Data augmentation uses artificially created examples to support supervised machine learning, adding robustness to the resulting models and helping to account for limited availability of labelled data. We apply and evaluate a synthetic data approach to relationship classification in digital libraries, generating artificial books with relationships that are common in digital libraries but not easier inferred from existing metadata. Artificial books are generated by remixing existing texts into synthetically constructed formats. We find that for classification on whole–part relationships between books, synthetic data improves a deep neural network classifier by 91%. Furthermore, we consider the ability of synthetic data to learn a useful new text relationship class from fully artificial training data.
BammanDCarneyMGillickJet al. Estimating the date of first publication in a large-scale digital library. In: 2017 ACM/IEEE joint conference on digital libraries (JCDL), Toronto, ON, Canada, 19–23 June 2017, pp. 1–10. New York: IEEE.
2.
OrganisciakPShetenhelmSVasquesDFAet al. Characterizing same work relationships in large-scale digital libraries. In: TaylorNGChristian-LambCMartinMHet al. (eds.) Information in contemporary society. Lecture notes in computer Science. Cham: Springer International Publishing, 2019, pp. 419–425.
SchofieldAThompsonLMimnoD. Quantifying the effects of text duplication on semantic models. In: Conference on empirical methods on natural language processing, Copenhagen, 7–11 September 2017.
5.
American Library Association, Canadian Library Association, Chartered Institute of Library and Information Professionals, and Joint Steering Committee for Development of RDA. RDA: Resource Description and Access: 2013 Revision, 2013. DOI: 10.29085/9781856047159
6.
IFLA Study Group on the Functional Requirements for Bibliographic Records. Functional requirements for bibliographic records. Technical report, IFLA, Munich, 1998.
7.
MichelJBShenYKAidenAPet al. Quantitative analysis of culture using millions of digitized books. Science2011; 331(6014): 176–182.
8.
YorkJ. Building a future by preserving our past: the preservation infrastructure of HathiTrust digital library. In: World library and information congress: 76th IFLA general conference and assembly, Gothenburg, 10–15 August 2010, pp. 10–15, https://www.hathitrust.org/documents/hathitrust-ifla-201008.pdf
9.
HamiltonWLLeskovecJJurafskyD.Diachronic word embeddings reveal statistical laws of semantic change. arXiv pre print arXiv:1605.09096, 2016, https://arxiv.org/abs/1605.09096
10.
LiebermanEMichelJBJacksonJet al. Quantifying the evolutionary dynamics of language. Nature2007; 449: 713–716.
11.
MorettiF.Distant reading. London: Verso Books, 2013.
SmithDACordellR.A research agenda for historical and multilingual optical character recognition. Technical report, NULab, Northeastern University, 2018, https://ocr.northeastern.edu/report
14.
ManmathaRFengS. A hierarchical, HMM-based automatic evaluation of OCR accuracy for a digital library of books. In: Proceedings of the 6th ACM/IEEE-CS joint conference on digital libraries (JCDL ’06), Chapel Hill, NC, 11–15 June 2006, pp. 109–118. New York: IEEE.
15.
DongRSmithD. Multi-input attention for unsupervised OCR correction. In: Proceedings of the 56th annual meeting of the association for computational linguistics (Volume 1: Long Papers), Melbourne, VIC, Australia, 15–20 July 2018, pp. 2363–2372. Melbourne, VIC, Australia: Association for Computational Linguistics.
MikołajczykAGrochowskiM. Data augmentation for improving deep learning in image classification problem. In: 2018 international interdisciplinary PhD workshop (IIPhDW), Swinoujscie, 9–12 May 2018, pp. 117–122. New York: IEEE.
18.
WongSCGattAStamatescuVet al. Understanding data augmentation for classification: when to warp? In: 2016 international conference on digital image computing: techniques and applications (DICTA), Gold Coast, QLD, Australia, 30 November–2 December 2016, pp. 1–6. New York: IEEE.
19.
PerezLWangJ.The effectiveness of data augmentation in image classification using deep learning. arXiv: 1712.04621, 2017, http://arxiv.org/abs/1712.04621
20.
YouZYeJLiKet al. Adversarial noise layer: regularize neural network by adding noise. In: 2019 IEEE International Conference on Image Processing (ICIP), Taipei, Taiwan, 22–25 September 2019, pp. 909–913. New York: IEEE.
21.
PengXSunBAliKet al. Learning deep object detectors from 3D models. arXiv: 1412.7122, 2015, http://arxiv.org/abs/1412.7122
22.
RichterSRVineetVRothSet al. Playing for data: ground truth from computer games. In: LeibeBMatasJSebeNet al. (eds) Computer vision ECCV 2016. Lecture notes in computer science. Cham: Springer International Publishing, 2016, pp. 102–118.
23.
WeiJZouK.EDA: easy data augmentation techniques for boosting performance on text classification tasks. arXiv: 1901.11196, 2019, http://arxiv.org/abs/1901.11196
24.
VaibhavVSinghSStewartCet al. Improving robustness of machine translation with synthetic noise. arXiv: 1902.09508, 2019, http://arxiv.org/abs/1902.09508
25.
ChironGDoucetACoustatyMet al. ICDAR2017 competition on post-OCR text correction. In: 2017 14th IAPR international conference on document analysis and recognition (ICDAR), vol. 1, Kyoto, Japan, 9–15 November 2017, pp. 1423–1428. New York: IEEE.
26.
RigaudCDoucetACoustatyMet al. ICDAR 2019 competition on post-OCR text correction. In: 2019 International Conference on Document Analysis and Recognition (ICDAR), Sydney, NSW, Australia, 20–25 September 2019, pp. 1588–1593. New York: IEEE.
27.
Fonseca CachoJRTaghvaK. Aligning ground truth text with OCR degraded text. In: AraiKBhatiaRKapoorS (eds) Intelligent computing. Advances in intelligent systems and computing. Cham: Springer International Publishing, 2019, pp. 815–833.
28.
DrobacSKauppinenPLindK. OCR and post-correction of historical Finnish texts. In: Proceedings of the 21st Nordic conference on computational linguistics, Gothenburg, 22–24 May 2017, pp. 70–76. Melbourne, VIC, Australia: Association for Computational Linguistics.
29.
ChernyshovaYSGayerAVSheshkusAV. Generation method of synthetic training data for mobile OCR system. In: Tenth international conference on machine vision (ICMV 2017), vol. 10696, Vienna, 13–15 November 2017, p. 106962G. Bellingham, WA: International Society for Optics and Photonics.
KangLRusinolMFornesAet al. Unsupervised adaptation for synthetic-to-real handwritten word recognition. In: 2020IEEE winter conference on applications of computer vision (WACV), Snowmass, CO, 1–5 March, pp. 3502–3511
32.
PuriRSpringRPatwaryMet al. Training question answering models from synthetic data. arXiv: 2002.09599, 2020, http://arxiv.org/abs/2002.09599
33.
ZhuXLiuYLiJet al. Emotion classification with data augmentation using generative adversarial networks. In: PhungDTsengVSWebbGIet al. (eds) Advances in knowledge discovery and data mining. Lecture notes in computer science. Cham: Springer International Publishing, 2018, pp. 349–360.
Frid-AdarMKlangEAmitaiMet al. Synthetic data augmentation using GAN for improved liver lesion classification. In: 2018 IEEE 15th international symposium on biomedical Imaging (ISBI 2018), Washington, DC, 4–7 April 2018, pp. 289–293. New York: IEEE.
PenningtonJSocherRManningC. GloVe: global vectors for word representation. In: Proceedings of the 2014 conference on empirical methods in natural language processing (EMNLP), pp. 1532–1543, https://nlp.stanford.edu/pubs/glove.pdf
38.
SrivastavaNHintonGKrizhevskyAet al. Dropout: a simple way to prevent neural networks from overfitting. J Mach Learn Res2014; 15(1): 1929–1958.
