Generating natural language descriptions of an image, namely image captioning, has received much attention in computer vision and natural language processing. Recent image captioning models are mainly based on the encoder-decoder framework in which visual information is extracted by an encoder, e.g. using convolutional neural network (CNN), and captions are generated by a decoder, e.g. using recurrent neural network (RNN). Although this framework is promising for image captioning, there are still issues in the RNN decoder for exploiting the visual information to generate grammatically and semantically correct captions. More specifically, the RNN decoder has limited ability in dealing with long-term complex dependencies, leading to ineffective use of contextual information from the encoded data. To address this issue, in this paper, we introduce a multi-layer gated recurrent unit (ML-GRU) within the conventional RNN decoder, which enables the modulation of the relevant information flow inside the unit, and thus leads to the generation of semantically coherent captions. The proposed ML-GRU-based RNN decoder has been extensively evaluated on the MSCOCO dataset, and experimental results demonstrate the advantage of our proposed approach over the state-of-the-art approaches across multiple performance metrics.

Summary Sentiment analysis is one of the major tasks of natural language processing, in which attitudes, thoughts, opinions, or judgments toward a particular subject has been extracted. Web is an unstructured and rich source of information containing many text documents with opinions and reviews. The recognition of sentiment can be helpful for individual decision makers, business organizations, and governments. In this article, we present a deep learning‐based approach to sentiment analysis on product reviews obtained from Twitter. The presented architecture combines TF‐IDF weighted Glove word embedding with CNN‐LSTM architecture. The CNN‐LSTM architecture consists of five layers, that is, weighted embedding layer, convolution layer (where, 1‐g, 2‐g, and 3‐g convolutions have been employed), max‐pooling layer, followed by LSTM, and dense layer. In the empirical analysis, the predictive performance of different word embedding schemes (ie, word2vec, fastText, GloVe, LDA2vec, and DOC2vec) with several weighting functions (ie, inverse document frequency, TF‐IDF, and smoothed inverse document frequency function) have been evaluated in conjunction with conventional deep neural network architectures. The empirical results indicate that the proposed deep learning architecture outperforms the conventional deep learning methods.