Papers by Mark Stevenson

9 papers
Combining Counting Processes and Classification Improves a Stopping Rule for Technology Assisted Review (2023.findings-emnlp)

Copied to clipboard

Challenge: Experiments on multiple data sets show that the proposed approach consistently improves performance and outperforms several alternative methods.
Approach: They propose to integrate a text classifier into an existing TAR stopping rule to train it without the need for additional annotations.
Outcome: Experiments on multiple data sets show the proposed approach outperforms other methods and achieves the desired level of recall with a lower cost than the existing method based on counting processes alone.
ParaPat: The Multi-Million Sentences Parallel Corpus of Patents Abstracts (2020.lrec-1)

Copied to clipboard

Challenge: Existing parallel corpora for patents and scientific texts are not available due to the need for correct alignment and human curation.
Approach: They develop a parallel corpus from the open access Google Patents dataset . they use Hunalign algorithm to align sentences and tokens using the largest 22 languages .
Outcome: The proposed corpus is available in TSV format and with a SQLite database, with complementary information regarding patent metadata.
Highly Efficient Knowledge Graph Embedding Learning with Orthogonal Procrustes Analysis (2021.naacl-main)

Copied to clipboard

Challenge: Knowledge Graph Embeddings (KGEs) have been explored in recent years due to their promise for a wide range of applications.
Approach: They propose a KGE framework which can reduce the training time and carbon footprint by orders of magnitudes compared with state-of-the-art approaches.
Outcome: The proposed framework reduces the training time and carbon footprint by orders of magnitudes compared with state-of-the-art approaches while producing competitive performance.
HiDE: a Tool for Unrestricted Literature Based Discovery (C18-2)

Copied to clipboard

Challenge: Literature based discovery (LBD) is an automatic technique that infers as yet unobserved connections from literature.
Approach: They propose a literature based discovery tool which allows fast access to hidden connections generated from all abstracts in Medline.
Outcome: The tool allows users to explore the full range of hidden connections generated by an LBD system.
Modelling Stopping Criteria for Search Results using Poisson Processes (D19-1)

Copied to clipboard

Challenge: Document retrieval systems often return large sets of documents, especially when applied to large collections.
Approach: They propose a method that predicts the rate at which relevant documents occur using a Poisson process and allows a user to specify a minimum desired level of recall to achieve.
Outcome: The proposed method is compared with previous methods on a public dataset and compares it with existing methods.
Topic or Style? Exploring the Most Useful Features for Authorship Attribution (C18-1)

Copied to clipboard

Challenge: Existing approaches to authorship attribution rely on individual's writing style and/or preferred topics.
Approach: They analyse four widely used datasets to explore how different types of features affect authorship attribution accuracy under varying conditions.
Outcome: The proposed model outperforms the state-of-the-art on two out of the four datasets used.
Document Set Expansion with Positive-Unlabeled Learning Using Intractable Density Estimation (2024.lrec-main)

Copied to clipboard

Challenge: Existing approaches to Document Set Expansion (DSE) rely on the unrealistic assumption of knowing the class prior for positive samples in the collection.
Approach: They propose a novel method that utilizes intractable density estimation models to learn the class prior for positive samples in the collection.
Outcome: The proposed method is based on a set of examples from PubMed and Covid datasets in a transductive setting.
Robustness and Reliability of Gender Bias Assessment in Word Embeddings: The Role of Base Pairs (2020.aacl-main)

Copied to clipboard

Challenge: Existing methods to quantify gender bias in word embeddings are not robust and cannot identify common types of bias.
Approach: They propose to quantify gender bias by using cosine similarity to a pair of gender words and using analogies.
Outcome: The proposed methods are not robust and cannot identify common types of bias, while analogies are unsuitable indicators.
Cross-Lingual Word Embedding Refinement by ℓ1 Norm Optimisation (2021.naacl-main)

Copied to clipboard

Challenge: Existing methods for building high-quality CLWEs learn mappings that minimise the l2 norm loss function but this optimisation objective has been shown to be sensitive to outliers.
Approach: They propose a simple post-processing step to improve cross-lingual word embeddings using the Manhattan norm goodness-of-fit criterion.
Outcome: The proposed approach outperforms four state-of-the-art baselines in bilingual lexicon induction and cross-lingual transfer tasks.

What is GenGO?

GenGO is an NLP powered publication search system. It currenctly indexes 30k+ papers from ACL Anthology, and implements multi-aspect summarization, semantic search, and more!

Information

About
Limitations