IBM Showcases Key AI Research @ AAAI-20

Share this post:


The IBM T.J. Watson Research Center in Yorktown Heights, New York

The thirty-fourth AAAI Conference on Artificial Intelligence (AAAI) will be held February 7-12 in New York, New York. IBM Research and its scientists have participated in AAAI for decades, but — given this year’s conference is in the city where IBM Research was founded 75 years ago and near where a significant number of our AI researchers live and work at the IBM T.J. Watson Research Center in Yorktown Heights, New York — we’ve made a concerted effort to showcase some of our most strategic work in AI at AAAI-20.

IBM Research will present more than fifty technical papers at the conference, as well a rich set of demos of our latest work, reflecting our focus on key areas of AI research including AutoAI, mastering language, planning, computational argumentation, the future of work and security. We believe progressing these areas of research is essential to developing a broader form of AI that goes beyond today’s narrow, task-specific applications to learn and reason across domains using multiple modalities, while being explainable, secure, and fair.

AutoAI is a novel approach of designing, training and optimizing machine learning models automatically. With AutoAI, anyone could soon build machine learning pipelines from raw data directly, without writing complex code and performing tedious tuning and optimization, automating complicated, labor-intensive tasks. Several IBM papers selected for the AAAI-20 conference in New York demonstrate the value of AutoAI and different approaches to it.

For example, in a research first, IBM researchers created an automated AI tool that enables users to build their own customized, end-to-end machine learning models. This framework eliminates several significant challenges facing enterprise AI users today, including that the models are too time-consuming to build and can’t be customized, and that the data needed is often in a “black box.” Additionally, researchers from the MIT-IBM Watson AI Lab and MIT CSAIL have created a new neural network architecture called EvolveGCN that for the first time enables graph representations of data to change as the information changes, offering new opportunities in areas where tracking data in real-time is critical (e.g., finance, healthcare, and social media.)

Mastering Language
At AAAI, we will also present multiple papers aimed at overcoming one of the biggest obstacles to advancing AI: the technology’s difficulty interacting with people in the natural language they typically use when speaking and writing. Human language is filled with nuance, hidden meaning and context that machines are currently unable to fully comprehend.

One IBM Research paper proposes an improved approach to augment data used to classify text, a crucial piece to training NLP systems. Another describes efforts to improve an NLP system’s ability to reason, through a process known as textual entailment, by complementing training data with information from an external source. Additional work from IBM Research at AAAI-20 tackles the need for NLP systems to learn how words relate to one another and fit into larger categories.

Autonomous agent must be able to plan for its next action in order to function in its environment. This was among the first tasks posed during the formulation of the notion of artificial intelligence more than half a century ago. Yet it remains a largely unsolved core AI competency due to the complexity of the task both in terms of representation as well as reasoning. Research from IBM presented at AAAI-20 explores key aspects of adopting automated planners, including scaling to very large problems, interfacing with end users, and providing guarantees on the quality of solution.

Computational Argumentation


(credit: Visually Attractive for IBM)

IBM Research will also present the latest from Project Debater, including a new end-to-end, high-precision tool and dataset of nearly 30,500 annotated arguments (the largest corpus ever released) to advance argument mining, which helps machines reason and ultimately assemble an argument on a topic. The paper will also reveal how these capabilities work over a massive corpus of ~10B sentences with an improvement in quality since the March 2019 debate going from ~85-90% to 95% precision.

The Future of Work
Knowing exactly how AI will impact employees’ day-to-day tasks remains largely unknown and unstudied, even as the technology continues being adopted in workplaces. That is why researchers from the MIT-IBM Watson AI Lab and MIT Sloan School of Management recently applied a predictive model commonly used for financial forecasting to a dataset of 170 million job postings that uncovered a new way of understanding how occupations and tasks will evolve with AI.

The new model can predict, with high confidence, how tasks may shift within an occupation. The findings will be presented at the Artificial Intelligence, Ethics & Society (AIES) conference as part of AAAI, on February 7.

Homomorphic Encryption (HE) is a security tool introduced in 2009 to help organizations process information while still protecting privacy and security. If encrypted homomorphically, data stays encrypted, preserving privacy and confidentiality — all the time, even when processing the data with AI.

IBM scientists have been improving HE to make it faster and more useable, even recently putting a library in GitHub. At AAAI-20, IBM researchers from the UK will present that they have demonstrated another machine learning tool with HE: Gaussian process regression, one of the most common machine learning algorithms. In the paper, the team demonstrates how the technique could be applied to pharmaceutical data to improve the discovery process across multiple parties without exposing any IP. While dependent on the data set, the model was only seconds slower than if it ran decrypted, which is suitable for such an application.

Where to find us at AAAI-20
These are just some of the papers IBM Research is excited to share and discuss with you at AAAI-20. Please visit us in booth #103 to learn more about our research and meet our scientists, who will be giving demos on some of our emerging technologies, including:

  • Command Line AI Toolkit (CLAI)—aims to bring the power of AI to the shell by augmenting the user experience with natural language support, troubleshooting, and automation, as well as providing researchers with an easily extensible API to develop their own AI plugins.
  • Go Ahead Ask Me Anything (GAAMA)—offers a (multilingual) reading comprehension for question answering.
  • AI automation for visual inspection—enables civil engineers to efficiently create and manage AI models for visual inspection of infrastructure like bridges.
  • Data quality for AI—demonstrates how AI techniques can be used for AI readiness evaluation.
  • COnversational assistant for business Process Automation (COPA)—a slackbot agent created for digital process automation on a travel use case.
  • Language-model-based data augmentation (LAMBADA)—provides a new data augmentation method for text classification tasks.
  • Causal knowledge extraction through large-scale text mining—automatically extracts causal statements from text that answer questions in the form “Could X cause Y?,” where X and Y are phrases without any semantic constraints.
  • Doc2Dial—offers a framework for dialogue composition grounded in documents.
  • TraceHub—acts as a platform to bridge the gap between state-of-the-art, time-series analytics and datasets.
  • Human-in-the-loop linguistic Expressions wIth Deep Learning (HEIDL)— enables human-machine co-creation of explainable AI models by learning rich and complex linguistic expressions in first-order logic that may be modified by domain experts to improve generalization.
  • Proof of Analog Results Through Numerically Equivalent Routine (PARTNER)—works as a
    human-in-the-loop deep learning system that identifies entity name structures with minimal human effort and allows users to construct complex normalization and variant generation functions without coding skills.

At our booth, you can also learn more about career opportunities, including internships and residencies We look forward to seeing you in New York!

Poster/Demo Reception 1
Americas Hall I / II
Sunday, February 9
7:30 – 9:30PM

DEMO388: PARTNER: Human-in-the-Loop, Entity Name Understanding with Deep Learning
Kun Qian, Poornima Chozhiyath Raman, Yunyao Li, Lucian Popa

DEMO425: Causal Knowledge Extraction through Large-Scale Text Mining
Oktie Hassanzadeh, Debarun Bhattacharjya, Mark Feblowitz, Kavitha Srinivas, Michael Perrone, Shirin Sohrabi, Michael Katz

Poster Spotlight Presentation 4483: Sunday, February 9 | 3:45-5:15 PM, Beekman
APP4483: Reinforcement Learning-Based Portfolio Management with Augmented Asset Movement Prediction States
Yunan Ye, Hengzhi Pei, Boxin Wang, Pin-¬Yu Chen, Yada Zhu, Jun Xiao, Bo L

Poster Spotlight Presentation 4483: Sunday, February 9 | 2:00-3:15 PM, Beekman
APP5592: CASTER: Predicting Drug Interactions with Chemical Substructure Representation
Kexin Huang, Cao Xiao, Trong Nghia Hoang, Lucas M. Glass, Jimeng Sun

Oral Presentation 1698: Sunday, February 9 | 3:45-5:15 PM, Clinton
HAI1698: Towards Socially Responsible AI: Cognitive Bias-Aware Multi-Objective Learning
Procheta Sen, Debasis Ganguly

Poster Spotlight Presentation 175: Sunday, February 9 | 3:45-5:15 PM, Concourse A
ML175: Characterizing Membership Privacy in Stochastic Gradient Langevin Dynamics
Bingzhe Wu, Chaochao Chen, Shiwan Zhao, Cen Chen, Yuan Yao, Guangyu Sun, Li Wang, Xiaolu Zhang, Jun Zhou

Poster Spotlight Presentation 1768: Sunday, February 9 | 3:45-5:15 PM, Murray Hill
ML1768: Privacy-Preserving Gaussian Process Regression –A Modular Approach to the Application of Homomorphic Encryption
Peter Fenner, Edward O. Pyzer-Knapp

Poster Spotlight Presentation 1768: Sunday, February 9 | 11:15-12:30 PM, Murray Hill
ML2932: Towards Certificated Model Robustness Against Weight Perturbations
TsuiWei Weng, Pu Zhao, Sijia Liu, Pin-Yu Chen, Xue Lin, Luca Daniel

Poster Spotlight Presentation 4305: Sunday, February 9 | 3:45-5:15 PM, Concourse A
ML4305: Event-Driven Continuous Time Bayesian Networks
Debarun Bhattacharjya, Karthikeyan Shanmugam, Tian Gao, Nicholas Mattei, Kush R. Varshney, Dharmashankar Subramanian

Poster Spotlight Presentation 5847: Sunday, February 9 | 11:15-12:30 PM, Gramercy
ML5847: CAG: A Real-Time Low-Cost Enhanced-Robustness High-Transferability Content-Aware Adversarial Attack Generator
Huy Phan, Yi Xie, Siyu Liao, Jie Chen, Bo Yuan

Poster Spotlight Presentation 8794: Sunday, February 9 | 9:30-10:45 AM, Concourse A
ML8794: Sanity Checks for Saliency Metrics
Richard Tomsett, Dan Harborne, Supriyo Chakraborty, Prudhvi Gurram, Alun Preece

Poster Spotlight Presentation 4347: Sunday, February 9 | 3:45-5:15 PM, Trianon
NLP4347: Embedding Compression with Isotropic Iterative Quantization
Siyu Liao, Jie Chen, Yanzhi Wang, Qinru Qiu, Bo Yuan

Poster Spotlight Presentation 5584: Sunday, February 9 | 3:45-5:15 PM, Sutton North
NLP5584: A Large- Scale Dataset for Argument Quality Ranking: Construction and Analysis
Shai Gretz, Roni Friedman, Edo Cohen-Karlik, Assaf Toledo, Dan Lahav, Ranit Aharonov, Noam Slonim

Poster Spotlight Presentation 7478: Sunday, February 9 | 9:30-10:45 AM, Clinton
RU7478: Parallel AND/OR Search for Marginal MAP
Radu Marinescu, Akihiro Kishimoto, Adi Botea

Poster/Demo Reception 2
Americas Hall I / II
Monday, February 10
7:20 – 9:20 PM

Not Participating in Poster/Demo Reception
Oral Presentation 5922: Monday, February 10 | 11:15-12:30 PM, Gibson
5922: Reshaping Diverse Planning
Michael Katz, Shirin Sohrabi

Poster Spotlight Presentation 9423: Monday, February 10 | 3:45-5:15 PM, Trianon
AIW9423: Infusing Knowledge into the Textual Entailment Task Using Graph Convolutional Networks
Pavan Kapanipathi, VeronikaThost, Siva Sankalp Patel, Spencer Whitehead, Ibrahim Abdelaziz, Avinash Balakrishnan, Maria Chang, Kshitij Fadnis, Chulaka Gunasekara, Bassem Makni, Nicholas Mattei, Kartik Talamadupula, Achille Fokoue

Poster Spotlight Presentation 4406: Monday, February 10 | 3:45-5:15 PM, Madison
HAI4406: GaSPing for Utility
Mengyang Gu, Debarun Bhattacharjya, Dharmashankar Subramanian

Poster Spotlight Presentation 1980: Monday, February 10 | 9:30-10:45 AM, Concourse A
ML1980: An ADMM Based Framework for AutoML Pipeline Configuration
Sijia Liu, Parikshit Ram, Deepak Vijaykeerthy, Djallel Bouneffouf, Gregory Bramble, Horst Samulowitz, Dakuo Wang, Andrew Conn, Alexander Gray

Poster Spotlight Presentation 2046: Monday, February 10 | 11:15-12:30 PM, Murray Hill
ML2046: Joint Modeling of Local and Global Temporal Dynamics for Multivariate Time Series Forecasting with Missing Values
Xianfeng Tang, Huaxiu Yao, Yiwei Sun, Charu Aggarwal, Prasenjit Mitra, Suhang Wang

Poster Spotlight Presentation 2928: Monday, February 10 | 3:45-5:15 PM, Concourse A
ML2928: Towards Query-Efficient Black Box Adversary with Zeroth-Order Natural Gradient Descent
Pu Zhao, Pin-Yu Chen, Siyue Wang, Xue Li

Poster Spotlight Presentation 3684: Monday, February 10 | 2:00-3:15 PM, Regent
ML3684: Towards Fine-Grained Temporal Network Representation via Time-Reinforced Random Walk
Zhining Liu, Dawei Zhou, Yada Zhu, Jinjie Gu, Jingrui He

Poster Spotlight Presentation 4837: Monday, February 10 | 2:00-3:15 PM, Gibson
ML4837: Fatigue-Aware Bandits for Dependent Click Models
Junyu Cao, Wei Sun, Zuo-Jun (Max) Shen, Markus Ettl

Poster Spotlight Presentation 5585: Monday, February 10 | 3:45-5:15 PM, Concourse A
ML5585: Seq2Sick: Evaluating the Robustness of Sequence-to-Sequence Models with Adversarial Examples
Minhao Cheng, Jinfeng Yi, Pin-Yu Chen, Huan Zhang, Cho-Jui Hsieh

Poster Spotlight Presentation 5679: Monday, February 10 | 3:45-5:15 PM, Gramercy
ML5679: EvolveGCN: Evolving Graph Convolutional Networks for Dynamic Graphs
Aldo Pareja, Giacomo Domeniconi, Jie Chen, Tengfei Ma, Toyotaro Suzumura, Hiroki Kanezashi, Tim Kaler, Tao B. Schardl, Charles E. Leiserson

Oral Presentation 5986: Monday, February 10 | 9:30-10:45 AM, Murray Hill
ML5986: Pursuit of Low-Rank Models of Time-Varying Matrices Robust to Sparse and Measurement Noise
Albert Akhriev, Jakub Marecek, Andrea Simonetto

Poster Spotlight Presentation 6005: Monday, February 10 | 3:45-5:15 PM, Regent
ML6005: A Multi-Channel Neural Graphical Event Model with Negative Evidence
Tian Gao, Dharmashankar Subramanian, Karthikeyan Shanmugam, Debarun Bhattacharjya, Nicholas Mattei

Poster Spotlight Presentation 6879: Monday, February 10 | 11:15-12:30 PM, Murray Hill
ML6879: Building Calibrated Deep Models via Uncertainty Matching with Auxiliary Interval Predictors
Jayaraman J. Thiagarajan, Bindya Venkatesh, Prasanna Sattigeri, Peer-Timo Bremer

Poster Spotlight Presentation 1741: Monday, February 10 | 3:45-5:15 PM, Sutton North
NLP1741: Corpus Wide Argument Mining – a Working Solution
Liat Ein-Dor, Eyal Shnarch, Lena Dankin, Alon Halfon, Benjamin Sznajder, Ariel Gera, Carlos Alzate, Martin Gleize, Leshem Choshen, Yufang Hou, Yonatan Bilu, Ranit Aharonov, Noam Slonim

Poster Spotlight Presentation 4027: Monday, February 10 | 3:45-5:15 PM, Trianon
NLP4027: Do Not Have Enough Data? Deep Learning to the Rescue!
Ateret Anaby-Tavor, Boaz Carmeli, Esther Goldbraich, Amir Kantor, George Kour, Segev Shlomov, Naama Tepper, Naama Zwerdling

Poster Spotlight Presentation 7162: Monday, February 10 | 3:45-5:15 PM, Trianon
NLP7162: Multi-Label Patent Categorization with Non Local Attention Based Graph Convolutional Network
Pingjie Tang, Meng Jiang, Bryan (Ning) Xia, Jed W. Pitera, Jeffrey Welser, Nitesh V. Chawla

Poster Spotlight Presentation 7266: Monday, February 10 | 9:30-10:45 AM, Trianon
NLP7266: End-to-End Argumentation Knowledge Graph Construction
Khalid Al-Khatib, Yufang Hou, Henning Wachsmuth, Charles Jochim, Francesca Bonin, Benno Stein

Oral Presentation 8479: Monday, February 10 | 11:15-12:30 PM, Trianon
NLP8479: Mask & Focus: Conversation Modelling by Learning Concepts
Gaurav Pandey, Dinesh Raghu, Sachindra Josh

Oral Presentation 5664: Monday, February 10 | 2:00-3:15 PM, Gibson
PRS5664: Online Planner Selection with Graph Neural Networks and Adaptive Scheduling
Tengfei Ma, Patrick Ferber, Siyu Huo, Jie Chen, Michael Katz

Oral Presentation 5932: Monday, February 10 | 11:15-12:30 PM, Gibson
 Top-Quality Planning: Finding Practically Useful Sets of Best Plans
Michael Katz and Shirin Sohrabi and Octavian Udrea

Poster/Demo Reception 3
Americas Hall I / II
Tuesday, February 11
6:30 – 8:30 PM

DEMO541: Doc2Dial: A Framework for Dialogue Composition Grounded in Documents
Song Feng, Kshitij Fadnis, Q. Vera Liao, Luis A. Lastras

DEMO556: TraceHub – A Platform to Bridge the Gap between State-¬of-¬the-¬Art Time-¬Series Analytics and Datasets
Shubham Agarwal, Christian Muise, Mayank Agarwal, Sohini Upadhyay, Zilu Tang, Zhongshen Zeng, Yasaman Khazaeni

Poster Spotlight Presentation 9334: Tuesday, February 11 | 9:30-10:45 AM, Clinton
CSO9334: Delay-Adaptive Distributed Stochastic Optimization
Zhaolin Ren, Zhengyuan Zhou, Linhai Qiu, Ajay Deshpande, Jayant Kalagnanam

Poster Spotlight Presentation 1680: Tuesday, February 11 | 9:30-10:45 AM, Regent
ML1680: Uncorrected Least-Squares Temporal Difference with Lambda-Return
Takayuki Osogami

Poster Spotlight Presentation 5350: Tuesday, February 11 | 2:00-3:30 PM, Sutton North
ML5350: Guiding Attention in Sequence-to-Sequence Models for Dialogue Act Prediction
Pierre Colombo, Emile Chapuis, Matteo Manica, Emmanuel Vignon, Giovanna Varni, Chloe Clavel

Poster Spotlight Presentation 6994: Tuesday, February 11 | 11:15-12:30 PM, Murray Hill
ML6694: On the Role of Weight Sharing During Deep Option Learning
Matthew Riemer, Ignacio Cases, Clemens Rosenbaum, Miao Liu, Gerald Tesauro

Poster Spotlight Presentation 7690: Tuesday, February 11 | 9:30-1045 AM, Concourse A
ML7690: Weighted Sampling for Combined Model Selection and Hyperparameter Tuning
Dimitrios Sarigiannis, Thomas Parnell, Haralampos Pozidis

Oral Presentation 9089: Wednesday, February 12 | 2:00-3:20 PM, Murray Hill
ML9089: Generalizable Resource Allocation in Stream Processing via Deep Reinforcement Learning
Xiang Ni, Jing Li, Mo Yu, Wang Zhou, Kun-Lung Wu

Poster Spotlight Presentation 3461: Tuesday, February 11 | 9:30-1045 AM, Sutton South
NLP3461: Modeling Dialogues with Hashcode Representations: A Nonparametric Approach
Sahil Garg, Irina Rish, Guillermo Cecchi, Palash Goyal, Sarik Ghazarian, Shuyang Gao, Greg Ver Steeg, Aram Galstyan

Poster Spotlight Presentation 8319: Tuesday, February 11 | 9:30-10:45 AM, Sutton North
NLP8319: Translucent Answer Predictions in Multi-Hop Reading Comprehension
G P Shrivatsa Bhargav, Michael Glass, Dinesh Garg, Shirish Shevade, Saswati Dana, Dinesh Khandelwal, L Venkata Subramaniam, Alfio Gliozzo

Oral Presentation 9092: Wednesday, February 12 | 11:30-12:30 PM, Sutton South
NLP9092: Hypernym Detection Using Strict Partial Order Networks
Sarthak Dash, Md Faisal Mahbub Chowdhury, Alfio Gliozzo, Nandana Mihindukulasooriya, Nicolas Rodolfo Fauceglia

Oral Presentation 1813: Tuesday, February 11 | 9:30-10:45 AM, Gibson
PRS1813: Expectation-Aware Planning: A Unifying Framework for Synthesizing and Executing Self-Explaining Plans for Human-Aware Planning
Sarath Sreedharan, Tathagata Chakraborti, Christian Muise, Subbarao Kambhampati

Poster Spotlight Presentation 5371: Tuesday, February 11 | 2:00-3:30 PM, Nassau
VIS5371: Location-Aware Graph Convolutional Networks for Video Question Answering
Deng Huang, Peihao Chen, Runhao Zeng, Qing Du, Mingkui Tan, Chuang Gan

Poster Spotlight Presentation 8881: Tuesday, February 11 | 2:00-3:30 PM, Gramercy
VIS8881: Crowd Counting with Decomposed Uncertainty
Min hwan Oh, Peder A. Olsen, Karthikeyan Natesan Ramamurthy

IBM Fellow, IBM Research AI and Future of Computing, IBM Research

More AI stories

We’ve moved! The IBM Research blog has a new home

In an effort better integrate the IBM Research blog with the IBM Research web experience, we have migrated to a new landing page:

Continue reading

Pushing the boundaries of human-AI interaction at IUI 2021

At the 2021 virtual edition of the ACM International Conference on Intelligent User Interfaces (IUI), researchers at IBM will present five full papers, two workshop papers, and two demos.

Continue reading

From HPC Consortium’s success to National Strategic Computing Reserve

Founded in March 2020 just as the pandemic’s wave was starting to wash over the world, the Consortium has brought together 43 members with supercomputing resources. Private and public enterprises, academia, government and technology companies, many of whom are typically rivals. “It is simply unprecedented,” said Dario Gil, Senior Vice President and Director of IBM Research, one of the founding organizations. “The outcomes we’ve achieved, the lessons we’ve learned, and the next steps we have to pursue are all the result of the collective efforts of these Consortium’s community.” The next step? Creating the National Strategic Computing Reserve to help the world be better prepared for future global emergencies.

Continue reading