商品搜尋技術對於電商營運的績效至關重要，直接影響到庫存的商品是否能精準地呈現在搜尋客戶的眼前。然而，當下電商搜尋技術的兩大瓶頸在於：欠缺代表性的公開商品搜尋資料庫，作為評測與訓練搜尋模型之用。此外，以詞彙 (lexical) 為基礎的主流搜尋方法已近極限，無法找出與搜尋詞雖在字形上不相似，但在詞義上卻實為相近的商品。本論文「基於深度學習網絡之意圖語意的電子商務資訊檢索技術」的研究目的是針對電商應用：一、建置足夠充份與代表，且具有維護性 (maintainability) 與可擴張性 (scalability) 的測試集蒐集流程。二、設計能反應搜尋詞意圖顆粒度 (granularity) 的語意搜尋 (semantic search) 技術。三、尋求互補詞彙與語意的優點，達到搜尋效果的優化。據此，本論文的貢獻有三：首先是建置公允且代表性的電子商務測試集：由實際電商運營的使用者搜尋記錄中，以分層取樣方式定義了250組測試搜尋詞。透過資訊檢索中 TREC 的流程規範，蒐集與標註搜尋結果商品之相關性。其次是提供以語意為基礎的商品搜尋技術：採用深度學習中能萃取字詞語意的BERT編碼器，以三元組網路架構做訓練。利用使用者搜尋與點擊的記錄，加入能反應搜尋意圖的負樣本組成三元組訓練資料，讓模型學習出搜尋詞與點擊商品名稱之間的語意關係，並得到與詞彙搜尋方法相當的績效。最後是透過融合詞彙與語意的方法達到顯著優化效果：經由合併詞彙與語意模型各別的相似度排序，最終在公正測試集中，相對於BM25詞彙方法而言，於二元指標MAP@50上約提升6%，並在反映多階層相關性的NDCG@50 指標上，提升高達5% 的績效。

Product item search is crucial to the performance of e-commerce operations in that it decides whether the products in stock can be accurately presented to requesting customers. However, there are two bottlenecks affecting current e-commerce search development: One is the lack of a representative and open product search database for evaluating and training search models. The other is the performance limitation posed by the mainstream lexical-based search methods in that product items of no lexical resemblance yet semantically similar to search queries are out of their reach. The objective of this thesis address three issues critical to E-commerce information retrieval technique based on intent semantics: First, to establish a comprehensive and representative test set collection process with maintainability and scalability. Second, to design a semantic search method reflecting granularity of search intent. Third, to take complement both lexical and semantic search methods for better search performance. Accordingly, the contributions of this thesis are three-fold: The first is an exemplary e-commerce test set containing 250 representative search terms, which is obtained by stratified sampling from trace logs of user query from actual e-commerce operation. For each of the search query, TREC compliance for information retrieval is enforced throughout the collection and relevance annotation for each product in the corresponding query results. The second is a semantic-based search technique, which is a refined BERT encoder fined tuned through a triplet network, with triplet training data made of users' search queries and clicks from trace logs, with additionally augmented negative samples that reflect query intention granularity. The BERT-based triplet network thus trained models nicely the semantic relationship between the search terms and the clicked product names, with a satisfactory performance comparable to lexical-based search methods. Finally, there is a superior hybrid search method through re-ranking of combined weighted scores inferenced individually from lexical and semantic search. In comparison to BM25 lexical baseline, the hybrid search method, conducted through the e-commerce test set, attains about 6% improvement in MAP@50, and about 5% improvement for NDCG@50, a performance measure indicating multilevel relevance of search results.

[1] J. Guo, Y. Cai, Y. Fan, F. Sun, R. Zhang, and X. Cheng, “Semantic models for the first-stage retrieval: A comprehensive review,” ACM Transactions on Information Systems (TOIS), vol. 40, no. 4, pp. 1–42, 2022.
[2] H. P. Luhn, “The automatic creation of literature abstracts,” IBM Journal of research and development, vol. 2, no. 2, pp. 159–165, 1958.
[3] K. S. Jones, “A statistical interpretation of term specificity and its application in retrieval,” Journal of documentation, 1972.
[4] S. Robertson, H. Zaragoza, and others, “The probabilistic relevance framework: BM25 and beyond,” Foundations and Trends® in Information Retrieval, vol. 3, no. 4, pp. 333–389, 2009.
[5] J. M. Ponte and W. B. Croft, “A language modeling approach to information retrieval,” in ACM SIGIR Forum, vol. 51, pp. 202– 208, ACM New York, NY, USA, 2017.
[6] P.-S. Huang, X. He, J. Gao, L. Deng, A. Acero, and L. Heck, “Learning deep structured semantic models for web search using clickthrough data,” in Proceedings of the 22nd ACM international conference on Information & Knowledge Management, pp. 2333– 2338, 2013.
[7] B. Hu, Z. Lu, H. Li, and Q. Chen, “Convolutional neural network architectures for matching natural language sentences,” Advances in neural information processing systems, vol. 27, 2014.
[8] Z. Lu and H. Li, “A deep architecture for matching short texts,” Advances in neural information processing systems, vol. 26, 2013.
[9] J. Guo, Y. Fan, Q. Ai, and W. B. Croft, “A deep relevance matching model for ad-hoc retrieval,” in Proceedings of the 25th ACM international on conference on information and knowledge management, pp. 55–64, 2016.
[10] S. Li, F. Lv, T. Jin, G. Lin, K. Yang, X. Zeng, X.-M. Wu, and Q. Ma, “Embedding-based product retrieval in taobao search,” in Proceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery & Data Mining, pp. 3181–3189, 2021.
[11] J.-T. Huang, A. Sharma, S. Sun, L. Xia, D. Zhang, P. Pronin, J. Padmanabhan, G. Ottaviano, and L. Yang, “Embedding-based retrieval in facebook search,” in Proceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, pp. 2553–2561, 2020.
[12] L. Kumar and S. Sarkar, “Neural Search: Learning Query and Product Representations in Fashion E-commerce,” arXiv preprint arXiv:2107.08291, 2021.
[13] W.-C. Chang, F. X. Yu, Y.-W. Chang, Y. Yang, and S. Kumar, “Pre-training tasks for embedding-based large-scale retrieval,” arXiv preprint arXiv:2002.03932, 2020.
[14] J. I. Choi, S. Kallumadi, B. Mitra, E. Agichtein, and F. Javed, “Semantic product search for matching structured product catalogs in e-commerce,” arXiv preprint arXiv:2008.08180, 2020.
[15] Y. Luan, J. Eisenstein, K. Toutanova, and M. Collins, “Sparse, dense, and attentional representations for text retrieval,” Transactions of the Association for Computational Linguistics, vol. 9, pp. 329–345, 2021.
[16] L. Gao, Z. Dai, T. Chen, Z. Fan, B. V. Durme, and J. Callan, “Complement lexical retrieval model with semantic residual embeddings,” in European Conference on Information Retrieval, pp. 146–160, Springer, 2021.
[17] T. Mikolov, I. Sutskever, K. Chen, G. S. Corrado, and J. Dean, “Distributed representations of words and phrases and their compositionality,” Advances in neural information processing systems, vol. 26, 2013.
[18] J. Pennington, R. Socher, and C. D. Manning, “Glove: Global vectors for word representation,” in Proceedings of the 2014 conference on empirical methods in natural language processing (EMNLP), pp. 1532–1543, 2014.
[19] J. Sarzynska-Wawer, A. Wawer, A. Pawlak, J. Szymanowska, I. Stefaniak, M. Jarkiewicz, and L. Okruszek, “Detecting formal thought disorder by deep contextualized word representations,” Psychiatry Research, vol. 304, p. 114135, 2021.
[20] A. Radford, K. Narasimhan, T. Salimans, and I. Sutskever, “Improving language understanding by generative pre-training,” 2018.
[21] A. Vaswani et al., “Attention is all you need,” Advances in neural information processing systems, vol. 30, 2017.
[22] J. Devlin, M.-W. Chang, K. Lee, and K. Toutanova, “Bert: Pre-training of deep bidirectional transformers for language understanding,” arXiv preprint arXiv:1810.04805, 2018.
[23] R. Hadsell, S. Chopra, and Y. LeCun, “Dimensionality reduction by learning an invariant mapping,” in 2006 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR’06), vol. 2, pp. 1735–1742, IEEE, 2006.
[24] G. Koch, R. Zemel, R. Salakhutdinov, et al., “Siamese neural networks for one-shot image recognition,” in ICML deep learning workshop, vol. 2, p. 0, Lille, 2015.
[25] F. Schroff, D. Kalenichenko, and J. Philbin, “Facenet: A unified embedding for face recognition and clustering,” in Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 815–823, 2015.
[26] A. Hermans, L. Beyer, and B. Leibe, “In defense of the triplet loss for person re-identification,” arXiv preprint arXiv:1703.07737, 2017.
[27] P. Neculoiu, M. Versteegh, and M. Rotaru, “Learning text similarity with siamese recurrent networks,” in Proceedings of the 1st Workshop on Representation Learning for NLP, pp. 148–157, 2016.
[28] A. Conneau, D. Kiela, H. Schwenk, L. Barrault, and A. Bordes, “Supervised learning of universal sentence representations from natural language inference data,” arXiv preprint arXiv:1705.02364, 2017.
[29] N. Reimers and I. Gurevych, “Sentence-bert: Sentence embeddings using siamese bert-networks,” arXiv preprint arXiv:1908.10084, 2019.
[30] V. Sanh, L. Debut, J. Chaumond, and T. Wolf, “DistilBERT, a distilled version of BERT: smaller, faster, cheaper and lighter,” arXiv preprint arXiv:1910.01108, 2019.
[31] Y. Liu et al., “Roberta: A robustly optimized bert pretraining approach,” arXiv preprint arXiv:1907.11692, 2019.
[32] M. Sanderson et al., “Test collection based evaluation of information retrieval systems,” Foundations and Trends® in Information Retrieval, vol. 4, no. 4, pp. 247–375, 2010.
[33] I. Soboroff, S. Huang, and D. Harman, “Trec 2018 news track overview.,” in TREC, 2018.
[34] A. Bondarenko, M. Fröbe, M. Beloucif, L. Gienapp, Y. Ajjour, A. Panchenko, C. Biemann, B. Stein, H. Wachsmuth, M. Potthast, et al., “Overview of touché 2020: argument retrieval,” in International Conference of the Cross-Language Evaluation Forum for European Languages, pp. 384–395, Springer, 2020.
[35] E. Voorhees, T. Alam, S. Bedrick, D. Demner-Fushman, W. R. Hersh, K. Lo, K. Roberts, I. Soboroff, and L. L. Wang, “Trec-covid: constructing a pandemic information retrieval test collection,” in ACM SIGIR Forum, vol. 54, pp. 1–12, ACM New York, NY, USA, 2021.
[36] D. Lewandowski, “Evaluating the retrieval effectiveness of web search engines using a representative query sample,” Journal of the Association for Information Science and Technology, vol. 66, no. 9, pp. 1763–1775, 2015.
[37] M. Sanderson and J. Zobel, “Information retrieval system eval- uation: effort, sensitivity, and reliability,” in Proceedings of the 28th annual international ACM SIGIR conference on Research and development in information retrieval, pp. 162–169, 2005.
[38] J. Li, A. Sun, J. Han, and C. Li, “A survey on deep learning for named entity recognition,” IEEE Transactions on Knowledge and Data Engineering, vol. 34, no. 1, pp. 50–70, 2020.
[39] F. Souza, R. Nogueira, and R. Lotufo, “Portuguese named entity recognition using BERT-CRF,” arXiv preprint arXiv:1909.10649, 2019.
[40] M. Wen, D. K. Vasthimal, A. Lu, T. Wang, and A. Guo, “Building large-scale deep learning system for entity recognition in e-commerce search,” in Proceedings of the 6th IEEE/ACM International Conference on Big Data Computing, Applications and Technologies, pp. 149–154, 2019.
[41] X. Cheng, M. Bowden, B. R. Bhange, P. Goyal, T. Packer, and F. Javed, “An end-to-end solution for named entity recognition in ecommerce search,” in Proceedings of the AAAI Conference on Artificial Intelligence, vol. 35, pp. 15098–15106, 2021.
[42] S. Mittal and M. K. Agarwal, “Named entity recognition in local intent web search queries,” in International Conference on Database and Expert Systems Applications, pp. 407–417, Springer, 2019.
[43] ckiplab/ckip-transformers. CKIP Lab, 2022. Accessed: Jun. 16, 2022. [Online]. Available: https://github.com/ckiplab/ckip-transformers
[44] Parker, Robert, et al. “Chinese Gigaword Fifth Edition LDC2011T13,” Philadelphia: Linguistic Data Consortium, 2011. [Online]. Available: https://catalog.ldc.upenn.edu/LDC2011T13
[45] L. Van der Maaten and G. Hinton, “Visualizing data using t-sne.,” Journal of machine learning research, vol. 9, no. 11, 2008.
[46] X. Zhang, C. Zhang, X. Li, X. L. Dong, J. Shang, C. Falout- sos, and J. Han, “Oa-mine: Open-world attribute mining for e-commerce products with weak supervision,” in Proceedings of the ACM Web Conference 2022, pp. 3153–3161, 2022.
[47] H. Xu, W. Wang, X. Mao, X. Jiang, and M. Lan, “Scaling up open tagging from tens to thousands: Comprehension empowered attribute value extraction from product title,” in Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pp. 5214–5223, 2019.
[48] G. Zheng, S. Mukherjee, X. L. Dong, and F. Li, “Opentag: Open attribute value extraction from product profiles,” in Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, pp. 1049–1058, 2018.
[49] J. Johnson, M. Douze, and H. Jégou, “Billion-scale similarity search with gpus,” IEEE Transactions on Big Data, vol. 7, no. 3, pp. 535–547, 2019.