PlagiaCorpus: A Bilingual, Adversarial, and Cross-Lingual Dataset for Forensic Plagiarism Detection

The user wants me to act as a professional news layout engineer and convert a plain text news body into a semantic HTML fragment. Let me analyze the text and apply the rules: 1. Output only HTML fragment, no markdown fences, no explanations 2. Each natural paragraph wrapped in

, long texts use

/

for sections 3. Lists use

• , quotes use

  , code uses

  , important sentences can use 
  4. Preserve all original information, no additions, deletions, or modifications
  5. No document-level tags like //
  
  The text is about PlagiaCorpus dataset. Let me identify the structure:
  - First paragraph introduces what PlagiaCorpus is
  - Second part talks about the dataset contents (198,643 records, source corpus, queries)
  - Third part lists the five reuse classes
  - Fourth part about linking and train/test split
  - Fifth part about curation process
  - Sixth part about technical validation
  - Seventh part about distribution formats
  
  Let me structure this:
  - The intro can be a paragraph with  for key terms
  - The dataset contents section could be a paragraph
  - The reuse classes list could be a 
  
  - The curation process could use  for key terms
  - Distribution info as a paragraph
  
  Let me draft the HTML:
  
  PlagiaCorpus is a bilingual Arabic-English dataset for benchmarking plagiarism and text-reuse detection, including reuse produced by large language models (LLMs) and reuse across languages.
  
  It contains 198,643 text records: a source corpus of 34,885 passages drawn from English and Arabic Wikipedia and two openly licensed book collections, and 163,758 queries (147,381 for training and 16,377 for testing).
  
  The queries cover five reuse classes of increasing difficulty:
  
             
  verbatim copying
             
  character-level obfuscation
             
  same-language paraphrasing by two different LLMs
             
  English–Arabic translation with structural paraphrasing
  
  Wait, the text says "five reuse classes" but then lists: verbatim copying, character-level obfuscation, same-language paraphrasing by two different LLMs, and English–Arabic translation with structural paraphrasing — that's four listed plus a Safe class. Let me re-read.
  
  "The queries cover five reuse classes of increasing difficulty, namely verbatim copying, character-level obfuscation, same-language paraphrasing by two different LLMs, and English–Arabic translation with structural paraphrasing, together with a Safe class of non-reused passages."
  
  So it says "five reuse classes" but lists four, then adds a Safe class. Actually reading more carefully - the text says "five reuse classes" and lists four items, then says "together with a Safe class". This seems like the text itself might be slightly inconsistent (says five but lists four + Safe = five total, or four reuse + Safe). I should preserve the original text exactly. So I should keep the wording as-is.
  
  Let me convert: "The queries cover five reuse classes of increasing difficulty, namely verbatim copying, character-level obfuscation, same-language paraphrasing by two different LLMs, and English–Arabic translation with structural paraphrasing, together with a Safe class of non-reused passages."
  
  I could make this a list but need to preserve the exact wording. Let me think... I could do:
  
  The queries cover five reuse classes of increasing difficulty, namely:
  
             
  verbatim copying
             
  character-level obfuscation
             
  same-language paraphrasing by two different LLMs
             
  English–Arabic translation with structural paraphrasing
  together with a Safe class of non-reused passages.
  
  Hmm, but that changes the sentence structure. The rule says "不得增删改内容" (no additions, deletions, modifications). Splitting a sentence into a list might be considered modifying. Let me be careful.
  
  Actually, converting inline enumerations to lists is a standard HTML semantic conversion and doesn't change content. But to be safe, I could keep it as a paragraph. Let me think about what's best.
  
  Actually, the rule says lists should use 
  
             
  . The text has "namely X, Y, Z" which is an enumeration. Converting to a list is a reasonable semantic HTML transformation. But the "together with a Safe class" part is a separate clause. Let me keep it as a paragraph to preserve exact wording, or split carefully. I think the safest approach that still demonstrates good HTML structure: 
              
  The queries cover five reuse classes of increasing difficulty, namely verbatim copying, character-level obfuscation, same-language paraphrasing by two different LLMs, and English–Arabic translation with structural paraphrasing, together with a Safe class of non-reused passages.
   Or I could use a list for the enumeration. Let me go with keeping it mostly as paragraphs but using  for key terms, and perhaps a list for the reuse classes since it's clearly an enumeration. Let me reconsider. The instruction says "列表内容用 
               
  
                
  " - list content should use ul/li. If the content is naturally list-like, I should use a list. The reuse classes are naturally list-like. I'll do: 
                 
  The queries cover five reuse classes of increasing difficulty, namely:
                 
  
                  
  Verbatim copying
                  
  Character-level obfuscation
                  
  Same-language paraphrasing by two different LLMs
                  
  English–Arabic translation with structural paraphrasing
                 
                 
  together with a Safe class of non-reused passages.
   Hmm, but this splits the sentence. Let me think about whether this violates "保留原文所有信息，不得增删改内容" (preserve all original information, no additions, deletions, modifications). Converting a sentence with enumerated items into a list + paragraphs is a presentation/semantic change, not a content change. All words are preserved. I think this is acceptable and actually demonstrates good HTML skills. But to be really safe, I could keep the whole thing as one paragraph. Let me go with a hybrid - keep it as a paragraph but it's fine. Actually, let me reconsider the overall structure. The text is fairly short (one long paragraph in the original). I need to break it into logical sections. Let me identify natural breaks: 1. What is PlagiaCorpus 2. Dataset contents (records, corpus, queries) 3. Reuse classes 4. Linking and train/test split 5. Curation process 6. Technical validation 7. Distribution I could use 
                 
  for some of these if it makes sense, but the text is short enough that paragraphs with  might be sufficient. The rule says "长文用 
                   
  /
                   
  划分小节" (long texts use h2/h3 to divide sections). This is a medium-length text, so I could use h3 for a few sections. Let me draft:
                   
  Overview
                   
  PlagiaCorpus is a bilingual Arabic-English dataset for benchmarking plagiarism and text-reuse detection, including reuse produced by large language models (LLMs) and reuse across languages.
   Hmm, adding "Overview" as a heading would be adding content. The rule says no additions. So I should only use headings if the text already has them. Since the original text has no headings, I should not add them. Let me just use paragraphs and lists. Wait, re-reading: "长文用 
                   
  /
                   
  划分小节" - this says long texts should use h2/h3 to divide sections. It doesn't say I must add headings if there