or mixed. Mixed sentiment means that the respondents are both positive and negative in the same answer.
The Sentiment Chart is a doughnut chart and shows all the answers from which sentiment has been extracted . The chart type cannot be changed.
The Sentiment Chart is shown by default but can also be added through the context menu of the dash.
Note that the Sentiment Chart only shows answers that have been processed by Text analysis and that it is not the same as respondents’ answers on the survey.
Sentiment Chart
The Sentiment Chart shows the distribution of how the respondents were feeling when they answered a question. The feelings are shown as polarity, positive, neutral, negative or mixed. Mixed sentiment means that the respondents are both positive and negative in the same answer.
The Sentiment Chart is a doughnut chart and shows all the answers from which sentiment has been extracted . The chart type cannot be changed.
The Sentiment Chart is shown by default but can also be added through the context menu of the dash.
Note that the Sentiment Chart only shows answers that have been processed by Text analysis and that it is not the same as respondents’ answers on the survey.
Sentiment Table
The Sentiment Table shows the distribution of how the respondents were feeling when they answered a question in both values and percentages. The feelings are shown as polarity, positive, neutral, negative or mixed. Mixed sentiment means that the respondents are both positive and negative in the same answer.
The values shown indicates how many text answers have sentiment. The percentages show what percentage of all answers express each sentiment.
The answer shows the number of answers the sentiment have been extracted from. Note that the answer number is not the same as respondents.
The Sentiment Table is shown by default but can also be added through the context menu of the dash.
Note that the Sentiment Table only shows answers that have been processed by Text Analysis and that it is not the same as respondents’ answers on the survey.
Keyword Cloud
The Keyword Cloud shows all keywords that have been extracted from the answers and how often they were mentioned (frequency). Larger keywords have a higher frequency than smaller words. Hovering a keyword shows information about how many times they were mentioned.
It is possible to filter the report based on keywords. Filters available are include or exclude answers with selected keyword.
You can customize the Keyword Cloud to make it more suitable for your needs.
The Keyword Cloud is shown by default but can also be added through the context menu of the dash.
Note that the Keyword Cloud only shows keywords from answers that have been processed by Text Analysis.
Keyword Chart
Keyword Chart shows all keywords that have been extracted from the answers and mentioned in answers (frequency).
The bubble size shows the frequency, the number of answers a keyword has been mentioned in.
There are no filter options within the Keyword Chart.
You can customize the Keyword Cloud, to make it more suitable for your needs.
The Keyword Chart is shown by default but can also be added through the context menu of the dash.
Note that the Keyword Chart only shows keyword from answers that have been processed by Text Analysis and that it is not the same as respondents’ answers on the survey.
Keyword List
Keyword List shows all keywords that have been extracted from the answers if they have been grouped and their mentioned in answers (frequency).
There are no filter options within the Keyword List.
It is possible to sort keywords alphabetically or sort mentioned in answers by highest or lowest value.
In the Grouped with column, shows words that have been grouped together by lemmatization. Both keywords and key phrases can be grouped.
The Keyword List is shown by default but can also be added through the context menu of the dash.
Note that the Keyword List only shows answers that have been processed by Text Analysis and that it is not the same as respondents’ answers on the survey.
Keyword Table
The Keyword Table shows some basic statistics about how many keywords have been extracted from the answers for a question. You can also the total number of keywords which lets you know how many times they have been mentioned in answers.
The Keyword Table is shown by default but can also be added through the context menu of the dash.
Note that the Keyword Table only shows answers that have been processed by Text Analysis and that it is not the same as respondents’ answers on the survey.
Answer List
The Answer List shows answers and their extracted sentiment and keywords.
There are quick filters that you can use on sentiment to show only selected sentiment. The quick filters work only locally, within the Answer List box.
Available in the Answer List box is also search. You can search for answers or keywords.
The Answer List is shown by default but can also be added through the context menu of the dash.
Note that the Answer List only shows answers that have been processed by Text Analysis and that it is not the same as respondents’ answers on the survey.
Apply Sentiment Breakdown
Using the Sentiment Breakdown options in the reports creates comparison series in your report. This makes it possible to compare the results of respondents that have been e.g., had positive or negative feelings in their text answers.
Apply Sentiment Breakdown is done through the context menu of the dash. It is possible to use the Sentiment Breakdown on multiple questions with sentiment. The questions name is added to the name of each series, to make it clear which question the sentiment where broken down on.
When applying Sentiment Breakdown, the series are change according to the color of the sentiment.
Share result
PDF export
If you want to share the Text Analysis result you can do so by exporting to PDF. To include all Text Analysis related boxed you need to “Include open text answers” in the export settings.
Share Report via Link or Email
If you want to share the Text Analysis result you can do so via Share report. To include all Text analysis related boxed you need to include ”View text answers” in the share settings.
Tips and Tricks
The engine doesn’t read the question when performing analysis. This is good to keep in mind when writing the survey questions. If you for example have a follow up question based on a “negative” score the respondent might write a neutral answer since it’s based on a negative question.
One of the most frequently used questions within Netigate is the Net Promotor Score (NPS/eNPS). Combining NPS with text box questions makes it even more powerful. The Netigate tool allows you to apply breakdown on the NPS, with this you can dig deeper into what the detractors, passives and promotors are thinking and feeling. Making it possible for you to catch areas for improvement too, for example keep a promoter or turn a detractor around.
It’s possible to include a certain keyword in the keyword cloud to see related keywords and draw insights. In the example below you see how the keyword cloud changes after including “management”, meaning the report is now filtered on data containing “management” in free text answers. In the second keyword cloud you can see what respondents are talking about in relation to “management”.
FAQ
Does the insight engine handle synonyms?
No.
How does the insight engine handle misspelled words?
The engine doesn’t do any spellchecks. A misspelled word can become a keyword, but it will not be grouped to correct spelled word.
How does the insight engine handle irony/sarcasm?
No this is not possible at the moment.
How many answers are required for Text analysis to be performed?
There is no limitation on how many answers are needed to perform Text Analysis. However, the more data, the better the insights that can be drawn from the data set. It’s hard for the engine to draw insights from short answers with just a few words.
Skulle du kunna ge förslag på hur vi kan göra denna artikel bättre?