# Data Management In-Depth

# Error handling

Before moving on to the model building we must make sure that our database doesn't contain any errors that might lead to unexpected results or failed model training.

The error information is displayed at the bottom of the project view.

An example of a database error.

If we have errors in our database, an easy way to locate them is to click on the "Status" heading in the "Segments" tab. This sorts the segment table by status (ok/warning/error). The red cross sign means that the segment is not correct and you get a detailed error message by hovering the mouse pointer over the symbol.

If right clicking on the whole line of the segment, we can "Open Session"/"Open Directory"/"Copy path" to locate the segment and fix the error.

# Errors

Errors must be fixed, otherwise model training will fail.

Example of an error. The selected label track contains overlapping labels for the file button.label.

# Warnings

Warnings are problems in the data that might affect the result of your training. Warnings should be fixed, or looked through carefully to make sure that they will not affect the training in an unexpected way.

Example of a warning. The selected label track is empty for the file corrected.label. Having data segments with empty label tracks is sometimes desired, for example these segments might contain data to which the model should not react.