Looking at Understanding Convolutional Neural Networks for NLP, Convolutional Neural Networks (CNNs) seem to be suitable not only for image recognition, but also for NLP.
Are CNNs in general the best choice for NLP text analysis, e.g. for sentiment analysis?
If not, is there an overview or comparison of different algorithms with regards to recognition performance (F1 score or something similar)?
The previously selected good answer is true in the sense that CNN and RNN where the bests choices the last few years for NLP (combined with unsupervised methods like word2vec, glove or wordpiece). But recent works use the attention neural network called the Transformer. See Attention Is All You Need and BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding. These model now achieves state of the art performance in many NLP tasks.
I don't have any reference right now, but you should find a lot of papers about it on Google Scholar. For all I know, CNN and RNN (mostly LSTMs) are the two best types of models for sentiment analysis. You will even find some CNN+LSTM architecture that work really well on this kind of tasks.
