Deep Neural Networks for Natural Language Processing

    Sergey works part-time as a senior applied research scientist at AI2, on the Semantic Scholar research team. He's worked on many different projects, including:

    • A paper about gender bias in clinical trial recruitment published in JAMA Network Open, along with news coverage.
    • A complete overhaul of the Semantic Scholar author disambiguation system, described in a published paper and a blog post. Also, see the open-sourced code & data
    • Two published methods for high quality academic paper embeddings: Citeomatic (code) and SPECTER (code).
    • Improving the Semantic Scholar search engine, described in a detailed blog post. Code is available as well.
    • A blog post and paper about the association between posting your papers on ArXiV before review and subsequent citations.


    This work is ongoing since March, 2016.

    Machine Learning Strategy Consulting

    The Healthy Birth, Growth, and Development (HBGD) program was launched in 2013 by the Bill & Melinda Gates Foundation.


    The Knowledge Integration (Ki) initiative aims facilitates collaboration between researchers, quantitative experts, and policy makers in fields related to HBGD. The broad goal is to aggregate data from past longitudinal studies about pathways and risk factors that affect birth, growth, and neurocognitive development in order to better predict Ki outcomes.


    Sergey works closely with Ki leadership - designing and overseeing data science contests; managing external collaborations with academic research labs and software companies; and modeling many diverse global health datasets (an example is described here).


    This work is ongoing since February, 2015.

    Open Source Contributions

    For: Everyone

    We contribute to the Python data science ecosystem.


    Most notably, Sergey co-wrote and maintains the imputation package fancyimpute, and merged IterativeImputer into the machine learning uber-library scikit-learn. Some other packages we've worked on:

    • https://github.com/allenai/S2AND/
    • https://github.com/allenai/s2_fos
    • https://github.com/allenai/specter/
    • https://github.com/allenai/scidocs/
    • https://github.com/allenai/s2search/
    • https://github.com/sergeyf/SmallDataBenchmarks/
    • https://github.com/allenai/citeomatic/

    Improving Reading Comprehension


    Actively Learn makes a reading tool that enables teachers to guide, monitor, and improve student learning. With our help, they wrote and were awarded an NSF SBIR grant to answer the key question: "How can we personalize reading instruction so as to increase comprehension & learning?" We are diving deep into the data with sophisticated machine learning tools, and bringing back testable hypotheses about what helps and hinders students.


    This work is ongoing since April, 2014.

    Contributing to Technical Books


    Jenny Dearborn, Chief Learning Officer and Senior Vice President at SAP, has written Data Driven, a "practical guide to increasing sales success, using the power of data analytics," and The Data Driven Leader (with David Swanson), "a clear, accessible guide to solving important leadership challenges through human resources-focused and other data analytics."


    We helped her and her team come up with clear and compelling ways to communicate the deep mathematical models that are at the core of the book, as well as contributed to the plot and characterizations.

    Pro Bono Data Science


    Seattle Against Slavery mobilizes the community in the fight against labor and sex trafficking through education, advocacy, and collaboration with local and national partners. We are proud to provide them with analytics and statistics services on a volunteer basis.

    Multiple Projects


    Long Tail NLP-Based Recommendations. Most e-commerce recommendation engines have difficulty highlighting less frequently bought products, which is an issue that compounds itself and ends up recommending the same popular products over and over. We developed a language-based model for RichRelevance that identifies good recommendations based on comparisons of the product descriptions and description metadata rather than purchase data. This evens the playing field between newer products and the old standbys, so the recommendations have more variety and are generally more applicable.


    Bayesian A/B Testing. RichRelevance swears by their top-notch recommendations. But what's the right way to measure their efficacy? Sergey put together an intuitive, comprehensive Bayesian A/B testing system that works for any KPI, and can provide direct answers to key customer questions like "What is the probability that algorithm A has at least 5% lift over algorithm B?


    Read all about this work in Sergey's three (archived) blog posts: [1], [2], and [3].


    Bandits for Online Recommendations. The most important piece of RichRelevance's impressive big data pipeline is their core recommendation system. It serves thousands of recommendations every minute, and it has to learn quickly from new data. Working with their analytics team, Sergey engineered a modern bandit-based approach to online recommendations that learns from less data, adapts easily to any optimization metric, and does not compromise quality at production-scale.


    Three (now archived) blog posts describe the results of our research: [1], [2], and [3].


    Preva Group

    Preva Group is dedicated to helping organizations achieve large scale social change by combining existing structured and unstructured data, powered by sophisticated analytics and machine learning, delivered through simple user-centered interfaces. Data Cowboys operates a strategic employee sharing partnership with Preva Group.

  • Publications

    Journal Papers

    • Michael Cafarella, Michael Anderson, Iz Beltagy, Arie Cattan, Sarah Chasins, Ido Dagan, Doug Downey, Oren Etzioni, Sergey Feldman, Tian Gao, Tom Hope, Kexin Huang, Sophie Johnson, Daniel King, Kyle Lo, Yuze Lou, Matthew Shapiro, Dinghao Shen, Shivashankar Subramanian, Lucy Lu Wang, Yuning Wang, Yitong Wang, Daniel S. Weld, Jenny Vo-Phamhi, Anna Zeng, and Jiayun Zou, "Infrastructure for rapid open knowledge network development", AI Magazine 43: 59–68, 2022. [pdf]
    • Sean MacAvaney, Sergey Feldman, Nazli Goharian, Doug Downey, and Arman Cohan, "ABNIRML: Analyzing the Behavior of Neural IR Models," Transactions of the Association for Computational Linguistics, 2022. [pdf]
    • Sergey Feldman, Waleed Ammar, Kyle Lo, Elly Trepman, Madeleine van Zuylen, and Oren Etzioni, "Quantifying Sex Bias in Clinical Studies at Scale With Automated Data Extraction," JAMA Network Open, 2019. [link]
    • Sergey Feldman, Maya R. Gupta, and Bela A. Frigyik, "Revisiting Stein's Paradox: Multi-Task Averaging," Journal of Machine Learning Research, 2014. [link]
    • Eric K. Garcia, Sergey Feldman, Maya R. Gupta, and Santosh Srivastava, "Completely Lazy Learning," IEEE Trans. on Knowledge and Data Engineering, 2010. [pdf] [code + data]
    • Vagisha Sharma, Jimmy K. Eng, Sergey Feldman, Priska von Haller, Michael J. MacCoss, and William S. Noble, "Precursor Charge State Prediction for Electron Transfer Dissociation Tandem Mass Spectra," Journal of Proteome Research, 2010. [pdf]

    Conference Papers

    • Shivashankar Subramanian, Daniel King, Doug Downey, and Sergey Feldman, "S2AND: A Benchmark and Evaluation System for Author Name Disambiguation," JCDL, 2021. [pdf] [code] [blog]
    • Asia J. Biega, Fernando Diaz, Michael D. Ekstrand, Sergey Feldman, Sebastian Kohlmeier, "Overview of the TREC 2020 Fair Ranking Track," TREC 2020. [pdf]
    • Sean MacAvaney, Andrew Yates, Sergey Feldman, Doug Downey, Arman Cohan, and Nazli Goharian, "Simplified Data Wrangling with ir_datasets," SIGIR, 2021. [pdf] [code]
    • Arman Cohan, Sergey Feldman, Iz Beltagy, Doug Downey, and Daniel S. Weld, "SPECTER: Document-level Representation Learning using Citation-informed Transformers,"  ACL, 2020. [link] [code] [data]
    • Chandra Bhagavatula, Sergey Feldman, Russell Power, and Waleed Ammar, "Content-Based Citation Recommendation," NAACL-HLT, 2018. [pdf
    • Waleed Ammar, Dirk Groeneveld, Chandra Bhagavatula, Iz Beltagy, Miles Crawford, Doug Downey, Jason Dunkelberger, Ahmed Elgohary, Sergey Feldman, Vu Ha, Rodney Kinney, Sebastian Kohlmeier, Kyle Lo, Tyler Murray, Hsu-Han Ooi, Matthew Peters, Joanna Power, Sam Skjonsberg, Lucy Lu Wang, Chris Wilhelm, Zheng Yuan, Madeleine van Zuylen, and Oren Etzioni, "Construction of the Literature Graph in Semantic Scholar," NAACL-HLT, 2018. [pdf]
    • Sergey Feldman, Maya R. Gupta, and Bela A. Frigyik, "Multi-Task Averaging," NIPS, 2012. [pdf]
    • Luca Cazzanti, Sergey Feldman, Maya R. Gupta, and Michael Gabbay, "Multi-Task Regularization of Generative Similarity Models," Lecture Notes in Computer Science, 2011. [pdf]
    • Sergey Feldman, Marius A. Marin, Mari Ostendorf, and Maya R. Gupta, "Part-of-Speech Histogram Features for Genre Classification of Text," IEEE ICASSP, 2009. [pdf]
    • Marius A. Marin, Sergey Feldman, Mari Ostendorf, and Maya R. Gupta, "Filtering Web Text to Match Target Genres," IEEE ICASSP, 2009. [pdf]
    • Sergey Feldman, Marius A. Marin, Julie Medero, and Mari Ostendorf, "Classifying Factored Genres with Part-of-Speech Histograms," NAACL-HLT, 2009. [pdf]


    Theses & Technical Reports

    • Sergey Feldman, Kyle Lo, and Waleed Ammar, "Citation Count Analysis for Papers with Preprints," 2018. [pdf]
    • Sergey Feldman, "Multi-Task Averaging: Theory and Practice," University of Washington PhD Thesis, 2012. [pdf]
    • Sergey Feldman, Barbara Frewen, Michael J. MacCoss, and Maya R. Gupta, "Filtering Tandem Mass Spectra for Quality," University of Washington Dept. of Electrical Engineering Technical Report UWEETR-2012-0001, 2012. [pdf] [code + data]
All Posts