Text analysis can be used to gain insights from large amounts of text or to find texts related to relevant topics. Quicksearch supports text analysis in most languages, with particularly strong support for Swedish and English. There is also support for custom external models or third-party providers such as Gavagai, IBM, Google, and Microsoft Azure.
Activate Text Analysis
Quicksearch's built-in text analysis is included as part of the service and can be activated by clicking "Edit" on a data source in Analytics and selecting "Enable" and "Quicksearch Text Analysis." If you wish to use other text analysis sources, the process is the same, but you choose the source you want to use, which requires you to have a service with that provider.
Brief Overview of Text Analysis in Quicksearch
Text analysis can be performed on all texts using the Analytics tool. Each result in Analytics is associated with a base language, with the most common being English or Swedish, but there is support for approximately 50 other languages.
| Swedish | The text is analyzed in Swedish | Results in Swedish |
| English | The text is analyzed in English | Results in English |
| Other Languages | The text is translated into English and then analyzed in English | Results in English |
To Achieve Good Results in Text Analysis
It is not enough for the text to be analyzed in a certain language—different contexts may require the analysis to learn things about the context. For example, ignoring certain words that have no value in your context. These are called Stop Words, and they can be specified per question.
Some words may need to be combined for the text analysis to understand them. These are called Synonyms, and you specify which word should be used instead of its synonyms. For example, "Doctor" may be the same as "Doctors," "Physician," "Dr.," etc., for those who respond.
Quicksearch provides templates for various common topics to facilitate setting synonyms and stop words.
Finding Stop Words and Synonyms
One way to find common stop words and synonyms is through word clouds. Word clouds can show which words are frequently used, and it is usually easy for a person to identify words they want to remove (Stop Words) and words that mean the same thing or are similar enough in context (Synonyms).
Step 1. Activate the Analysis Tools on the Question
To activate synonyms, stop words, and word clouds on a question, click the pencil icon above the text responses. There is an option to apply pre-made templates for stop words and synonyms. When you save the question, you can view and analyze the responses through the word cloud.
Step 2. Review and Fill in Stop Words and Synonyms
When you see the word cloud, it is easy to find words that are synonyms or stop words that should not be part of the analysis. Click the pencil icon again to fill in the synonyms and stop words you have identified. Repeat steps 1 & 2 until the desired result is achieved.
Step 3. Create Text Analysis Question
Quicksearch handles text analysis as a new question. This provides many advantages in analysis work because common question types like scale questions can be used in combination with the results of text analysis.
Go to the question list under "Questions" and click on "Add New Question -> Text Analysis Question."
To create a new text analysis question, specify which free text question will be the source for the analysis and then the type of analysis to be performed.
| Sentiment/Tone | Analyze the tone of the text | Scale question with a scale of 0 to 10 where 0 is most negative and 10 is most positive |
| Topic/Subject | Identify the subjects in the text | Multiple-choice question where the subjects identified in all texts are options and the subjects identified in a specific response are selected |
| Language Identification | Determine which language was used | Single-choice question indicating which language was identified by the system. |
| Additional Tone | Some analysis sources (e.g., Gavagai) can analyze more dimensions of a text than just positive/negative. They can also interpret anger, sadness, happiness, etc. | Scale question with a scale of 0 to 10 where 0 is the weakest option on the dimension and 10 is the strongest option on the dimension. |