The REVIEW Section

The Review section is the last section of the Composer area that follows the process of writing on the NLG Cloud. In this area the focus is on quality assurance, e.g. reviewing the possible outcome and correcting mistakes.

For this you can generate one text a time for all the test objects in the Review tab. Below the generated text appear all relevant errors and warnings.

Clicking on the Regenerate button will render a different outcome of the same data set. This is very useful for quality assurance, since you can review here a lot of different texts in every language you wrote

Test Objects

The test objects you created in the Data Sources area appear in a column on the left panel. You can search for a concrete test object in the search box. By clicking in one object, you see in the middle panel the uploaded data associated to that object – in JSON format – and in the right panel the corresponding generated text. This gives you an overview of all the important elements of text generation.

Result

In the Result tab you get one variant of the text which would correspond to this specific test object and the chosen story. The text can be regenerated to check different variants and even other statements depending on the story.

Debug

The debug tag lists other elements of the NLG that are important for evaluating the role of the data, the correctness of the rulesets, and the language dependencies.

  1. Rendered statements The list of the name and one possible output of the rendered statements. The statements as well as the containers within a statement are marked and linked to the original container in the Write Tab.
  2. Rendered nodes A table shows details to the nodes that are rendered based on the test object. The categories are name of the node, truth value (true or false), the value of the data field that is used in the nodes and the vocabulary used in the statement to express this node.
  3. Missing Lexicon entries A list of the words used in containers that have no lexicon entry.