In folk song research, string methods have been widely used to retrieve highly similar tunes or to perform tune family classification. In this study, we investigate how various string methods perform on a fundamentally different classification task, which is to classify folk tunes into genres, the genres being the dance types of the tunes. A new data set Dance-9 is therefore introduced. The dif...

The phrase-based and N-gram-based SMT frameworks complement each other. While the former is better able to memorize, the latter provides a more principled model that captures dependencies across phrasal boundaries. Some work has been done to combine insights from these two frameworks. A recent successful attempt showed the advantage of using phrasebased search on top of an N-gram-based model. W...

由於現代科技以及 3C 產品的普及,使得孩子頻繁的接觸電視、網路、手機...等,因此 容易缺乏與人之間互動、溝通以及情感的表達,相對的,學生寫的作文常常是以流水帳 交代經過,有的學校甚至不考作文,但隨著教育政策的變動,國中教育會考加入了作文 評量的項目,使的作文再度受到學生及家長的重視。可是受限於學校教學時數,作文較 弱的學生容易缺少補救的機會。我們認為未來自學作文以及在家練習,可以藉由自動化 的作文教學系統輔助。而本系統開發作文教學系統之句子流暢度偵測,經由系統回饋的 診斷結果可以讓學生對詞句組合的理解力有所提升,幫助學生寫出較流暢的句子,藉此 提高他們的作文分數。系統所依賴的 N-gram 語言模型,它的特性是計算字詞間組合的 機率,機率越高的話字詞組合的正確性越高也就是越流暢,而語言模型效果相當依賴大 型的訓練語料,這是語言模型然能待克服的缺點,例如資料稀疏(Data spar...

We study the renormalization of the O(N) model using the auxiliary field formalism (HubbardStratonovich transformation) in the 1/N expansion at finite temperature. We provide the general strategy of renormalization for arbitrary order, and make calculation up to next-to-leading order. We show that renormalization is possible for any values of the condensates, prove the temperature independence ...

隨著科技的發展,現在 3C 產品可說是非常的普遍,也因為如此現在非常多的孩子從小就 接觸電腦、手機、平板等 3C 產品,使得現在學生更有可能以電腦作為寫作文的工具。雖 然教育政策將作文納入考試評分項目,使得學生跟家長再度重視寫作能力,但是受限於 教學時數,可以練習寫作的時間實在是不足以將那些寫作能力較弱的學生作有效提升。 因此我們認為未來可以藉由自動化的作文教學系統幫助學生在家自學作文。而我們所開 發作文教學之句子流暢度偵測系統,經由系統回傳的診斷結果,幫助學生提升詞句組合 的理解能力以寫出較順暢的句子,藉此提升他們作文的分數。本系統依賴 N-gram 的語言 模型[1],其特色是計算字詞間組合的機率,機率越高字詞組合的正確性就越高句子也就 越順暢,然而語言模型其效果相當依賴大型的訓練語料,這是語言模型仍待克服的問題, 而且如果訓練語料的性質跟要測試的文章性質越不相關,效果就會越差,...

This paper presents a novel approach to low resource language modeling. Here we propose a model for word prediction which is based on multi-variant ngram abstraction with weighted confidence level. We demonstrate a significant improvement in word recall over ”traditional” KneserNey back-off model for most of the examined low resource languages.