Automatic classification of speech overlaps: Feature representation and algorithms.

•The annotation of speech overlap categories over a large amount of ecological conversational data.•Investigation of different lexical feature representation (n-grams, word-embeddings).•Study the effect of acoustic and lexical feature combination techniques – feature space and hidden space combination.•Comparative analysis using SVM, Feed-Forward Neural Network, Convolutional network, and Long-Short-Term-Memory.•Modeled the concurrent speech information from both speakers for classification.