Question

I wrote a program in C# to calculate TF-IDF to rank documents.

I used the following XML to store the word frequencies within documents. I was criticised heavily for using this structure. Even though I use the text of the word within the Tag, as per me its efficient and consumes less space. Also, I can make a search using XDocument pretty easily since its a nice tree structure. Can you help me understand why was I criticised heavily?

Criticism: How can you add information within meta-data? (For me its innovative).

<word>
   <siddhartha>
      <doc1> 4 </doc4>
      <doc2> 5 </doc2>

   <insipration>
      <doc1> 4 </doc1>
      <doc6> 5 </doc6>

   ....
</word>

I was suggested something like this:

   <word>
   <text> siddhartha </text>
   <doc1> 4 </doc1>
   <text> inspiration </text>
   <doc1> 4 </doc1>
   ...
   </word>

Answer 1

Your structure, with word name as node, will be hard to parse with generic parsers. There is no defined structure: you need to read the whole document to know it.

I may have done something like this (I tried to stay closed to your idea):

<words>
   <word id="siddhartha">
      <freq id="doc1"> 4 </freq>
      <freq id="doc2"> 5 </freq>
   </word>
   ....
</words>