EthCC’s Post

Looking to empower your RAGs-based LLM Pipeline, how about exploring Knowledge Graphs? 🤔📊 We explore Knowledge Graphs and their ability to empower LLM Pipelines by boosting their contextual understanding. Read more in our publication's latest blog! ✍🏼 🔗: https://lnkd.in/e5gw_6QU #llm #knowledgegraph #opensourcellms #developerpractices

New York University students and alumni can have unlimited access for One.

Another paper recently published and presented at 26th International Conference on Enterprise Information Systems (ICEIS - 2024): "Module of Contrastive Analysis for a Phonological Assessment Software in Development ." We based our study in a Digital Platform that is currently under development and is built on React. The contrastive analysis in speech therapy is basically a comparison between two phonetic transcriptions for a single word: the one we want to analyze and the expected. The speech therapists want to see the exacly differences between both, so they can use it to measure a indication of speech disorder, specially in children. The entire process of contrastive analysis is done manually with paper/pencil by most of them, and they needed to do it for several words. So, in this paper, we proposed a module to generate the entire contrastive analysis data, by processing all the phonetic transcriptions from an assessment when it's finished. We applied the Levenshtein Distance Algorithm combined with a Case-Based Reasoning approach to implement this module in our digital platform. Paper available at ResearchGate: https://lnkd.in/dVifRkrZ João Carlos D. Lima Marcia Keske-Soares https://lnkd.in/dByCWkZj

Big update to our Document Q/A application, finally we've integrated the new PDF viewer (which is, by the way, a Streamlit component, under active development) which allows the visualisation of the retrieved context directly in the PDF document. This makes use of the full potential of Grobid for text extraction. I had just woke up, as you can see from my face 😅 Demo: https://lnkd.in/gZG5BVfv https://lnkd.in/gApKuXiY

📨 Our Spring 2024 United Way Update is available, focused on the ALICE data tool and reports! Dive deeper into the details about the population known as ALICE–an acronym standing for Asset Limited, Income Constrained, Employed–and learn how to put this data into action. Check out the update: https://lnkd.in/ee7nMyd4

It is quite interesting how would new llama 3.2 and 4o-mini approach a classic statistics problem, where incremental reasoning is required. Following the recent post by Simon Villani, PhD, I modified it a bit to avoid model parroting historic discussions: Imagine playing a simple game, where there are five identical doors with prize behind one and noting for the remaining doors. The game hosts who knows where the prize is asks you to pick a door. Next, he asks which door you chose and opens two of the remaining four doors that do not have a prize behind them. Finally, he asks you if you want to change your door selection. Is it in your interest to keep your original selection?

Good to see the Arts Council have released their annual 'Basic Dataset for Libraries' of library locations: https://lnkd.in/edMRf9MU It's very useful, but the dataset is also an illustration of how we need one data source, which is permanently maintained, and used in applications. An annual survey is always out of date. One particular typo has Porlock library written as TA24 7HD, which should be TA24 8NP. An easy mistake, as the next library in the list (Priorswood) is TA2 7HD - so they've been mixed up. The problem is, I corrected that typo in 2016, and have reported it wrong, along with other corrections, in all the iterations since then. But when there's no mechanism to correct or change the data, the previous year gets sent to library services to validate or correct. Which is then an arduous task.

to demonstrate the solution I posted in : https://lnkd.in/dR_qtgEE Screen recording demonstrating local inference with a remote LLM ( llama3-8b-gguf)

Update: Relay will continue to operate from July 1st. It will be moving to a subscription model in the coming weeks but, for now, it's available for everyone to use free of charge and ad-free! https://buff.ly/431OFTC via https://buff.ly/3sqkucb #facts #truefact #truefacts #telegramfacts #telegram #lessbullshit #fakenews #factcheck #factchecks #truereality FactsTelegramFreaks : Important topics analysed with a scientific approach and facts - Knowledge is the new gold, not bullshit! https://buff.ly/3E4dofW

Great guide put together by CAI on all things reserves. It covers everything from the various state laws, the updated reserve study standards, Reserve Study Best Practices Report, among other topics. I usually try to point out that reserves aren't complicated, but at the same time there are a lot of details and aspects to a thorough reserve study, ensuring regulatory compliance, and leveraging a reserve study to maximize value. If you don't want to read all 80 pages 😲 then reach out to Reserve Advisors, LLC (Reserve Studies, Insurance Appraisals). We're always conducting continuing education courses to help teach to and your team the right steps when it comes to reserves 📊