INTERACT ERC

We study interactive machine learning for compositional models of natural language. Led by @ariadnaquattoni at the UPC, supported by @ERC_Research

[INTERACT ERC @interacterc]

🚀 We complement this paper with a set of experiments that test the performance of different LLMs in a variety of learning settings.

[INTERACT ERC @interacterc]

🧠 Our paper ZOGRASCOPE: A New Benchmark for Semantic Parsing over Property Graphs by @francescoczz, Justin Kleindienst, Sofia Marquez Gomez and @AriadnaQuattoni is being presented at EMNLP today!
Check it out here 👉 https://t.co/OlkF4CzwAj
Repo 🔗 https://t.co/0GaeGJbLr0

[INTERACT ERC @interacterc]

🧩 Our benchmark features a diverse set of manually annotated queries with varying complexity.
It’s organized into three partitions: IID, Compositional, and Length-based, to test different generalization abilities.

[INTERACT ERC @interacterc]

📚 Alongside SPOT we also release the data for ZOGRASCOPE, a manually annotated benchmark for the cypher language: https://t.co/0GaeGJbLr0
[5/5]

[INTERACT ERC @interacterc]

📢 Excited to share that our paper "SPOT: Zero-Shot Semantic Parsing Over Property Graphs" has been accepted to Findings of #ACL2025!
Joint work with @francescoczz, Justin Kleindienst, Sofia Márquez Gomez and @AriadnaQuattoni
🧑‍💻 Code: https://t.co/SQ0H8yruuE
[1/5]

[INTERACT ERC @interacterc]

📈 We demonstrate the effectiveness of SPOT on two property graph benchmarks utilizing the Cypher query language. In addition, we show that our approach can also be applied effectively to RDF graphs (with SPARQL)
[4/5]

[INTERACT ERC @interacterc]

👋 Want to hear more? Don't forget to join our presentation ❗️TODAY❗️(17th June 2024) at the 16:00 session of the Efficient/Low-Resources and Evaluation Methods for NLP track #NAACL2024

[INTERACT ERC @interacterc]

📢 Our paper "Leveraging the Structure of Pre-trained Embeddings to Minimize Annotation Effort" has been accepted for publication in #NAACL2024.

[INTERACT ERC @interacterc]

📅 This was a joint work of our members Cesar Gonzalez-Gutierrez, @aprimadh, @francescoczz, and @AriadnaQuattoni, to be presented at Insights Workshop 2024 at #NAACL2024 on 20 Jun.

[INTERACT ERC @interacterc]

📢 Our work "Does Fine-tuning a Classifier Help in Low-budget Scenarios? Not Much" has been accepted for publication in Insights Workshop 2024 (colocated with #NAACL2024).

[INTERACT ERC @interacterc]

🍋️ In low-budget scenarios, fine-tuning might not be worth it. But pre-trained representations can still be useful.

[INTERACT ERC @interacterc]

📅 This was a joint work of our members Cesar Gonzalez-Gutierrez and @AriadnaQuattoni, who will be presenting at #NAACL2024 conference on Monday, 17 Jun. See you there!

[INTERACT ERC @interacterc]

📢 Our paper "Leveraging the Structure of Pre-trained Embeddings to Minimize Annotation Effort" has been accepted for publication in #NAACL2024.

[INTERACT ERC @interacterc]

📊 Our experiments show that the strategy is effective in realistic scenarios, outperforming both label propagation techniques and self-training approaches.