Multi-Source Summarizer

Description
Given a block of text summarizers (e.g.<a href="https://algorithmia.com/algorithms/nlp/Summarizer"> this one</a>) are able to create a good summary. &#160;However, there are times when a source string is made up of multiple strings on the same topic. &#160;If multiple strings are concatenated together into a single string, the example Summarizer will usually reject or include entire sub-strings. &#160;For example, from 5 strings concatenated into a single block of text, the <a href="https://algorithmia.com/algorithms/nlp/Summarizer">Summarizer</a> will usually pick 2 out of the original 5 as the summary.<div><br/></div><div>Multiple strings on the same topic might contain differing views or additional information so it would be useful to give the algorithm freedom to blend across multiple strings. &#160;The output created is probably less likely to include whole paragraphs verbatim from the source strings.</div><div><br/></div><div>Approaches that came to me (very new to nlp)...</div><div><br/></div><div>Maybe could be achieved by first breaking into sentences, randomizing the order and then summarizing?<br/><br/>To handle differing sentiment perhaps the sentences could be grouped by sentiment, then randomized in groups?</div>
Discussion
  • {{comment.username}}
Status
Active
submission(s) pending review
Bounty expires in
Bounty expired
Bounty
50
Tags
(no tags)