JinKaidou/Team-Cebuano-TD-GAT-Attention-Over-Attention

Domaine:

natural language processing

Type de record:

model

Créateur:

Jin

Hôte:

Enhancing low-resource sentiment analysis: A syntactically-aware and noise-robust for low-resource language model for the cebuano ## Bisakol Sentiment Analyzer (Cebuano/Bisaya) **Bisakol Sentiment Analyzer** is a Cebuano/Bisaya sentiment analysis project powered by **DistilBERT-Cebuano + Graph Attention Network (GAT) + Attention-over-Attention (AoA)**. It includes **Bisaya-only datasets**, **LICR-based preprocessing**, multiple **model experiments**, and a working **prototype web app** for single-text, bulk-text, and file-based sentiment analysis. --- ## Project Structure - **`Datasets/`** - Contains the **Bisaya-only text datasets** used in this project (LICR datasets). - Note: Large dataset JSON files may be excluded from GitHub due to GitHub’s 100MB file limit. - **`Data Preprocessing/`** - Notebooks/scripts for preprocessing the LICR dataset (cleaning, normalization, formatting for training/evaluation). - **`Models/`** - Model training and experimentation notebooks for the approaches used in this project. - **`Bisakol_Sentiment_Analysis_Prototype/`** - Prototype application for running sentiment predictions via UI. --- ## Datasets (Purely Bisaya Text) This project uses datasets that contain **purely Bisaya/Cebuano text** (e.g., Cebuano comments) with sentiment labels. - **LICR datasets** - `LICR_Bisakol_Dataset.json` - `LICR_Cebert_Dataset.json` > If the dataset files are not present in the GitHub repo, they were likely kept locally or shared separately because of GitHub file size limits. to view the file please proceed to this link = drive.google.com --- ## Data Preprocessing (LICR) Preprocessing follows a LICR-focused pipeline implemented in the notebook(s) under: - `Data Preprocessing/` (e.g., `licr-data-preprocessing.ipynb`) Typical steps included: - Text cleaning (noise removal, spacing/punctuation handling) - Normalization for Cebuano/Bisaya text - Formatting into train/validation/test-ready structure - Exporting processed outputs for model training and prototype inference --- ## Models Used We exp …

Visit

github.com

Similaires

Graph Attention Networks (GAT) Enhance Epilepsy Detection from EEG Using Accessible Hardware in Low-Resource SettingsContextual Graph Attention for Answering Logical Queries over Incomplete Knowledge GraphsA Quantum-assisted Attention U-Net for Building Segmentation over Tunis using Sentinel-1 DataInvestor Attention and Stock Market PerformanceSelf-Attention with Structural Position Representationsnileshbansal/Neural-Machine-Translation-with-Attention

Graph Attention Networks (GAT) Enhance Epilepsy Detection from EEG Using Accessible Hardware in Low-Resource Settings

Poster summarizing Mazurek, Szymon, Stephen Moore, and Alessandro Crimi. "Graph Attention Networks

Contextual Graph Attention for Answering Logical Queries over Incomplete Knowledge Graphs

Recently, several studies have explored methods for using KG embedding to answer logical queries. Th

A Quantum-assisted Attention U-Net for Building Segmentation over Tunis using Sentinel-1 Data

Building segmentation in urban areas is essential in fields such as urban planning, disaster respons

Investor Attention and Stock Market Performance

This data is strictly related to the Ghanaian economy from September 2005 to December 2020. Monthly

Self-Attention with Structural Position Representations

Although self-attention networks (SANs) have advanced the state-of-the-art on various NLP tasks, one

nileshbansal/Neural-Machine-Translation-with-Attention

NMT with Attention to translate French into Fongbe and Ewe (two African languages - collectively cal