24小时热门版块排行榜

返回列表

当前只显示满足指定条件的回帖，点击这里查看本话题的所有回帖

mpustc

银虫 (初入文坛)

应助: 0 (幼儿园)
金币: 284.3
帖子: 45
在线: 15.1小时
虫号: 778579
注册: 2009-05-24
专业: 信息安全

[交流] 向细心的2010Information Hiding Conference匿名评审人致敬已有12人参与

看看人家的评审意见多么的细心（仅一条），只要对着一条条的改正就可以，人家不是指出来错误，而且是给出来一个更好的表述方式，开始感觉这么多错误，怎么改啊，因为鄙人英语很不好，原来人家给出的是正确的表述方式。

我们很多人也都是论文审稿人，我们什么时候向人家这么仔细。无论做学问还是做事情，怕就怕在较真上。美国能在科学技术上有今天的成就，我想和做事的态度也很有关系。相反，我们有些时候是不是有些太浮躁了，感觉文章写好了，多检查一遍有时都嫌太耽误时间。多向人家学习学习做事做研究的态度吧！！

4) The writing still needs quite some work.  While it does not hinder understanding,
   the writing does provide a distraction.  I'm just going to point out some
   high-level points:

   a) Avoid contractions (don't => do not, cannot => cannot)
   b) Avoid "very" (for example, in the first sentence of the abstract; "secure"
      here is already a strong statement, and with "very" you're rather contradicting
      the rest of your paper...)
   c) Section X and Table X (capitalize first letter), but do not capitalize
      section or table if not followed by a number
   d) Use the URL package for URLs (~ missing in [17]); check formatting of
      references [3, 5] as well
   e) Be careful with whitespace (use math-mode in 2.2, add a space before '(SVM)'
      at the end of section 4, etc.)
   f) English:
      - Introduction: which uses precomputed; method called NICETEXT;
      classify given text segments; normal texts. This method can accurately detect;
      detection method for NICETEXT, which; to classify text segments;
      may be translated twice to A1 and three times to A2; ... from different
      translators, a high-frequency word in the cover text thus has a
      reasonable chance of becoming two low-frequency words in the stegotext.
      Then new paragraph.  an SVM classifier; A series of experiments is given;
      - TBS: LiT. ; The encoder then selects the sentence; allows the sender to not
      transmit; bits correspond to the;  a hidden message, he first;
      - SATBS: So in a German cover text which; 1000 differnt types of texts;
      from 1 to 15.  The texts; by the Google, Systran and Prompt machine;
      words for each frequency; given texts into stegotexts; we need to expand;
      A word which is translated; translators and always yields the same result;
      a word is a one-to-one; Words which are translated; and generate different
      results are called one-to-many-words; and the stegotext, if we delete;
      stegotext (after deleting...; We find the word; greater than it is in;
      frequency differences between; also 2-gram (two adjacent words) frequency
      differences. Deleting one-to-one 2-grams ...; one-to-one words) from;
      stegotext also expands the; exists significant n-gram frequency;
      - FG: Our detection schema is an instance of two-class pattern recognition.
      A given text; can be formalized to; text, we delete (space!);
      following word, we; into a collection of; from the collection.  the refined
      collection as:; we extracted are defined; word types that appear in
      - Experiments: translators.  The texts; using the LiT prototype;
      (English) and German; from the Europarl; texts from the Europal; sentence.  Every sentence;
      we collected a total of 9784 one-to-one words and a total of 39638;
      text size increases.  when the text size; represents data; not need to be repeated
      in each subsequent; texts. Translated texts; our method is accurate at classifying;
      - Conclusion: we presented; semantically coherent; Also: "normal texts" (normal =
      untranslated? nobody cares! what about machine-translated texts?);
      First sentence after (4) is also irrelevant for your conclusion.

回复此楼