Great benchmark! It highlights an important but often downstream problem. In real-world pipelines, the bigger issue comes before this: extracting tables from PDFs or scans without breaking their layout. Once the structure is lost (merged headers, nested cells, footnotes, etc.), no data format can fully recover it.
Check out LLMWhisperer from Unstract —> it preserves table and layout fidelity when converting documents for LLM use. You can try it on complex PDFs or forms here: https://pg.llmwhisperer.unstract.com (no signup needed)
Layout preservation upstream often improves downstream accuracy more than choosing between CSV, JSON, or Markdown. Find more details here: https://unstract.com/llmwhisperer/
A friend of mine runs Code Wizards HQ, a startup that teaches coding to kids through teacher-led online classes. Check out their website: http://www.codewizardshq.com/
@danso: Yes, that is the bio search mode. It is simply a bio search sorted by the number of followers. When you switch to influencer search mode, you see results sorted by shares of the articles related to the keyword you searched for. We actually look for keywords in articles they shared by going through their content with NLP.
The free account allows a lot more searches. Do try.
@sbuccini This is the exact problem we have solved with our "influencer search" mode. There, we show you influencers based on keywords in content they've shared. :) Not based on their bio text. Though to do, but solves the exact problem you're talking about!
You can achieve this for a few accounts with the Twitter API, but due to API limits, it is tough to get it done at scale and make a product out of it. There are creative ways of achieving this, though.
Point in case: take a look at how Nuzzle works for Twitter accounts. It is very close to what you are asking for.
About filtering by location, Twitter allows users to fill free text in the LOCATION field. Hence the disparity. In your example, Professor Louis Appleby has set location as "Stockport station". :)
Naren, co-founder of Tweeple Search here. Here are a few points:
1.The Twitter API is not a practical tool you can use to search for tens of thousands of people. Think rate limits.
2. Results there are not sorted by the number of followers (which is decent marker for influence)
3. Check out Tweeple Search's filters, they are very powerful
4. We can give you a list of influencers based on keywords in content they shared as opposed to keywords in their bios (very tough to do this!)
5. Search is only the beginning. Check out our analysis tools
I'm one of the co-founders. There is no disagreeing that the search results page is structurally similar to that of Buzzsumo, which is a very good product. We wouldn't have bothered to develop Social Animal if it was going to be just a Buzzsumo clone. We process every article that enters the system through our Natural Language Processing (NLP) engine which figure out the primary and secondary keywords. These keywords are then displayed for every article you see. With this, things get way more interesting:
- You can click on any keyword to display more top performing articles for that keyword
- Just because 2 articles on the same topic perform well, it doesn't mean they are the same. You can compare any 2 articles to see their unique and common keywords. This saves you time, since you almost never have to read those articles.
- Headline analyzer is built in, which you can invoke for any article
- The insights summary feature not only tells you the best time to publish an article for any keyword, it also tells you which title and article lengths work best. There is detailed analysis available as well.
- If you setup keywords, Social Animal will proactively send you every day:
- Top performing articles for keywords you care about
- Competitor mentions
- Potential, new influencers
- You don't have to manually search every day, saving you a ton of time
- You have to use our Influencer search tool to see how good it is :) It is based on a much better algorithm.
- The InspireMe module lets you explore various, related topics for you to get writing inspiration. It's a lot of fun to use with a nice UI.
- The InspireMe module also tells you, for any given keyword, which secondary topics articles are written about so you can find your niche.
Check out LLMWhisperer from Unstract —> it preserves table and layout fidelity when converting documents for LLM use. You can try it on complex PDFs or forms here: https://pg.llmwhisperer.unstract.com (no signup needed)
Layout preservation upstream often improves downstream accuracy more than choosing between CSV, JSON, or Markdown. Find more details here: https://unstract.com/llmwhisperer/