The text analysis functionality makes it possible to gain insight into and get a grip on given open answers. Via a manual action (read: classification) you can manage sentiment yourself and link sentiments to given open answers. This functionality is available to everyone and is included in the basic package.
Op deze pagina:
Soon we will launch the 'Artificial Intelligence (AI) add-on' to this functionality. The AI add-on makes it possible to perform automated sentiment analyzes via Machine Learning. Manual classification is therefore no longer necessary. However, manual classification remains available. It is also possible to make corrections to the received Machine Learning results. The AI add-on is ideally suited for investigations where manual work takes too much time and/or effort. We will inform you as soon as the AI add-on is available.
In order to have open answers shown on the Text analysis page, a study must be selected in a first step. This can be done in two ways.
The regular route is to select an investigation within the main menu (click on 'Manage' > 'Home'). You can then click through to the main menu item 'Text analysis' > 'Text analysis'. The survey is immediately loaded in the dropdown at the top left 'Study'. Then select the question for which you want to see the open answers. You do this via the drop-down 'Question'. After selecting the desired question, the relevant open answers are presented in the right column per row.
With the alternative route you go directly via the main menu to 'Text analysis' > 'Text analysis' and select the relevant study and the desired question there as described above.
To further segment the open answers presented, three filters have been made available. Below is an explanation.
- The first filter called 'Filter' is the global filter that is widely available across the entire platform. As a result, the open answers shown can be filtered on, for example: data enrichment, variable, group, date, profile variable or language.
- With the second filter called 'Filter category' it is possible to show given open answers that have been given a certain 'text category'. Read more about manually categorizing open answers here.
- With the third filter called 'Filter sentiment' it is possible to show given open answers that have been given a certain sentiment. The select filter options are: all sentiments, positive, neutral, negative, undefined.
A sentiment results overview has been placed above the right-hand column to provide a clear and effective insight at a glance regarding the total number of data per sentiment and the mutual relationship to each other via the percentage stated in brackets. The overview of results is particularly useful for studies where many open answers have been given to questions.
Manually add sentiment
When you are on the text analysis page and you want to link a new sentiment to an open answer, click on the pencil icon on the right side of the relevant open answer. Then the three 'face icons' below are shown.
- A positive sentiment is shown by means of a smiling face
- A neutral sentiment is shown by means of a neutral face
- A negative sentiment is shown by means of a sad looking face
Then click on the desired sentiment to link that sentiment to the relevant open answer. The selected sentiment is automatically saved. The sentiment can be changed at any time by clicking the pencil icon again and selecting a different sentiment.
Correct the AI generated sentiment input via manual sentiment input
Note: The AI add-on is currently not available yet. We will inform you when the AI add-on is available.
If you have used the AI sentiment analysis and you think it is incorrect, it is possible to manually correct the sentiment generated by the AI - note: this does not improve the underlying AI model (read: trained) . You correct the AI generated sentiment by clicking on the brush icon. Then the three 'face icons' above are shown. You can then select one of the 'face icons' to link another sentiment. If you want to reset the manual sentiment input to the AI generated sentiment, you can do this by clicking on the brush icon again after performing the manual sentiment input and selecting the 'face icon' under which the text 'Choice of AI'.
It is important to know that the manually entered correction is also implemented in the underlying database and the related result overviews.