1. A multi-target stance detection algorithm based on a bidirectional long short-term memory (Bi-LSTM) network with position-weight is proposed to solve the problem of mutual influence of content describing different targets.
2. The position information and output from the Bi-LSTM layer are fused by the position-weight fusion layer, and stances of different targets are predicted using the LSTM network and softmax classification.
3. The results demonstrate that the Bi-LSTM network with position-weight achieves an advantage of 1.4% in macro average F1 value in comparison to recent algorithms.
The article is generally reliable and trustworthy, as it provides evidence for its claims in the form of a multi-target stance detection corpus used to validate the proposed method, which demonstrates an advantage of 1.4% in macro average F1 value compared to recent algorithms. However, there are some potential biases that should be noted. Firstly, there is no mention of any counterarguments or alternative methods that could be used for multi-target stance detection, which could lead to a one-sided view on this topic. Secondly, there is no discussion about possible risks associated with using this method, such as potential errors or inaccuracies due to incorrect data input or other factors. Finally, there is no mention of any ethical considerations related to using this method for multi-target stance detection, such as privacy concerns or potential misuse of data collected through this method.