Longseek
AI Tool for Long Source Researchers
Longseek is an experimental LLM AI tool that helps researchers and others conveniently analyse long, unstructured information with stored, re-usable, batches of multiple questions.
-
Scrapes, reads & stores very long information sources – entire large websites & high volumes of documents.
-
Answers questions, accurately, with supporting evidence - citations and excerpts.
Sample output report here.
If you'd like to explore what Longseek can do, please contact us.
Capabilities
Researchers can feed Longseek very large, heteregenous data sources of up to 1.3 million words in length and query them all simultaneously with 50 to 200 questions at a time - stored in a convenient Question File for re-use. Longseek generates a PDF report within a few minutes. Most LLM answer engines can’t do this. Longseek is also unusual in that:
-
It only searches the trusted information sources the user instructs it to – not the whole internet (unless that’s what is required).
-
For each set of questions, Longseek is simultaneously aware of its entire question set and all its very detailed source information. It therefore has a deep ‘understanding’ of the context of the questions. This is not the case for most LLMs which typically process material in much smaller pieces, and/or one question at a time.
-
Its deeper contextual understanding enables Longseek to extract information, answer questions, and support its answers, potentially more accurately.
To see a sample Longseek report, click here.
Use Cases
We believe there are many use-cases for Longseek and welcome opportunities to explore them with new partners. A few examples:
-
Metastudies in healthcare and elsewhere.
-
Pro bono legal work.
-
Research grant applications.
MHF is a nonprofit. We support nonprofits and pro bono work by donating AI services.
We continue to develop Longseek to process larger corpora, with greater accuracy. Longseek is exeprimental. Always check results.
.
Case Study - ZSL
We're excited to be collaborating with ZSL, the international conservation charity. See the announcement here. ZSL’s SPOTT team provide sustainability indicators on commodity companies in the Palm Oil, Timber and Rubber sectors.
Current Workflow
SPOTT produces up to 180 sustainability indicators on over 300 companies. Analysts laboriously extract information from hundreds of websites.
Longseek
Longseek aims to make SPOTT more efficient. For each company, it reads hundreds or thousands of web pages and simultaneously answers hundreds of detailed questions. Draft indicators and reports are created automatically. See here.
Example Question
Does the company's sustainable palm oil policy or commitment apply to all suppliers?
Longseek's Response
Yes.
"This policy applies to all of Goodhope’s operations and subsidiaries, including all our palm oil mills, refineries, and plantations, and includes provisions for our associates, contractors, and third-party suppliers."
Goodhope Sustainability Policy 2022
https://goodhopeholdings.com/wp-content/uploads/Sustainability_Policy_2022.pdf
---
Longseek gives a clear answer, and supports it by quoting text from a source document, providing the document's name, and a citation URL link. See all 50 Q & A's here.
ZSL
Established in 1826, ZSL produces crucial research and information about the natural world, and mankind's impact on it.
SPOTT enables importers, retailers and investors to identify and do business with commodity companies with good ESG practices, thereby putting pressure on other companies to improve their own practices.
Project Status
SPOTT is currently dual-running Longseek against its existing manual processes. If trials demonstrate acceptable accuracy, SPOTT hopes to fully integrate Longseek into its workflow, increasing its analysts’ capacity to monitor more companies, and additional commodity sectors.
