The introduction of Transformer neural networks has changed the landscape of\nNatural Language Processing (NLP) during the last years. So far, none of the\nvisualization systems has yet managed to examine all the facets of the\nTransformers. This gave us the motivation of the current work. We propose a new\nNLP Transformer context-sensitive visualization method that leverages existing\nNLP tools to find the most significant groups of tokens (words) that have the\ngreatest effect on the output, thus preserving some context from the original\ntext. First, we use a sentence-level dependency parser to highlight promising\nword groups. The dependency parser creates a tree of relationships between the\nwords in the sentence. Next, we systematically remove adjacent and non-adjacent\ntuples of \\emph{n} tokens from the input text, producing several new texts with\nthose tokens missing. The resulting texts are then passed to a pre-trained BERT\nmodel. The classification output is compared with that of the full text, and\nthe difference in the activation strength is recorded. The modified texts that\nproduce the largest difference in the target classification output neuron are\nselected, and the combination of removed words are then considered to be the\nmost influential on the model's output. Finally, the most influential word\ncombinations are visualized in a heatmap.\n