Zhang, Wei

Wei Zhang is a research scientist in NLP, deep learning, reinforcement learning, human computer interaction and speech. His goal is to create the world's best product search and discovery experience using ML and NLP. He has been interested in the E-commerce domain, with the positions of machine learning engineer at Saks Fifith Avenue, and senior staff scientist at Wayfair . Before E-commerce, he was a research scientist at IBM Thomas J. Watson Research Center until 2021. He joined IBM right after his graduation from CMU LTI in 2014.

He has a broad interest in AI. Yet, he recently focuses on the following:
  • Chat Search with large language models and knowledge grounded Chat
  • reinforcement learning and RLHF
  • HCI factors in Chat Search
  • Platform-agnostic integrations of Chat Search into E-commerce sites
  • self-adaptive prompt engineering
  • active learning by prompt engineering
  • Explainability of large language models
  • Multi-agent communication and AI self-learning
  • AI's applications to Finance

News

Dec 2022: Our IEEE SLT 2022 paper achieved the state-of-the-art on our IEMOCAP emotion recognition dataset! The use of Transformers as long-range attention over MFCC inputs is a successful supplement to CNN over LogMel inputs!

Feb 2022: Is BERT-based language models truly generalize in the E-commerce product search domain? Checkout our ACL 2022 Ecommerce for NLP paper.

May 2021: Our ACL 2021 paper identifies issues when applying sample-based explanation methods in NLP, and proposed a solution to enhance the interpretability of explanations that are friendly to long documents.

Feb 2021: We recently proved for the first time that BERT can encode lexical semantics better than GloVe on a intrinsic evaluation using psycholinguistic data, help explaining why pre-trained Transformers are better encoders.

Jan 2020: Our CHI 2020 paper studying human behavior in human-AI communication has been nominated the best paper award!

He recently serves as a PC member for: Interspeech 2023, ACL2022, KDD 2022, ARR May 2021, ACL 2021, EMNLP 2021, ACL 2020, EMNLP 2020, NAACL 2021, EACL 2021, SLT 2021, NeurIPS 2020.


Google Scholar

Selected Publications

[IEEE SLT 2022] Xiaoming Zhang, Fan Zhang, Xiaodong Cui, Wei Zhang. Speech Emotion Recognition with Complementary Acoustic Representations

[ECNLP@ACL 2022] Wei Zhang, et al. Towards Generalizeable Semantic Product Search by Text Similarity Pre-training on Search Click Logs

[ACL 2021] Wei Zhang, Ziming Huang, Yada Zhu, Guangnan Ye, Xiaodong Cui, Fan Zhang. On Sample Based Explanation Methods for NLP: Faithfulness, Efficiency and Semantic Evaluation.

[AAAI 2021] Wei Zhang, Murray Campbell, Yang Yu, Sadhana Kumaravel. Circles are like Ellipses, or Ellipses are like Circles? Measuring the Degree of Asymmetry of Static and Contextual Word Embeddings and the Implications to Representation Learning. [paper]

[NeurIPS 2020] Zahra Ashktorab, James Johnson, Qian Pan, Wei Zhang, Casey Dugan. The Design and Development of Games with a Purpose for AI System. Workshop of Human And Machine in-the-Loop Evaluation and Learning Strategies.

[CHI 2020] Katy Ilonka Gero, Zahra Ashktorab, Casey Dugan, Qian Pan, James Johnson, Werner Geyer, Maria Ruiz, Sarah Miller, David R Millen, Murray Campbell, Sadhana Kumaravel, Wei Zhang. Mental Models of AI Agents in a Cooperative Game Setting. (Best Paper Award)

[AAAI 2018] Shuohang Wang, Mo Yu, Xiaoxiao Guo, Zhiguo Wang, Tim Klinger, Wei Zhang, Shiyu Chang, Gerald Tesauro, Bowen Zhou, Jing Jiang. R $^ 3$: Reinforced Reader-Ranker for Open-Domain Question Answering.

[ICLR 2018] Shuohang Wang, Mo Yu, Jing Jiang, Wei Zhang, Xiaoxiao Guo, Shiyu Chang, Zhiguo Wang, Tim Klinger, Gerald Tesauro, Murray Campbell. Evidence Aggregation for Answer Re-Ranking in Open-Domain Question Answering.

[arxiv 2017] Wei Zhang, Bowen Zhou. Learning to update Auto-associative Memory in Recurrent Neural Networks for Improving Sequence Memorization. arxiv. preprint arXiv:1709.06493 (2017). [paper]

[arxiv 2016] Wei Zhang*, Yang Yu*, Kazi Hasan, Mo Yu, Bing Xiang, Bowen Zhou. Dynamic Chunk Reader for Machine Reading Comprehension arxiv. preprint: arXiv:1610.09996 (2016) (* equal contribution)[paper]

[NIPS 2015] Wei Zhang, Yang Yu, Bowen Zhou. Structured Memory for Neural Turing Machines Reasoning, Memory and Attention Workshop [slides][paper]

[arxiv 2015] Yang Yu, Wei Zhang, Chung-Wei Hang, and Bowen Zhou. Empirical Study on Deep Learning Models for Question Answering. arXiv preprint arXiv:1510.07526 (2015).

[Journal 2014] Wei Zhang, and Judith Gelernter. Geocoding location expressions in Twitter messages: A preference learning method. Journal of Spatial Information Science 2014, no. 9 (2014): 37-70.

[ACM GIR 2013] Judith Gelernter, and Wei Zhang. Cross-lingual geo-parsing for non-structured data. In Proceedings of the 7th Workshop on Geographic Information Retrieval, pp. 64-71. ACM, 2013. [paper]

[SemEval 2007] Yuhang Guo, Wanxiang Che, Yuxuan Hu, Wei Zhang, and Ting Liu. HIT-IR-WSD: A wsd system for english lexical sample task. In Proceedings of the ACL SemEval. (2007). (System won 1st place on SemEval 2007 Task 11)[paper]

Talks

Explainability with Influence Functions ++ and TracIn ++ [Youtube] at ACL'21

"On Machine Reading Comprehension and Question Answering" [slides] at Harvard NLP Reading Group

"Structured Memory for Neural Turing Machines" [slides] on NIPS 2015 RAM workshop