• Home
• Papers & Talks
• CV
• Teaching

Ash (Afshin) Rahimi

Applied Scientist
Amazon, Alexa

Email
Google Scholar
Github
Linkedin

My research interests fall within the fields of Natural Language Processing, Social Network Analysis and Machine Learning. I am specifically interested in exploiting both structured and unstructured data to help machines understand conversational language in Emergency Situations and Health Informatics.

News

• After three good years at UQ, I'm joining Amazon to work on exciting user-facing projects.
• Class-imbalance and fairness are often studied separately, our new paper, "Fairness-aware Class Imbalanced Learning" (EMNLP2021) tries to bridge the gap. Paper, Slides and Code.
• Two papers accepted to COLING2020, one on aligning Wikipedia and UMLS, and another on Indonesian pretrained models and benchmarks (see my GScholar).
• Paper with Gaurav Arora in ALTA2019 on catastrophic forgetting in NLP applications.
• I joined The University of Queensland as a lecturer (Assisstant Professor). If you're interested in working with me on NLP/Social Media/Health, please email me.
• There are more than 6000 languages without any annotation, our new ACL-19 paper on transfer learning (few shot and zero shot NER) for those languages+ slides .
• Geotagged data is scarce, our new ACL-18 paper on semi-supervised user geolocation using graph convolutional networks + slides .
• I defended my PhD, and submitted my thesis. The slides are available here.
• Our EMNLP2017 paper on using Mixture Density Networks for geolocation and RBF networks for lexical dialectology is published. The code and slides are available.
• Our ACL 2017 paper on geolocation (predict location given text) and lexical dialectology (predict text given location) where we talk about utilising both network and text information for geolocation and also use the geolocation model to retrieve dialect words within U.S. is available. The code and the evaluation dataset can be found here. It was selected as one of the outstanding papers!
• The Youtube Demo and Web UI of our geolocation tool is ready to use. The corresponding paper is published at ACL 2016, demonstration papers.
• Our paper on geolocation of social media users is accepted at ACL 2015. We use social relationships and also tweet content of Twitter users to find where they live!
• Our paper on geolocation of social media users is accepted at NAACL 2015. We use social relationships and also tweet content of Twitter users to find where they live!

Selected Papers

Fairness-aware Class Imbalanced Learning
Shivashankar Subramanian, Afshin Rahimi, Timothy Baldwin, Trevor Cohn, Lea Frermann. In EMNLP 2021., 2019.
Abstract

×

Fairness-aware Class Imbalanced Learning

Class imbalance is a common challenge in many NLP tasks, and has clear connections to bias, in that bias in training data often leads to higher accuracy for majority groups at the expense of minority groups. However there has traditionally been a disconnect between research on class-imbalanced learning and mitigating bias, and only recently have the two been looked at through a common lens. In this work we evaluate long-tail learning methods for tweet sentiment and occupation classification, and extend a margin-loss based approach with methods to enforce fairness. We empirically show through controlled experiments that the proposed approaches help mitigate both class imbalance and demographic biases.

Code

WikiUMLS: Aligning UMLS to Wikipedia via Cross-lingual Neural Ranking
Afshin Rahimi, Timothy Baldwin, Karin Verspoor. In COLING 2020., 2020.
Abstract

×

WikiUMLS: Aligning UMLS to Wikipedia via Cross-lingual Neural Ranking

We present our work on aligning the Unified Medical Language System (UMLS) to Wikipedia, to facilitate manual alignment of the two resources. We propose a cross-lingual neural reranking model to match a UMLS concept with a Wikipedia page, which achieves a recall@1 of 72%, a substantial improvement of 20% over word- and char-level BM25, enabling manual alignment with minimal effort. We release our resources, including ranked Wikipedia pages for 700k UMLS concepts, and WikiUMLS, a dataset for training and evaluation of alignment models between UMLS and Wikipedia. This will provide easier access to Wikipedia for health professionals, patients, and NLP systems, including in multilingual settings.

Code

IndoLEM and IndoBERT: A Benchmark Dataset and Pre-trained Language Model for Indonesian NLP
F Koto, A Rahimi, JH Lau, T Baldwin. In COLING 2020., 2020.
Abstract

×

IndoLEM and IndoBERT: A Benchmark Dataset and Pre-trained Language Model for Indonesian NLP

Although the Indonesian language is spoken by almost 200 million people and the 10th mostspoken language in the world, it is under-represented in NLP research. Previous work on Indonesian has been hampered by a lack of annotated datasets, a sparsity of language resources, and a lack of resource standardization. In this work, we release the INDOLEM dataset comprising seven tasks for the Indonesian language, spanning morpho-syntax, semantics, and discourse. We additionally release INDOBERT, a new pre-trained language model for Indonesian, and evaluate it over INDOLEM, in addition to benchmarking it against existing resources. Our experiments show that INDOBERT achieves state-of-the-art performance over most of the tasks in INDOLEM.

Code

Massively Multilingual Transfer for NER
Afshin Rahimi, Yuan Li, and Trevor Cohn. In ACL 2019., 2019.
Abstract

×

Massively Multilingual Transfer for NER

In cross-lingual transfer, NLP models over one or more source languages are applied to a low- resource target language. While most prior work has used a single source model or a few carefully selected models, here we consider a “massive” setting with many such mod- els. This setting raises the problem of poor transfer, particularly from distant languages. We propose two techniques for modulating the transfer, suitable for zero-shot or few-shot learning, respectively. Evaluating on named entity recognition, we show that our tech- niques are much more effective than strong baselines, including standard ensembling, and our unsupervised method rivals oracle selec- tion of the single best individual model.

Code

Semi-supervised User Geolocation via Graph Convolutional Networks
Afshin Rahimi, Trevor Cohn and Timothy Baldwin. In ACL 2018., 2018.
Abstract

×

Semi-supervised User Geolocation via Graph Convolutional Networks

Social media user geolocation is vital to many applications such as event detection. In this paper, we propose GCN, a multiview geolocation model based on Graph Convolutional Networks, that uses both text and network context. We compare GCN to the state-of-the-art, and to two baselines we propose, and show that our model achieves or is competitive with the state-of-the-art over three benchmark geolocation datasets when sufficient supervision is available. We also evaluate GCN under a minimal supervision scenario, and show it outperforms baselines. We find that highway network gates are essential for controlling the amount of useful neighbourhood expansion in GCN.

Code

Continuous Representation of Location for Geolocation and Lexical Dialectology using Mixture Density Networks
Afshin Rahimi, Timothy Baldwin and Trevor Cohn. In EMNLP 2017., 2017.
Abstract

×

Continuous Representation of Location for Geolocation and Lexical Dialectology using Mixture Density Networks

We propose a method for embedding two- dimensional locations in a continuous vec- tor space using a neural network-based model incorporating mixtures of Gaussian distributions, presenting two model vari- ants for text-based geolocation and lexi- cal dialectology. Evaluated over Twitter data, the proposed model outperforms con- ventional regression-based geolocation and provides a better estimate of uncertainty. We also show the effectiveness of the rep- resentation for predicting words from loca- tion in lexical dialectology, and evaluate it using the DARE dataset.

Code

A Neural Model for User Geolocation and Lexical Dialectology
Afshin Rahimi, Trevor Cohn and Timothy Baldwin. In ACL 2017, Short papers., 2017.
Abstract

×

A Neural Model for User Geolocation and Lexical Dialectology

We propose a simple yet effective textbased user geolocation model based on a neural network with one hidden layer, which achieves state of the art performance over three Twitter benchmark geolocation datasets, in addition to producing word and phrase embeddings in the hidden layer that we show to be useful for detecting dialectal terms. As part of our analysis of dialectal terms, we release DAREDS, a dataset for evaluating dialect term detection methods.

Code

pigeo, A Python Geotagging Tool
Afshin Rahimi, Trevor Cohn and Timothy Baldwin. In ACL 2016, Demonstration papers., 2016.
Abstract

×

pigeo, A Python Geotagging Tool

We present pigeo, a Python geolocation prediction tool that predicts a location for a given text input or Twitter user. We discuss the design, implementation and application of pigeo, and empirically evaluate it. pigeo is able to geolocate informal text and is a very useful tool for users who require a free and easy-to-use, yet accurate geolocation service based on pre-trained models. Additionally, users can train their own models easily using pigeo’s API.

Code

Twitter User Geolocation Using a Unified Text and Network Prediction Model
Afshin Rahimi, Trevor Cohn and Timothy Baldwin. In ACL 2015, Short papers., 2015.
Abstract

×

Twitter User Geolocation Using a Unified Text and Network Prediction Model

Abstract We propose a label propagation approach to geolocation prediction based on Modified Adsorption, with two enhancements: (1) the removal of “celebrity” nodes to increase location homophily and boost tractability; and (2) the incorporation of text-based geolocation priors for test users. Experiments over three Twitter benchmark datasets achieve state-of-the-art results, and demonstrate the effectiveness of the enhancements.

Code

Exploiting text and network context for geolocation of social media users
Afshin Rahimi, Duy Vu, Trevor Cohn and Timothy Baldwin. In NAACL-HLT 2015, Short papers., 2015.
Abstract

×

Exploiting text and network context for geolocation of social media users

Abstract Research on automatically geolocating social media users has conventionally been based on the text content of posts from a given user or the social network of the user, with very little crossover between the two, and no benchmarking of the two approaches over comparable datasets. We bring the two threads of research together in first proposing a text-based method based on adaptive grids, followed by a hybrid network- and text-based method. Evaluating over three Twitter datasets, we show that the empirical difference between textand network-based methods is not great, and that hybridisation of the two is superior to the component methods, especially in contexts where the user graph is not well connected. We achieve state-of-the-art results on all three datasets.

Code