New
- Sticky notes in J‑Graph.
- Parameters for the Classic ML classifier.
Docs
- Updated data labeling article.
Sticky notes in J‑Graph
Now you can add colored sticky notes with comments to the canvas in J‑Graph.
This is convenient for collaborating on a project, especially at the design and review stage. For example, a script designer can leave comments for a developer about a particular state. A sticky note will display information about who created it and when, and who made the last change.
Parameters for the Classic ML classifier
In advanced settings of the Classic ML classifier, there are new parameters:
useTfIdf
determines which algorithm to use when vectorizing training phrases:TF-IDF
orCountVectorizer
.noteThis parameter changes the classifier behavior: the weight of the intents will be increased by several tenths. To cancel these changes, set
useTfIdf
totrue
.min_document_frequency
defines the minimum word frequency with which it can occur in training phrases, so that it can be vectorized and classified.
Documentation updates
We have updated the article about data labeling and added a detailed description of labeling methods.