Can We Survive without Labelled Data in NLP? Transfer Learning for Open Information Extraction

Injy Sarhan,Marco Spruit

doi:10.3390/app10175758

Abstract

Various tasks in natural language processing (NLP) suffer from lack of labelled training data, which deep neural networks are hungry for. In this paper, we relied upon features learned to generate relation triples from the open information extraction (OIE) task. First, we studied how transferable these features are from one OIE domain to another, such as from a news domain to a bio-medical domain. Second, we analyzed their transferability to a semantically related NLP task, namely, relation extraction (RE). We thereby contribute to answering the question: can OIE help us achieve adequate NLP performance without labelled data? Our results showed comparable performance when using inductive transfer learning in both experiments by relying on a very small amount of the target data, wherein promising results were achieved. When transferring to the OIE bio-medical domain, we achieved an F-measure of 78.0%, only 1% lower when compared to traditional learning. Additionally, transferring to RE using an inductive approach scored an F-measure of 67.2%, which was 3.8% lower than training and testing on the same task. Hereby, our analysis shows that OIE can act as a reliable source task.

Highlights

In deep learning for natural language processing (NLP), the collection of labelled data necessary for training and building models is expensive
It is worth noting that the dimensionality of the word embeddings refers to the length of the vector; in theory the size of the vector is directly proportional to the information it can store, which allows
We found an improvement of 12.8% when compared to transductive learning using a 4:1 ratio, with the Open information extraction (OIE) news dataset overtaking the higher ratio

Summary

Introduction

In deep learning for natural language processing (NLP), the collection of labelled data necessary for training and building models is expensive. This has further highlighted the urgency towards transfer learning research. The aim of transfer learning is to benefit from information gathered from previous training data in directly making predictions in the target task by utilizing the extracted information. Open information extraction (OIE) is a challenging task of extracting relation tuples from an unstructured corpus. The extracted tuples can be binary, ternary, or n-ary, where the relationship is expressed between more than two entities such as the Person–Location–BornIn–BornOn relation (Jack Adams, Michigan, California, 1975)

Objectives

Methods

Results

Conclusion