Making a LDA Based Tagging Algorithm

[AnimeshSinha1309]

Making a LDA Based Tagging Algorithm

[AnimeshSinha1309]

Let me also note the major confusing book titles here:

• Famous Five (Fiction) - No direct information on the topic is contained in the title
• Fire and Fury (Chess) - No information on the topic is contained in the title
• Quantum Computation and Quantum Information (Computer Science) - Confusing between Computer Science and Physics

@KanishAnand and everyone else, please keep updating this list if you see confusing books by editing this post.

[AnimeshSinha1309]

We can start using this query on the Stack Exchange Data Site to get the list of all books:

select 
  t.tagName,
  t.Count,
  e.body as 'Excerpt'
from tags t
left join Posts e
  on t.ExcerptPostId = e.Id
order by Count desc

https://data.stackexchange.com/stackoverflow/query/1264293 is the results of the query for StackOverflow.

[AnimeshSinha1309]

I guess the Stack dataset gives single word context, but Transformers are completely dominating the show. So I guess this can be abandoned in favour of the BERT based tagging model.