Hey everyone, I'm trying to use Scholarcy to summarize academic papers for my literature review.
I noticed the summary always includes the whole reference list at the end, which massively inflates the word count. My target is a concise summary of just the findings and methodology.
Is there a setting or a trick to exclude the bibliography from the summary output? I've looked through the dashboard but couldn't find anything obvious.
For example, I input a 10-page PDF, and the summary is great, but then it adds 2 pages of references, making it hard to track my actual summary word count.
Yeah, I've definitely run into this same issue when summarizing for team research briefs. There isn't a direct toggle in Scholarcy's settings for that, which is a bit frustrating.
What worked for me was a bit of a workaround: before you paste the text or upload the PDF, try manually deleting the reference section from the original document first. I know, it's an extra step, but I found Scholarcy is less likely to pull it back in if it's not there to begin with. For a 10-page paper, you can usually scroll to the end and cut everything from "References" onward pretty quickly.
Another thing to check is your export format. I noticed the word count bloat was worse in the plain text export, but a bit cleaner when I copied directly from the summary cards on their web interface. Have you tried comparing outputs that way?
Pipeline is king.
Try using the "Highlights and Abstract" export option. It usually strips out the reference list. The full summary export is the one that adds it back.
If that fails, you could set up a simple script. I run one to clean outputs. It looks for section headers like "REFERENCES" or "BIBLIOGRAPHY" and deletes everything from there to the end.
Example in Python:
```
import re
def strip_references(text):
match = re.search(r'(?i)^(references|bibliography)', text, flags=re.MULTILINE)
return text[:match.start()] if match else text
```
Otherwise, you're stuck with the manual method user742 mentioned.
Benchmarks don't lie.