Multilingual Search Optimization Using BERT AI and Product Knowledge Bases

References

[1] Radford, A., Narasimhan, K., Salimans, T., & Sutskever, I. (2018). Improving language understanding by generative pre-training. OpenAI Blog. Retrieved from https://openai.com/blog/language-unsupervised

[2] Schwartz, B. (2019). How BERT improves Google search results. Search Engine Land. Retrieved from https://searchengineland.com/how-bert-improves-google-search-results-32485

[3] Singhal, A. (2011). Google search and search engine optimization (SEO). Google Webmaster Central Blog. Retrieved from https://webmasters.googleblog.com/2011/05/more-guidance-on-building-high-quality.html

[4] Li, Y., Shen, S., Ma, H., Xiao, H., & Jin, X. (2020). Deep learning for query understanding in search engines: A comprehensive survey. ACM Transactions on Information Systems (TOIS), 38(4), 1-38.

[5] Dehghani, M., Zamani, H., Severyn, A., Kamps, J., & Croft, W. B. (2017). Neural ranking models with weak supervision. In Proceedings of the 40th International ACM SIGIR Conference on Research and Development in Information Retrieval (pp. 65-74). ACM.

[6] Yin, W., Schütze, H., Xiang, B., Zhou, B., & Zhang, B. (2019). Benchmarking neural network robustness to common corruptions and perturbations. In Proceedings of the 36th International Conference on Machine Learning (ICML) (Vol. 97, pp. 7204-7213). PMLR.

[7] Halevy, A.,Norvig, P., & Pereira, F. (2009). The unreasonable effectiveness of data. IEEE Intelligent Systems, 24(2), 8-12.

[8] Deng, L., & Yu, D. (2014). Deep learning: methods and applications. Foundations and Trends in Signal Processing, 7(3-4), 197-387.

[9] Radinsky, K., & Horowitz, E. (2013). The Google way: How to use Google to do everything!. Pearson Education India.

[10] Singhal, A. (2011). Google search and search engine optimization (SEO). Google Webmaster Central Blog. Retrieved from https://webmasters.googleblog.com/2011/05/more-guidance-on-building-high-quality.html

[11] D. Cer et al., “Universal Sentence Encoder,” EMNLP demonstration, Association for Computational Linguistics, pp. 1–7, 2018, doi: 10.48550/arXiv.1803.11175.

[12] K. M. Hermann et al., “Teaching machines to read and comprehend,” Advances in Neural Information Processing Systems, vol. 2015, pp. 1693–1701, 2015.

[13] M. E. Peters et al., “Deep contextualized word representations,” NAACL HLT 2018 - 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies - Proceedings of the Conference, vol. 1, pp. 2227–2237, 2018, doi: 10.18653/v1/n18-1202

[14] Y. Nie, S. Wang, and M. Bansal, “Revealing the importance of semantic retrieval for machine reading at scale,” EMNLP-IJCNLP 2019 - 2019 Conference on Empirical Methods in Natural Language Processing and 9th International Joint Conference on Natural Language Processing, Proceedings of the Conference, pp. 2553–2566, 2019, doi: 10.18653/v1/d19-1258

[15] L. Sharma, L. Graesser, N. Nangia, and U. Evci, “Natural language understanding with the quora question pairs dataset,” arXiv-Computer Science, pp. 1-10, 2019.

[16] Adhikari, A., Ram, A., Tang, R., Lin, J.: DocBERT: BERT for document classification.arXiv preprint arXiv:1904.08398 (2019)

[17] Beltagy, I., Lo, K., Cohan, A.: SciBERT: A pretrained language model for scientifictext. arXiv preprint arXiv:1903.10676 (2019)

[18] Benoit, K., Watanabe, K., Wang, H., Nulty, P., Obeng, A., M¨uller, S., Matsuo, A.:Quanteda: An R package for the quantitative analysis of textual data 3(30), 774(2018)

[19] Campello, R.J., Moulavi, D., Sander, J.: Density-based clustering based on hierarchicaldensity estimates. In: Pacific-Asia Conference on Knowledge Discovery and DataMining, pp. 160–172 (2013). Springer

[20] Devlin, J., Chang, M.-W., Lee, K., Toutanova, K.: BERT: Pre-training of deep bidi-rectional transformers for language understanding. In: Proceedings of naacL-HLT,vol. 1, p. 2 (2019)

[21] Hovy, D.: Demographic factors improve classification performance. In: Proceedingsof the 53rd Annual Meeting of the Association for Computational Linguistics andthe 7th International Joint Conference on Natural Language Processing (Volume 1:Long Papers), pp. 752–762 (2015)

[22] Liu, Y., Ott, M., Goyal, N., Du, J., Joshi, M., Chen, D., Levy, O., Lewis, M.,Zettlemoyer, L., Stoyanov, V.: RoBERTa: A robustly optimized BERT pretrainingapproach. arXiv preprint arXiv:1907.11692 (2019)

[23] Liu, Y., Ott, M., Goyal, N., Du, J., Joshi, M., Chen, D., Levy, O., Lewis, M.,Zettlemoyer, L., Stoyanov, V.: RoBERTa: A robustly optimized BERT pretrainingapproach. arXiv preprint arXiv:1907.11692 (2019)

[24] Qi, S., Wong, C.U.I., Chen, N., Rong, J., Du, J.: Profiling Macau cultural tourists byusing user-generated content from online social media. Information Technology &Tourism 20, 217–236 (2018)

[25] S. Qaiser and R. Ali, “Text Mining: Use of TF-IDF to Examine the Relevance of Words to Documents,” International Journal of Computer Applications, vol. 181, no. 1, pp. 25–29, 2018, doi: 10.5120/ijca2018917395.

@article{1701028,
      author = {Martin Louis},
      title = {Multilingual Search Optimization Using BERT AI and Product Knowledge Bases},
      journal = {Iconic Research And Engineering Journals},
      year = {2019},
      volume = {2},
      number = {9},
      pages = {208-223},
      issn = {2456-8880},
      url = {https://www.irejournals.com/formatedpaper/1701028.pdf},
      abstract = {Multilingual search optimization is important to help users obtain precise and relevant information in different languages. However, traditional search techniques typically have a hard time with semantic subtleties, cultural differences, and the difficulties of low-resource languages. In this work, we propose a new approach combining BERT's power (Bidirectional Encoder Representations from Transformers) with domain-specific product knowledge bases to improve multilingual search performance. With natural language understanding capabilities of the advanced BERT model and structured product knowledge integrated, our model enhances query understanding, reduces context irrelevancy, and improves the accuracy of search results. 
The proposed system is tested on high- and low-resource multilingual datasets, and it achieved considerable improvements in precision, recall, and Mean Reciprocal Rank (MRR) over baseline methods, including TF-IDF and BM25. Additionally, the model could better resolve ambiguous queries and provide domain-specific insights by adding product knowledge bases. The impact of this approach on user experience, cultural adaptation, and cross-linguistic consistency is also evaluated in the study. 
Thorough model comparison and ablation analysis were demonstrated to show that the combination of BERT and knowledge bases outperforms traditional and modern search optimization techniques. This research makes the case for AI-powered multilingual search optimization and offers a basis for continued innovation in global search systems.},
      keywords = {Multilingual Search, BERT, AI, Product Knowledge Bases, Search Optimization, Natural Language Processing, Cross-lingual Search},
      month = {March},
  }