Gaetano Rossiello
@tanoross
Followers
388
Following
2K
Media
17
Statuses
463
Research Scientist @IBMResearch | Natural Language Processing, Knowledge Representation and Reasoning, Machine Learning | Guitarist, Composer
New York, USA
Joined June 2014
Say hello to Text2SQL in watsonx․data intelligence. 👋 Simply describe what data you need in plain English and let Text2SQL generate the SQL query. ⭐ No more IT bottlenecks ⚡ Faster insights 🌐 Democratized data access Check it out ⤵️
4
9
18
#UMAP2025 is just around the corner 😍🗽 Get ready with us and take a look at our conference program 📅 https://t.co/g0Ttwge2pQ (more in the thread 👇) And you, have you reached #NYC?? The organizing chairs are ready to welcome you in the wonderful conference venue 🤗 (1/5)
1
5
10
#Reasoning and #Planning for #LLMs @ #ICLR2025 is happening now. If you're interested in enhancing datasets with parsable chain-of-thoughts via knowledge distillation, feel free to stop by! Paper: https://t.co/Pb6WC2NgZb
0
0
1
IBM's updated Granite 3.1 models are here! 4 sizes using two different architectures: Mixture of expert models are designed for low latency usage. ollama run granite3-moe:1b ollama run granite3-moe:3b Dense models are designed for tool-based use cases. ollama run
9
58
346
#DavidLetterman show opening in June 2025 be like: "Frooom Roosvelt Island, the heaaart of #NYC, it's UMAAAP 2025" 🎉🎺🎹 Have a look at the wonderful conference venue: the Verizon Executive Education Center 😍 https://t.co/zJXwsartnd
#UMAP2025 #usermodeling #personalization
1
6
6
We've got a pocket full of dreams. We are proud to announce the Final Call for Full and Short Papers for #UMAP2025, June 2025 in NYC. Please spread the word, as it promises to be a unique edition of our beloved conference. Check out the details here: https://t.co/3y6LXIDhp8
1
7
8
Oktie showing a demo of the work from IBM Research semantic automation team at ISWC 2023! @oktie @IBMResearch @IBMwatsonx @iswc_conf
0
2
10
Oktie presenting our work on semantic automation and conversational GUIs for discovering, enriching, connecting and augmenting data all without writing a single line of code. @oktie @IBMResearch
#iswc2023 @iswc_conf
0
2
20
Introducing IBM watsonx, our new enterprise-ready AI and #data platform designed to multiply the impact of #AI across your business. See what's coming: https://t.co/Bpud6oWavS
17
153
258
This week in 1961, we cut the opening ribbon for our Thomas J. Watson Research Center in New York 🍎 Designed by famed architect Eero Saarinen, the Yorktown Heights facility houses IBM Research, whose labs are developing what’s next in quantum, AI, semiconductors & cloud.
4
20
102
Looking for a new SOTA few-shot classifier? We present QAID! QAID uses batch contrastive learning with BERTScore & classifies by retrieving the intent name. SOTA on few-shot Intent Detection! 🎉 https://t.co/bMxNHwO0gc Accepted to @iclr_conf 2023 🥳 #NLProc #NLP #ML
2
37
84
Zero-shot domain adaptation with SoTa results for Document Retrieval using LLMs like GPT3.5 and FLAN-T5!! Joint work with @stanfordnlp @ChrisGPotts @lateinteraction , others and led by @JonSaadFalcon . Code will be available soon on PrimeQA.
Do you need zero-shot document retrieval in new domains? You're in luck! We propose UDAPDR (“You-dapter”) that efficiently leverages GPT-3.5 and Flan-T5 to address domain adaptation under annotation constraints and domain shifts, improving the SoTA ColBERTv2 by up to 13%.
0
10
45
I'm happy to share our work "KnowGL: Knowledge Generation and Linking from Text" presented at the #AAAI23 demo session. - Paper: https://t.co/9KdX0yTDAn - Model: https://t.co/ElhJCJ7LAe - Video: https://t.co/KYaZDzpSX6
1
6
42
"KnowGL: Knowledge Generation and Linking from Text" a tool that allows converting text into structured relational data for given Knowledge Graphs such as @Wikidata. (Rossiello et al, 2022) paper: https://t.co/4JGmeKj6YG model: https://t.co/8AY5SEwkqk.
6
73
290
I don’t train from scratch, I use RoBERTa🧐 Wait… Why not cross-encoder/stsb-roberta?facebook/muppet-roberta? We automatically identify the best models on 🤗(periodically) Just pick the best one and finetune on your task https://t.co/TO92t2hl2a
https://t.co/Y5FKjfEZk1
ibm.github.io
Model-recycling - the best model per architecture. Comparing finetuned models from HF, as base models for future finetuning.
8
53
245
KnowGL: Knowledge Generation and Linking from Text https://t.co/FcZw0PnumF by @tanoross et al. #KnowledgeGraph #Computation
deepai.org
10/25/22 - We propose KnowGL, a tool that allows converting text into structured relational data represented as a set of ABox assertions comp...
0
2
11
The KnowGL model, which generates triples from text, we used in this work is available in @huggingface. https://t.co/TivvqGGq1P
#iswc2022 #NLP #KnowledgeGraphs #SemanticWeb
@tanoross @faisalMChy @gliozzo @IBMResearch
huggingface.co
Nandana Mihindukulasooriya explains why and how to use knowledge graph in a corporate research community use case with the work "Knowledge Graph Induction enabling Recommending and Trend Analysis: A Corporate Research Community Use Case". #iswc_conf #iswc2022
1
1
5