Muralidhar Sortur
ABSTRACT
In E-Commerce, Product Matching is one of the fundamental problems for various use cases like (1) Competitive pricing of products, (2) deduplication of products in catalog, (3) grouping items from various merchants (4) Recommending products. The requirement is to match a product accurately against a catalog spread across tens of thousands of taxonomy nodes and millions of items. Product matching results must be accurate, and the margin for error is minimal to use Product Matching across use cases. This paper proposes a combination of Deep Learning models integrated into the scalable architecture to achieve the required results. Here we have approached the problem at the grass-root level consisting of five stages (1) Identifying attributes per taxonomy node (2) classification of products (3) Attribute Enrichment from NER (Text) and Image feature extraction (4) Search against multiple indices and filter results for mandatory attributes (5) Re-rank to improve the relevancy of shortlisted results. We have defined product data quality and measured it at every stage to improve the overall performance of Product Matching. This approach has yielded accurate product matching results at scale minimising false-positives
- [1] P. Langley, Crafting Papers on Machine Learning, 2000, pp.1207–1216, Pat Langley, Proceedings of the 17th International Conference on Machine Learning (ICML 2000), Stanford, CA, Morgan Kaufmann.Google Scholar
- [2] Sun, Chi and Qiu, Xipeng and Xu, Yige and Huang, XuanjingGoogle Scholar
- [3] Piotr Bojanowski and Edouard Grave and Armand Joulin and Tomás Mikolov Enriching Word Vectors with Subword Information CoRR abs/1607.04606 2016 http://arxiv.org/abs/1607.04606 arXiv 1607.0460 Mon, 28 Dec 2020 11:31:02 +0100 https://dblp.org/rec/journals/corr/BojanowskiGJM16.bib dblp computer science bibliography, https://dblp.orgGoogle Scholar
- [4] J. Clerk Maxwell, A Treatise on Electricity and Magnetism, 3rd ed., vol. 2. Oxford: Clarendon, 1892, pp.68–73.Google Scholar
- [5] I. S. Jacobs and C. P. Bean, “Fine particles, thin films and exchange anisotropy,” in Magnetism, vol. III, G. T. Rado and H. Suhl, Eds. New York: Academic, 1963, pp. 271–350.Google Scholar
- [6] K. Elissa, “Title of paper if known,” unpublished.Google Scholar
- [7] R. Nicole, “Title of paper with only first word capitalized,” J. Name Stand. Abbrev., in press.Google Scholar
- [8] Y. Yorozu, M. Hirano, K. Oka, and Y. Tagawa, “Electron spectroscopy studies on magneto-optical media and plastic substrate interface,” IEEE Transl. J. Magn. Japan, vol. 2, pp. 740–741, August 1987 [Digests 9th Annual Conf. Magnetics Japan, p. 301, 1982].Google ScholarCross Ref
- [9] M. Young, The Technical Writer’s Handbook. Mill Valley, CA: University Science, 1989.Google Scholar
- [10] Joulin, Armand and Grave, Edouard and Bojanowski, Piotr and Mikolov, Tomas,Bag of Tricks for Efficient Text Classification, 2016Google Scholar
- [11] Devlin, Jacob and Chang, Ming-Wei and Lee, Kenton and Toutanova, Kristina, BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding, 2018Google Scholar
- [12]Ajinkya, Attribute Extraction from Product Titles in eCommerce, 2016Google Scholar
- [13] Huang, Zhiheng and Xu, Wei and Yu, Kai, Bidirectional LSTM-CRF Models for Sequence Tagging, 2015Google Scholar
- [14] Stahlberg, Felix, Neural Machine Translation: A Review and Survey, 2019Google Scholar
- [15] Ratner, Alexander and Bach, Stephen H. and Ehrenberg, Henry and Fries, Jason and Wu, Sen and Ré, Christopher, Snorkel: Rapid Training Data Creation with Weak Supervision, 2017Google Scholar
- [16] Kaiming He and Georgia Gkioxari and Piotr Dollár and Ross Girshick, Mask R-CNN, 2018Google Scholar
- [17] Ajinkya More, Product Matching in eCommerce using deep learning, 2017Google Scholar
- [18] Janusz Tracz and Piotr Iwo Wójcik and Kalina Jasinska-Kobus and Riccardo Belluzzo and Robert Mroczkowski and Ireneusz Gawlik, BERT-based similarity learning for product matching, 2020Google Scholar
- [19] Edouard Grave, et.al Efficient softmax approximation for GPUs, 2017Google Scholar
- [20] Armand Joulin, Edouard Grave, Piotr Bojanowski, Tomas Mikolov, Bag of Tricks for Efficient Text Classification, 2016Google Scholar
- [21]Tomas Mikolov, Kai Chen, Greg Corrado, Jeffrey Dean, Efficient Estimation of Word Representations in Vector Space, 2013Google Scholar
- [22] Kilian Weinberger, Anirban Dasgupta, Josh Attenberg, John Langford, Alex Smola,Feature Hashing for Large Scale Multitask Learning, 2010Google Scholar
Index Terms
- E-Commerce Product Matching at Internet Scale
Recommendations
The WDC Training Dataset and Gold Standard for Large-Scale Product Matching
WWW '19: Companion Proceedings of The 2019 World Wide Web ConferenceA current research question in the area of entity resolution (also called link discovery or duplicate detection) is whether and in which cases embeddings and deep neural network based matching methods outperform traditional symbolic matching methods. ...
Using schema.org Annotations for Training and Maintaining Product Matchers
WIMS 2020: Proceedings of the 10th International Conference on Web Intelligence, Mining and SemanticsProduct matching is a central task within e-commerce applications such as price comparison portals and online market places. State-of-the-art product matching methods achieve F1 scores above 0.90 using deep learning techniques combined with huge amounts ...
ProMap: Product Mapping Datasets
Advances in Information RetrievalAbstractThe goal of product mapping is to decide, whether two listings from two different e-shops describe the same products. Existing datasets of matching and non-matching pairs of products, however, often suffer from incomplete product information or ...
Comments