Ighina
PhD Candidate in Speech and Language Processing, passionate about Digital Humanities and building things from scratch.
Queen Mary University of LondonLondon
Pinned Repositories
ARP_Score
Average Relative Proximity metrics and experiments used in the paper "When Cohesion Lies in the Embedding Space: New Framework and Methodologies for Embedding-Based Reference-Free Metrics for Topic Segmentation".
Audio-Topic-Segmentation
Repository for the paper "Exploring pre-trained Audio Neural Representations for Audio Topic Segmentation"
bad-boids
A deliberately badly programmed implementation of Boids for teaching
CERTIFAI
A python implementation of CERTIFAI framework for machine learning models' explainability as discussed in https://www.aies-conference.com/2020/wp-content/papers/099.pdf
DeepTiling
A TextTiling-based algorithm for text segmentation (aka topic segmentation) that uses neural sentence encoders, as well as extractive summarization and semantic search applications built on top of it.
Language-Modelling-with-RNNs
A simple series of programs to train gated recurrent neural networks with PyTorch and generate text based on them.
LatinWSD
Repository for Paper "Language Pivoting from Parallel Corpora for Word Sense Disambiguation of Historical Languages: a Case Study on Latin" presented at COLING-LREC 2024
MultiModalSA
MultiModal Sentiment Analysis architectures for CMU-MOSEI.
NSE-TopicSegmentation
A repository including a variety of neural architectures for supervised topic segmentation
VQ-VAE_Topic
An implementation of the paper [Vector-Quantization-Based Topic Modeling](https://dl.acm.org/doi/10.1145/3450946), providing a series of VQ-VAE models for topic modelling. The model reaches state-of-the-art performance on Ng20 and enables the extraction of dense topic vectors for downstream tasks.
Ighina's Repositories
Ighina/DeepTiling
A TextTiling-based algorithm for text segmentation (aka topic segmentation) that uses neural sentence encoders, as well as extractive summarization and semantic search applications built on top of it.
Ighina/MultiModalSA
MultiModal Sentiment Analysis architectures for CMU-MOSEI.
Ighina/CERTIFAI
A python implementation of CERTIFAI framework for machine learning models' explainability as discussed in https://www.aies-conference.com/2020/wp-content/papers/099.pdf
Ighina/VQ-VAE_Topic
An implementation of the paper [Vector-Quantization-Based Topic Modeling](https://dl.acm.org/doi/10.1145/3450946), providing a series of VQ-VAE models for topic modelling. The model reaches state-of-the-art performance on Ng20 and enables the extraction of dense topic vectors for downstream tasks.
Ighina/NSE-TopicSegmentation
A repository including a variety of neural architectures for supervised topic segmentation
Ighina/Language-Modelling-with-RNNs
A simple series of programs to train gated recurrent neural networks with PyTorch and generate text based on them.
Ighina/LatinWSD
Repository for Paper "Language Pivoting from Parallel Corpora for Word Sense Disambiguation of Historical Languages: a Case Study on Latin" presented at COLING-LREC 2024
Ighina/ARP_Score
Average Relative Proximity metrics and experiments used in the paper "When Cohesion Lies in the Embedding Space: New Framework and Methodologies for Embedding-Based Reference-Free Metrics for Topic Segmentation".
Ighina/Audio-Topic-Segmentation
Repository for the paper "Exploring pre-trained Audio Neural Representations for Audio Topic Segmentation"
Ighina/bad-boids
A deliberately badly programmed implementation of Boids for teaching
Ighina/Coursera_Capstone
Repository for the data science specialisation by IBM on Coursera
Ighina/demorepo
It's a demo
Ighina/DigitRecogniser
A very, very basic digit recogniser and gaussian calculators functions with basic Python
Ighina/FrequencyApp
Shiny app to discover and visualise the occurrences of words and/or word-sets (i.e. dictionaries) in given txt files (up to 5)
Ighina/git-is-great
RSE git Module
Ighina/git-is-great-1
RSE Git Module
Ighina/latin-bert-ise-wsd
Using Latin BERT for large scale word sense disambiguation on ISE corpus
Ighina/Latin-ISE-WSD
A large scale automatic analysis of selected lemmas sense change across centuries based on the Latin-ISE corpus and the original BERT-based word sense disambiguation system by Bamman et al. (2020)
Ighina/LLMClimate2024
Code for our paper at NLP4PI at EMNLP 2024
Ighina/MultimodalTopicSegmentation
Repository implementing multimodal topic segmentation in the embedding space as described in the paper Multimodal Topic Segmentation with pre-trained Neural Encoders.
Ighina/Ninety_Words
Repo for project for the British Council
Ighina/SemanticEgoNetwork
codes to perform exploratory semantic network analysis on one concept of interest
Ighina/SemanticNetworkVizR
codes to perform semantic network analysis on multiple concepts (defined as multiple words-set, i.e. dictionaries) across multiple texts with R
Ighina/TextGeneration
Qui c'è la ricetta di base per addestrare un nuovo modello neurale di generazione di testo partendo da testi arbitrari.