Low-resource Deep Entity Resolution with Transfer and Active Learning

Kasai, Jungo; Qian, Kun; Gurajada, Sairam; Li, Yunyao; Popa, Lucian

doi:10.18653/v1/p19-1586

Cited by 108 publications

(99 citation statements)

References 34 publications

(40 reference statements)

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…In addition we show the best reported result found in related work from matching systems using supervised learning while matching systems that use other types of learning (e.g. active learning or semisupervised learning) are excluded from this comparison [13,18,22]. The interpretation of the comparison to the results reported in related work should be made with attention to the differing, unfixed train, optimization, and test sets.…”

Section: Baseline Resultsmentioning

confidence: 91%

Profiling Entity Matching Benchmark Tasks

Primpeli

Bizer

2020

Proceedings of the 29th ACM International Conference on Information &Amp; Knowledge Management

View full text Add to dashboard Cite

Entity matching is a central task in data integration which has been researched for decades. Over this time, a wide range of benchmark tasks for evaluating entity matching methods has been developed. This resource paper systematically complements, profiles, and compares 21 entity matching benchmark tasks. In order to better understand the specific challenges associated with different tasks, we define a set of profiling dimensions which capture central aspects of the matching tasks. Using these dimensions, we create groups of benchmark tasks having similar characteristics. Afterwards, we assess the difficulty of the tasks in each group by computing baseline evaluation results using standard feature engineering together with two common classification methods. In order to enable the exact reproducibility of evaluation results, matching tasks need to contain exactly defined sets of matching and non-matching record pairs, as well as a fixed development and test split. As this is not the case for some widely-used benchmark tasks, we complement these tasks with fixed sets of non-matching pairs, as well as fixed splits, and provide the resulting development and test sets for public download. By profiling and complementing the benchmark tasks, we support researchers to select challenging as well as diverse tasks and to compare matching systems on clearly defined grounds.

show abstract

Section: Baseline Resultsmentioning

confidence: 91%

Profiling Entity Matching Benchmark Tasks

Primpeli

Bizer

2020

Proceedings of the 29th ACM International Conference on Information &Amp; Knowledge Management

View full text Add to dashboard Cite

show abstract

“…Recent literature on Machine Learning applied to RL includes Aiken et al (2019), that compares probabilistic, stochastic and machine learning approaches, showing that supervised methods outperform unsupervised ones; Dong & Rekatsinas (2018), that surveys state-of-the-art data integration solutions based on Machine Learning with a discussion of open research challenges; Kasai et al (2019), that leverages Deep Learning in a combination of Transfer and Active Learning aiming to save labeled data up to an order of magnitude; Di Cicco et al (2019), that presents an attempt of explainable Deep Learning exploiting LIME, a popular tool for prediction explanations in classification; Hou et al (2019), that propose a paradigm called ''gradual machine learning'' where data are labeled automatically through iterative factor graph inference, starting with the easiest instances and going up to the hardest ones.…”

Section: Related Workmentioning

confidence: 99%

Record linkage of banks and municipalities through multiple criteria and neural networks

Maratea

Ciaramella

Cianci

2020

PeerJ Computer Science

View full text Add to dashboard Cite

Record linkage aims to identify records from multiple data sources that refer to the same entity of the real world. It is a well known data quality process studied since the second half of the last century, with an established pipeline and a rich literature of case studies mainly covering census, administrative or health domains. In this paper, a method to recognize matching records from real municipalities and banks through multiple similarity criteria and a Neural Network classifier is proposed: starting from a labeled subset of the available data, first several similarity measures are combined and weighted to build a feature vector, then a Multi-Layer Perceptron (MLP) network is trained and tested to find matching pairs. For validation, seven real datasets have been used (three from banks and four from municipalities), purposely chosen in the same geographical area to increase the probability of matches. The training only involved two municipalities, while testing involved all sources (municipalities vs. municipalities, banks vs banks and and municipalities vs. banks). The proposed method scored remarkable results in terms of both precision and recall, clearly outperforming threshold-based competitors.

show abstract

“…Recently, deep learning, in particular, has received increasing attention [3,8,11]. The community also see the need for interactively querying the user for examples, and so exciting work has been done within active learning [5,7].…”

Section: Background and State Of The Artmentioning

confidence: 99%

Tailoring Entity Matching for Industrial Settings

Barlaug

2020

Proceedings of the 29th ACM International Conference on Information &Amp; Knowledge Management

View full text Add to dashboard Cite

Entity matching has received significant attention from the research community over many years. Despite some limited success, most state-of-the-art methods see no widespread usage in industry. In this paper, we present the author's PhD research, which aims at identifying issues that hold techniques and methods developed by the research community back from use in industry, and look at how they might be adapted to address those issues. In our proposed approach, we implement a modular framework, which will be used for real-world user testing and quantitative experiments of our adapted methods. We will have three main contributions from our research: 1) We develop a modular framework for interactive entity matching combining intra-and inter-session iterations. 2) We show how active learning methods for entity matching can be adapted to learn not only classification of matches but also classification of which records are of interest to the user jointly, and how it compares to current methods. 3) We show how deep learning can be used to synthesize interpretable rules for entity matching, and how it compares to traditional methods.

show abstract

Low-resource Deep Entity Resolution with Transfer and Active Learning

Cited by 108 publications

References 34 publications

Profiling Entity Matching Benchmark Tasks

Profiling Entity Matching Benchmark Tasks

Record linkage of banks and municipalities through multiple criteria and neural networks

Tailoring Entity Matching for Industrial Settings

Contact Info

Product

Resources

About