Tenghao Huang, Kinjal Basu, Ibrahim Abdelaziz, Pavan Kapanipathi, Jonathan May, Muhao Chen. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

Kinjal Basu, Ibrahim Abdelaziz, Kiran Kate, Mayank Agarwal, Maxwell Crouse, Yara Rizk, Kelsey Bradford, Asim Munawar, Sadhana Kumaravel, Saurabh Goyal, Xin Wang, Luis A. Lastras, Pavan Kapanipathi. Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing. 2025.

Kinjal Basu, Ibrahim Abdelaziz, Subhajit Chaudhury, Soham Dan, Maxwell Crouse, Asim Munawar, Vernon Austel, Sadhana Kumaravel, Vinod Muthusamy, Pavan Kapanipathi, Luis Lastras. Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2024.

Subhajit Chaudhury, Sarathkrishna Swaminathan, Daiki Kimura, Prithviraj Sen, Keerthiram Murugesan, Rosario Uceda-Sosa, Michiaki Tatsubori, Achille Fokoue, Pavan Kapanipathi, Asim Munawar, Alexander Gray. Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2023.

Maxwell Crouse, Pavan Kapanipathi, Subhajit Chaudhury, Tahira Naseem, Ramon Fernandez Astudillo, Achille Fokoue, Tim Klinger. Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2023.

Traditional automated theorem provers have relied on manually tuned heuristics to guide how they perform proof search. Recently, however, there has been a surge of interest in the design of learning mechanisms that can be integrated into theorem provers to improve their performance automatically. In this work, we describe TRAIL (Trial Reasoner for AI that Learns), a deep learning-based approach to theorem proving that characterizes core elements of saturation-based theorem proving within a neural framework. TRAIL leverages (a) an effective graph neural network for representing logical formulas, (b) a novel neural representation of the state of a saturation-based theorem prover in terms of processed clauses and available actions, and (c) a novel representation of the inference selection process as an attention-based action policy. We show through a systematic analysis that these components allow TRAIL to significantly outperform previous reinforcement learning-based theorem provers on two standard benchmark datasets (up to 36% more theorems proved). In addition, to the best of our knowledge, TRAIL is the first reinforcement learning-based approach to exceed the performance of a state-of-the-art traditional theorem prover on a standard theorem proving benchmark (solving up to 17% more theorems).

Subhajit Chaudhury, Sarathkrishna Swaminathan, Chulaka Gunasekara, Maxwell Crouse, Srinivas Ravishankar, Daiki Kimura, Keerthiram Murugesan, Ramón Fernandez Astudillo, Tahira Naseem, Pavan Kapanipathi, Alexander Gray. Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing. 2022.

Prithviraj Sen, Breno William Carvalho, Ibrahim Abdelaziz, Pavan Kapanipathi, Salim Roukos, Alexander Gray. Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing. 2022.

Knowledge Base Question Answering (KBQA) is a task where existing techniques have faced significant challenges, such as the need for complex question understanding, reasoning, and large training datasets. In this work, we demonstrate Deep Thinking Question Answering (DTQA), a semantic parsing and reasoning-based KBQA system. DTQA (1) integrates multiple, reusable modules that are trained specifically for their individual tasks (e.g. semantic parsing, entity linking, and relationship linking), eliminating the need for end-to-end KBQA training data; (2) leverages semantic parsing and a reasoner for improved question understanding. DTQA is a system of systems that achieves state-of-the-art performance on two popular KBQA datasets.

Automated theorem provers have traditionally relied on manually tuned heuristics to guide how they perform proof search. Deep reinforcement learning has been proposed as a way to obviate the need for such heuristics, however, its deployment in automated theorem proving remains a challenge. In this paper we introduce TRAIL, a system that applies deep reinforcement learning to saturation-based theorem proving. TRAIL leverages (a) a novel neural representation of the state of a theorem prover and (b) a novel characterization of the inference selection process in terms of an attention-based action policy. We show through systematic analysis that these mechanisms allow TRAIL to significantly outperform previous reinforcement-learning-based theorem provers on two benchmark datasets for first-order logic automated theorem proving (proving around 15% more theorems).

Knowledge graphs (KGs) are of great importance to many real world applications, but they generally suffer from incomplete information in the form of missing relations between entities. Knowledge graph completion (also known as relation prediction) is the task of inferring missing facts given existing ones. Most of the existing work is proposed by maximizing the likelihood of observed instance-level triples. Not much attention, however, is paid to the ontological information, such as type information of entities and relations. In this work, we propose a type-augmented relation prediction (TaRP) method, where we apply both the type information and instance-level information for the relation prediction. In particular, type information and instance-level information are encoded as prior probabilities and likelihoods of relations respectively, and are combined by following the Bayes' rule. Our proposed TaRP method achieves significantly better performance than state-of-the-art methods on four benchmark datasets: FB15K, FB15K-237, YAGO26K-906, and DB111K-174. In addition, we show that the TaRP achieves the significantly improved data efficiency. More importantly, the type information extracted from a specific dataset can generalize well to different datasets through the proposed TaRP model.

Text-based games have emerged as an important test-bed for Reinforcement Learning (RL) research, requiring RL agents to combine grounded language understanding with sequential decision making. In this paper, we examine the problem of infusing RL agents with commonsense knowledge. Such knowledge would allow agents to efficiently act in the world by pruning out implausible actions, and to perform look-ahead planning to determine how current actions might affect future world states. We design a new text-based gaming environment called TextWorld Commonsense (TWC) for training and evaluating RL agents with a specific kind of commonsense knowledge about objects, their attributes, and affordances. We also introduce several baseline RL agents which track the sequential context and dynamically retrieve the relevant commonsense knowledge from ConceptNet. We show that agents which incorporate commonsense knowledge in TWC perform better, while acting more efficiently. We conduct user-studies to estimate human performance on TWC and show that there is ample room for future improvement.

Anshuman Mishra, Dhruvesh Patel, Aparna Vijayakumar, Xiang Lorraine Li, Pavan Kapanipathi, Kartik Talamadupula. Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies. 2021.

Textual entailment is a fundamental task in natural language processing. Most approaches for solving this problem use only the textual content present in training data. A few approaches have shown that information from external knowledge sources like knowledge graphs (KGs) can add value, in addition to the textual content, by providing background knowledge that may be critical for a task. However, the proposed models do not fully exploit the information in the usually large and noisy KGs, and it is not clear how it can be effectively encoded to be useful for entailment. We present an approach that complements text-based entailment models with information from KGs by (1) using Personalized PageRank to generate contextual subgraphs with reduced noise and (2) encoding these subgraphs using graph convolutional networks to capture the structural and semantic information in KGs. We evaluate our approach on multiple textual entailment datasets and show that the use of external knowledge helps the model to be robust and improves prediction accuracy. This is particularly evident in the challenging BreakingNLI dataset, where we see an absolute improvement of 5-20% over multiple text-based entailment models.

Natural Language Inference (NLI) is fundamental to many Natural Language Processing (NLP) applications including semantic search and question answering. The NLI problem has gained significant attention due to the release of large scale, challenging datasets. Present approaches to the problem largely focus on learning-based methods that use only textual information in order to classify whether a given premise entails, contradicts, or is neutral with respect to a given hypothesis. Surprisingly, the use of methods based on structured knowledge – a central topic in artificial intelligence – has not received much attention vis-a-vis the NLI problem. While there are many open knowledge bases that contain various types of reasoning information, their use for NLI has not been well explored. To address this, we present a combination of techniques that harness external knowledge to improve performance on the NLI problem in the science questions domain. We present the results of applying our techniques on text, graph, and text-and-graph based models; and discuss the implications of using external knowledge to solve the NLI problem. Our model achieves close to state-of-the-art performance for NLI on the SciTail science questions dataset.

Hierarchical relationships play a key role in knowledge graphs. Particularly, large and well-known knowledge graphs such as DBpedia contain significant number of facts expressed with hierarchical relationships in comparison to the other types of relationships. These hierarchical relationships are extensively harnessed by applications such as personalization, question answering, and recommendation systems. However, the presence of large number of facts with hierarchical relationships makes the applications computationally intensive. Additionally, the applications can be domain-specific and may not require all the hierarchical facts available, but only require those that are specific to the domain. In this paper, we present an approach to extract domain-specific hierarchical subgraph from large knowledge graphs by identifying the domain-specificity of the categories in the hierarchy. Given a domain, the domain-specificity of categories are determined by combining different types of evidence using a probabilistic framework. We show the effectiveness of our approach with a recommendation use case for movie and book domains. Our evaluation demonstrates that the domain-specific hierarchical subgraphs extracted by our approach can reduce the baseline subgraph by 40% to 50% without compromising the accuracy of the recommendations. Furthermore, the presented approach outperforms the recommendation results obtained with a state-of-the-art domain-specific subgraph extraction technique which uses supervised learning.

Applications on the Web such as search engines and recommendation systems are increasingly adapting semantic approaches by leveraging knowledge graphs. While some applications require processing of the whole knowledge graph, most are domain-specific and require only a relevant subset of it. For example, a movie or a book recommendation system would require a subgraph that comprises knowledge relevant to the specific domain. In such scenarios, processing the whole knowledge graph, particularly the commonly used, large, and openly available knowledge graphs on the Web, is computationally intensive and the irrelevant portion may negatively impact the performance of the application. This necessitates the identification and extraction of relevant subgraphs that adequately captures entities and their relationships for a given application domain and/or task. In this work, we present an approach to identify a minimal domain-specific subgraph by utilizing statistic and semantic-based metrics. Our approach highlights the importance of relationships as first-class elements to capture the domain specificity of a subgraph. We demonstrate the applicability of this approach for a recommendation use case on two domains, i.e. movie and book. Our evaluation demonstrates a reduction of 80% to 90% of the knowledge graph with orders of magnitude decrease in time for computation without compromising accuracy.

Extracting and representing user interests on the Social Web is becoming an essential part of the Web for personalisation and recommendations. Such personalisation is required in order to provide an adaptive Web to users, where content fits their preferences, background and current interests, making the Web more social and relevant. Current techniques analyse user activities on social media systems and collect structured or unstructured sets of entities representing users' interests. These sets of entities, or user profiles of interest, are often missing the semantics of the entities in terms of: (i) popularity and temporal dynamics of the interests on the Social Web and (ii) abstractness of the entities in the real world. State of the art techniques to compute these values are using specific knowledge bases or taxonomies and need to analyse the dynamics of the entities over a period of time. Hence, we propose a real-time, computationally inexpensive, domain independent model for concepts of interest composed of: popularity, temporal dynamics and specificity. We describe and evaluate a novel algorithm for computing specificity leveraging the semantics of Linked Data and evaluate the impact of our model on user profiles of interests.

In this paper we discuss the collection, semantic annotation and analysis of real-time social signals from micro blogging data. We focus on users interested in analyzing social signals collectively for sense making. Our proposal enables flexibility in selecting subsets for analysis, alleviating information overload. We define an architecture that is based on state-of-the-art Semantic Web technologies and a distributed publish-subscribe protocol for real time communication. In addition, we discuss our method and application in a scenario related to the health care reform in the United States.