翻訳と辞書 Words near each other ・ Katyń Memorial ・ Katyń Memorial (Jersey City) ・ Katyń Museum, Warsaw ・ Katz ・ KATZ (AM) ・ Katz (surname) ・ Katz and Leavitt Apartment House ・ Katz Broadcasting ・ Katz Castle ・ Katz centrality ・ Katz Editores ・ Katz Group of Companies ・ Katz railway station ・ Katz syndrome ・ Katz v. United States ・ Katz's back-off model ・ Katz's Delicatessen ・ Katz, British Columbia ・ Katza ・ Katzbach (Kraichbach) ・ Katzbach Railway ・ Katzbalger ・ Katze ・ Katze (village) ・ Katze im Sack ・ Katzelmacher ・ Katzelsdorf ・ Katzen ・ Katzen (performer) ・ Katzen Arts Center Dictionary Lists mini英和辞書 mini和英辞書 Webster 1913 Latin-English FOLDOC Wikipedia English ウィキペディア

翻訳と辞書　辞書検索 [ 開発暫定版 ] スポンサード リンク Katz's back-off model ： ウィキペディア英語版 Katz's back-off model Katz back-off is a generative ''n''-gram language model that estimates the conditional probability of a word given its history in the ''n''-gram. It accomplishes this estimation by "backing-off" to models with smaller histories under certain conditions. By doing so, the model with the most reliable information about a given history is used to provide the better results. ==The method== The equation for Katz's back-off model is: 〔Katz, S. M. (1987). Estimation of probabilities from sparse data for the language model component of a speech recogniser. IEEE Transactions on Acoustics, Speech, and Signal Processing, 35(3), 400–401. 〕 :$$ \begin & P_ (w_i \mid w_ \cdots w_) \\() = \cdots w_} \dfracw_)})} & \text C(w_ \cdots w_i) > k \\() \alpha_} P_(w_i \mid w_ \cdots w_) & \text \end \end where : ''C''(''x'') = number of times ''x'' appears in training : ''w''''i'' = ''i''th word in the given context Essentially, this means that if the ''n''-gram has been seen more than ''k'' times in training, the conditional probability of a word given its history is proportional to the maximum likelihood estimate of that ''n''-gram. Otherwise, the conditional probability is equal to the back-off conditional probability of the "(''n'' − 1)-gram". The more difficult part is determining the values for ''k'', ''d'' and ''α''. $k$ is the least important of the parameters. It is usually chosen to be 0. However, empirical testing may find better values for k. $d$ is typically the amount of discounting found by Good–Turing estimation. In other words, if Good–Turing estimates $C$ as $C^$ *, then $d\; =\; \backslash frac$ To compute $\backslash alpha$, it is useful to first define a quantity β, which is the left-over probability mass for the (''n'' − 1)-gram: :$\backslash beta\_\}\; =\; 1\; -\; \backslash sum\_)\; >\; k\; \backslash \}\; \}\; d\_\}\; \backslash frac\; w\_)\})\}$ Then the back-off weight, α, is computed as follows: :$\backslash alpha\_\}\; =\; \backslash frac\}\}\; )\; \backslash leq\; k\; \backslash \}\; \}\; P\_(w\_i\; \backslash mid\; w\_\; \backslash cdots\; w\_)\}$ The above formula only applies if there is data for the "(''n'' − 1)-gram". If not, algorithm skips N-1 entirely and uses the Katz estimate for N-2. (and so on until an N-gram with data is found) 抄文引用元・出典: フリー百科事典『 ウィキペディア（Wikipedia）』 ■ウィキペディアで「Katz's back-off model」の詳細全文を読む スポンサード リンク 翻訳と辞書 : 翻訳のためのインターネットリソース Copyright(C) kotoba.ne.jp 1997-2016. All Rights Reserved.