Chinese-Uyghur Bilingual Lexicon Extraction Based on Weak Supervision

Bilingual lexicon extraction is useful, especially for low-resource languages that can leverage from high-resource languages. The Uyghur language a derivative language, and its resources are scarce noisy. Moreover, it difficult to find bilingual resource utilize the linguistic knowledge of other large languages, such as Chinese or English. There little related research on unsupervised Chinese-Uyghur existing methods mainly focus term based translated parallel corpora. Accordingly, effective, This paper proposes method extract dictionary by combining inter-word relationship matrix mapped neural network cross-language word embedding vector. A seed used weak supervision signal. small data map multilingual vectors into unified vector space. As word-particles these two not well-coordinated, stems main particles. strong semantic associate information. Two retrieval indicators, nearest neighbor cross-domain similarity local scaling, calculate dictionaries. experimental results show accuracy proposed in this improved 65.06%. helps improve machine translation, automatic extraction, translations.