Sentiment analysis is one of the natural language processing tasks used to find reviews expressed in online texts and classify them into different classes. One of the most important factors affecting the efficiency of sentiment analysis methods is the aggregation algorithm used for scores combination. Recently, Dempster–Shafer algorithm has been used for scores aggregation. This algorithm has a higher precision than common methods such as average, weighed average, product and voting, but the problem with this algorithm is the aggregation of a dominant high or low score that is always selected by the algorithm as the overall score. In the current research, a new method is proposed for scores aggregation that employs both the most and the second probable classes to predict the final score. The proposed approach considers every review as a set of sentences each of which has its own sentiment orientation and score and computes the probability of belonging of every sentence to different classes in a five-star scale using a pure lexicon-based system. These probabilities are then used for document-level sentiment detection. To this aim, two-point structure is used to improve the Dempster–Shafer aggregation algorithm. The proposed method is applied to review datasets of TripAdvisor and CitySearch which have been used in previous studies. The obtained results show that in comparison with the original Dempster–Shafer aggregation method, the precision of the proposed method for both datasets is 23% and 27% higher, respectively.

Despite the increasing popularity of the stance detection task, existing approaches are predominantly limited to using the textual content of social media posts for the classification, overlooking the social nature of the task. The stance detection task becomes particularly challenging in cross-target classification scenarios, where even in few-shot training settings the model needs to predict the stance towards new targets for which the model has only seen few relevant samples during training. To address the cross-target stance detection in social media by leveraging the social nature of the task, we introduce CT-TN, a novel model that aggregates multimodal embeddings derived from both textual and network features of the data. We conduct experiments in a few-shot cross-target scenario on six different combinations of source-destination target pairs. By comparing CT-TN with state-of-the-art cross-target stance detection models, we demonstrate the effectiveness of our model by achieving average performance improvements ranging from 11% to 21% across different baseline models. Experiments with different numbers of shots show that CT-TN can outperform other models after seeing 300 instances of the destination target. Further, ablation experiments demonstrate the positive contribution of each of the components of CT-TN towards the final performance. We further analyse the network interactions between social media users, which reveal the potential of using social features for crosstarget stance detection.