"High bias" represents that the model cannot capture the underlying the pattern of the data (i.e. underfitting).
Since the model is underfitting to even the training data, the difference between the score for the training data and that of test data is small.
In case of high bias, increasing the number of features and changing the algorithm would help to improve the performance.

"Hibh variance" represents that tha model is overfitting to the traning data.
Therefore the difference between the score for the training data and that of test data is large.
In case of high variance, getting more data and feature selection would help to improve the performance.