A Cross-lingual Sentence Similarity Calculation Method with Multifeature Fusion

Cross-language sentence similarity computation is among the focuses of research in natural language processing (NLP). At present, some researchers have introduced fine-grained word and character features to help models understand meanings, but they do not consider coarse-grained prior knowledge at level. Even if two cross-linguistic pairs same meaning, representations extracted by baseline approach may language-specific biases. Considering above problems, this paper, we construct a Chinese–Uyghur cross-lingual dataset propose method compute fusing multiple features. The based on pretraining model XLM-RoBERTa assists calculation introducing features, i.e., sentiment length time, eliminate possible biases vectors, whitened vectors different languages ensure that were all represented under standard orthogonal basis. combination has effects final performance model, introduce vector for comparison experiments basic feature splicing method. results show absolute value difference between can reflect sentences well. F1 our reaches 98.97%, which 19.81% higher than baseline.