Methods of Implicit Aspect Detection in Russian Publicistic Texts

This paper compares the performance of various methods of automatic implicit aspect detection in publicistic texts in Russian. The task of implicit aspect detection is an auxiliary task in the aspect-oriented sentiment analysis. The experiments are conducted on a corpus of sentences extracted from political campaign materials. The best results, with the F1-measure reaching 0.84, are obtained using the Navec embeddings and classifiers based on the support vector machine method. Fairly strong results, with the F1-measure reaching 0.77, are obtained using the bag-of-words model and the naive Bayesian classifier. The other methods have a lower performance. It is also found during the experiments that the detection quality can differ significantly between aspects. The detection quality is the highest for the aspects associated with characteristic marker words, for example, healthcare and holding elections. More general aspects, such as quality of governance, are the most difficult to identify.