Creating a query
To create a new query based on the new VOC MINE engine, go to Mine -> Queries V2, and click on the Add query button on the page that will open.
In the Add query modal, you will be prompted to select basic parameters for the query:
The available parameters are:
- Query name - name for your query;
- Source - data source that should be used to build query. Uploads, Surveys, and Virtual source can be equally selected for building a query.
- Text column - column of [TEXT] type that will be used for mining. If the selected data source has several TEXT columns, all of them can be used at the same time.
- Language - Language of text cases. In case the data source has text cases in several languages, language can be used as a filter in the query.
- Query Tag - tag for the query. Query tags is the new iteration of QueryLabels: they allow marking different filtered queries and visualizing their result as a virtual Tag column in VISUAL.
After the query is created, it contains all text cases from the selected [TEXT] columns. To narrow the query, query filters are used. To add a filter, please click on the "Add filters" as shown below:
Available filter types are:
- Classification results - allows filtering by classification result if the selected data source is applied to a classifier;
- Filter by date - allows filtering by date. By default, "Captured at"/"Questionnaire end time" columns are used. If your data source contains other [DATE] columns, it is possible to use them to filter text cases by
- Lemmata to include - after applying the filter you will see text cases which contain the selected lemmata.
- Lemmata to exclude - after applying the filter text cases which contain the selected lemmata will be excluded.
- Text include - after applying the filter you will see text cases which contain the selected texts.
- Text to exclude - after applying the filter text cases which contain the selected texts will be excluded.
- Filter by attribute - allows filtering the selection by metadata values.
- Advanced query - for more advanced filtering it is possible to use a more complex metadata filter in the same way as in Recoded Variables.
Each filter (except Filter by date) allows joining its condition with AND/OR operators.
Query action output
The right half of the new queries interface contains various views on the query's action output:
- Cloud - shows a number of most frequently appearing lemmata. In this view, it is possible to add lemmas to the Lemmata to include or Lemmata to exclude filters by clicking on the needed lemma.
- Text - shows every text case that appears in the query separately. In this view, it is possible to see text case's lemmata, associated metadata, and other details, such as classification results.
- Chart - shows the list of most frequently used lemmata along with a frequency of appearing in the resulting query. It is also possible to add any of the listed lemmata to the Lemmata to include or Lemmata to exclude filters by clicking on Exclude or Include buttons against the needed lemma.
- Co-occurence - shows the number of times words occurred together in one text.
- Correlation - shows how much more likely a pair of words will occur together in a sentence than separately
Positive correlation - there are more cases when words either appear together in one sentence or do not appear at all.
Negative correlation - there are more cases that they will appear separately than together or not at all.
In MINE, word pairs with positive correlation are displayed only.
Applying dictionaries to a query
To add a dictionary to a query click on the Dictionaries button.
In the modal window, you will be able to select a dictionary to be applied to a query and how this dictionary should be used: as a Stop Word or Synonym. It is possible to select multiple dictionaries at the same time.
Please sign in to leave a comment.