Yiqiao Jin, Kartik Sharma, Vineeth Rakesh, Yingtong Dou, Menghai Pan, Mahashweta Das, Srijan Kumar. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Chia-Yuan Chang, Zhimeng Jiang, Vineeth Rakesh, Menghai Pan, Chin-Chia Michael Yeh, Guanchu Wang, Mingzhi Hu, Zhichao Xu, Yan Zheng, Mahashweta Das, Na Zou. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

We consider the problem of test-time adaptation of predictive models trained on tabular data. Effective solution of this problem requires adaptation of predictive models trained on the source domain to a target domain, using only unlabeled target domain data, without access to source domain data. Existing test-time adaptation methods for tabular data have difficulty coping with the heterogeneous features and their complex dependencies inherent in tabular data. To overcome these limitations, we consider test-time adaptation in the setting wherein the logical structure of the rules is assumed to remain invariant despite distribution shift between source and target domains whereas the numerical parameters associated with the rules and the weights assigned to them can vary to accommodate distribution shift. TabLog discretizes numerical features, models dependencies between heterogeneous features, introduces a novel contrastive loss for coping with distribution shift, and presents an end-to-end framework for efficient training and test-time adaptation by taking advantage of a logical neural network representation of a rule ensemble. We present results of experiments using several benchmark data sets that demonstrate TabLog is competitive with or improves upon the state-of-the-art methods for test-time adaptation of predictive models trained on tabular data. Our code is available at https://github.com/WeijieyingRen/TabLog.

Recent advancements in deep learning techniques have transformed the area of semantic text matching (STM). However, most state-of-the-art models are designed to operate with short documents such as tweets, user reviews, comments, and so on. These models have fundamental limitations when applied to long-form documents such as scientific papers, legal documents, and patents. When handling such long documents, there are three primary challenges: (i) the presence of different contexts for the same word throughout the document, (ii) small sections of contextually similar text between two documents, but dissimilar text in the remaining parts (this defies the basic understanding of “similarity”), and (iii) the coarse nature of a single global similarity measure which fails to capture the heterogeneity of the document content. In this article, we describe CoLDE : Co ntrastive L ong D ocument E ncoder—a transformer-based framework that addresses these challenges and allows for interpretable comparisons of long documents. CoLDE uses unique positional embeddings and a multi-headed chunkwise attention layer in conjunction with a supervised contrastive learning framework to capture similarity at three different levels: (i) high-level similarity scores between a pair of documents, (ii) similarity scores between different sections within and across documents, and (iii) similarity scores between different chunks in the same document and across other documents. These fine-grained similarity scores aid in better interpretability. We evaluate CoLDE on three long document datasets namely, ACL Anthology publications, Wikipedia articles, and USPTO patents. Besides outperforming the state-of-the-art methods on the document matching task, CoLDE is also robust to changes in document length and text perturbations and provides interpretable results. The code for the proposed model is publicly available at https://github.com/InterDigitalInc/CoLDE .

Node embeddings obtained from information networks have been widely adopted for representing knowledge and driving various information retrieval and machine learning tasks. However, training node embeddings is computationally intensive, making it difficult to scale to larger graphs. Most existing works have addressed the scalability challenge by simply adding more hardware resources. For example, a common approach to speed up the training process is to distribute model computation across multiple machines and GPUs. This paper takes an orthogonal approach towards scalability by addressing the problem of computation complexity in training embeddings. We present bigg2vec for scaling up the embedding training process. bigg2vec introduces a novel polar coordinate-based system for internal representation and computation. It provides the following benefits: (a) It significantly reduces compute and memory requirements while improving embedding quality and (b) uses a novel graph organization to generate high-quality negative samples (this reduces the number of negative samples needed for training, which is especially beneficial f or skewed graphs). We have deployed bigg2vec to generate embeddings for multiple AI models within Visa. Our Global Personalized Restaurant Recommender System (GPR) is one such project that uses bigg2vec to periodically generate embeddings for over 450 million nodes connected by more than 3 billion edges. bigg2vec generates higher quality embeddings while training them faster than state-of-the-art methods on a single CPU-based machine.

Graph neural networks have achieved significant success in representation learning. However, the performance gains come at a cost; acquiring comprehensive labeled data for training can be prohibitively expensive. Active learning mitigates this issue by searching the unexplored data space and prioritizing the selection of data to maximize model's performance gain. In this paper, we propose a novel method SMARTQUERY, a framework to learn a graph neural network with very few labeled nodes using a hybrid uncertainty reduction function. This is achieved using two key steps: (a) design a multi-stage active graph learning framework by exploiting diverse explicit graph information and (b) introduce label propagation to efficiently exploit known labels to assess the implicit embedding information. Using a comprehensive set of experiments on three network datasets, we demonstrate the competitive performance of our method against state-of-the-arts on very few labeled data (up to 5 labeled nodes per class).

In the era of big data, online doctor review platforms, which enable patients to give feedback to their doctors, have become one of the most important components in healthcare systems. On one hand, they help patients to choose their doctors based on the experience of others. On the other hand, they help doctors to improve the quality of their service. Moreover, they provide important sources for us to discover common concerns of patients and existing problems in clinics, which potentially improve current healthcare systems. In this paper, we systematically investigate the dataset from one of such review platform, namely, ratemds.com, where each review for a doctor comes with an overall rating and ratings of four different aspects. A comprehensive statistical analysis is conducted first for reviews, ratings, and doctors. Then, we explore the content of reviews by extracting latent topics related to different aspects with unsupervised topic modeling techniques. As the core component of this paper, we propose a multi-task learning framework for the document-level multi-aspect sentiment classification. This task helps us to not only recover missing aspect-level ratings and detect inconsistent rating scores but also identify aspect-keywords for a given review based on ratings. The proposed model takes both features of doctors and aspect-keywords into consideration. Extensive experiments have been conducted on two subsets of ratemds dataset to demonstrate the effectiveness of the proposed model.

Several complex tasks that arise in organizations can be simplified by mapping them into a matrix completion problem. In this paper, we address a key challenge faced by our company: predicting the efficiency of artists in rendering visual effects (VFX) in film shots. We tackle this challenge by using a two-fold approach: first, we transform this task into a constrained matrix completion problem with entries bounded in the unit interval [0,1]; second, we propose two novel matrix factorization models that leverage our knowledge of the VFX environment. Our first approach, expertise matrix factorization (EMF), is an interpretable method that structures the latent factors as weighted user-item interplay. The second one, survival matrix factorization (SMF), is instead a probabilistic model for the underlying process defining employees' efficiencies. We show the effectiveness of our proposed models by extensive numerical tests on our VFX dataset and two additional datasets with values that are also bounded in the [0,1] interval.

Recommendation in the modern world is not only about capturing the interaction between users and items, but also about understanding the relationship between items. Besides improving the quality of recommendation, it enables the generation of candidate items that can serve as substitutes and supplements of another item. For example, when recommending Xbox, PS4 could be a logical substitute and the supplements could be items such as game controllers, surround system, and travel case. Therefore, given a network of items, our objective is to learn their content features such that they explain the relationship between items in terms of substitutes and supplements. To achieve this, we propose a generative deep learning model that links two variational autoencoders using a connector neural network to create Linked Variational Autoencoder (LVA). LVA learns the latent features of items by conditioning on the observed relationship between items. Using a rigorous series of experiments, we show that LVA significantly outperforms other representative and state-of-the-art baseline methods in terms of prediction accuracy. We then extend LVA by incorporating collaborative filtering (CF) to create CLVA that captures the implicit relationship between users and items. By comparing CLVA with LVA we show that inducing CF-based features greatly improve the recommendation quality of substitutable and supplementary items on a user level.

The growth of social media has created an open web where people freely share their opinion and even discuss sensitive subjects in online forums. Forums such as Reddit help support seekers by serving as a portal for open discussions for various stigmatized subjects such as rape. This paper investigates the potential roles of online forums and if such forums provide intended resources to the people who seek support. Specifically, the open nature of forums allows us to study how online users respond to seeker's queries or needs; through their response, we attempt to assess the range of topics covered by responders in regards to the issues, concerns and, obstacles faced by the victims of rape and sexual abuse, using rape-related posts from Reddit. We employ natural language processing techniques to extract topics of responses, examine how diverse these topics are to answer research questions such as whether responses are limited to emotional support; if not, what other topics are; what the diversity of topics manifests; and how online response differs from traditional response found in a physical world.

Modeling spillover effects from observational data is an important problem in economics, business, and other fields of research. It helps us infer the causality between two seemingly unrelated set of events. For example, if consumer spending in the United States declines, it has spillover effects on economies that depend on the U.S. as their largest export market. In this paper, we aim to infer the causation that results in spillover effects between pairs of entities (or units); we call this effect as paired spillover. To achieve this, we leverage the recent developments in variational inference and deep learning techniques to propose a generative model called Linked Causal Variational Autoencoder (LCVA). Similar to variational autoencoders (VAE), LCVA incorporates an encoder neural network to learn the latent attributes and a decoder network to reconstruct the inputs. However, unlike VAE, LCVA treats the latent attributes as confounders that are assumed to affect both the treatment and the outcome of units. Specifically, given a pair of units u and $\baru $, their individual treatment and outcomes, the encoder network of LCVA samples the confounders by conditioning on the observed covariates of u, the treatments of both u and $\baru $ and the outcome of u. Once inferred, the latent attributes (or confounders) of u captures the spillover effect of $\baru $ on u. Using a network of users from job training dataset (LaLonde (1986)) and co-purchase dataset from Amazon e-commerce domain, we show that LCVA is significantly more robust than existing methods in capturing spillover effects.

Online reviews have become an inevitable part of a consumer's decision making process, where the likelihood of purchase not only depends on the product's overall rating, but also on the description of its aspects. Therefore, e-commerce websites such as Amazon and Walmart constantly encourage users to write good quality re- views and categorically summarize different facets of the products. However, despite such attempts, it takes a significant effort to skim through thousands of reviews and look for answers that address the query of consumers. For example, a gamer might be interested in buying a monitor with fast refresh rates and support for Gsync and Freesync technologies, while a photographer might be interested in aspects such as color depth and accuracy. To address these chal- lenges, in this paper, we propose a generative aspect summarization model called APSUM that is capable of providing fine-grained sum- maries of online reviews. To overcome the inherent problem of aspect sparsity, we impose dual constraints: (a) a spike-and-slab prior over the document-topic distribution and (b) a linguistic su- pervision over the word-topic distribution. Using a rigorous set of experiments, we show that the proposed model is capable of out- performing the state-of-the-art aspect summarization model over a variety of datasets and deliver intuitive fine-grained summaries that could simplify the purchase decisions of consumers.

The pervasive growth of location-based services such as Foursquare and Yelp has enabled researchers to incorpo- rate better personalization into recommendation models by leveraging the geo-temporal breadcrumbs left by a plethora of travelers. In this paper, we explore Travel path recommendation, which is one of the applications of intelligent urban navigation that aims in recommending sequence of point of interest (POIs) to tourists. Currently, travelers rely on a tedious and time-consuming process of searching the web, browsing through websites such as Trip Advisor, and reading travel blogs to compile an itinerary. On the other hand, people who do not plan ahead of their trip find it extremely difficult to do this in real-time since there are no automated systems that can provide personalized itinerary for travelers. To tackle this problem, we propose a tour recommendation model that uses a probabilistic generative framework to incorporate user's categorical preference, influence from their social circle, the dynamic travel transitions (or patterns) and the popularity of venues to recommend sequence of POIs for tourists. Through comprehensive experiments over a rich dataset of travel patterns from Foursquare, we show that our model is capable of outperforming the state-of-the-art probabilistic tour recommendation model by providing contextual and meaningful recommendation for travelers.

Crowdfunding has gained widespread attention in recent years. Despite the huge success of crowdfunding platforms, the percentage of projects that succeed in achieving their desired goal amount is only around 40%. Moreover, many of these crowdfunding platforms follow "all-or-nothing" policy which means the pledged amount is collected only if the goal is reached within a certain predefined time duration. Hence, estimating the probability of success for a project is one of the most important research challenges in the crowdfunding domain. To predict the project success, there is a need for new prediction models that can potentially combine the power of both classification (which incorporate both successful and failed projects) and regression (for estimating the time for success). In this paper, we formulate the project success prediction as a survival analysis problem and apply the censored regression approach where one can perform regression in the presence of partial information. We rigorously study the project success time distribution of crowdfunding data and show that the logistic and log-logistic distributions are a natural choice for learning from such data. We investigate various censored regression models using comprehensive data of 18K Kickstarter (a popular crowdfunding platform) projects and 116K corresponding tweets collected from Twitter. We show that the models that take complete advantage of both the successful and failed projects during the training phase will perform significantly better at predicting the success of future projects compared to the ones that only use the successful projects. We provide a rigorous evaluation on many sets of relevant features and show that adding few temporal features that are obtained at the project's early stages can dramatically improve the performance.

Crowdfunding has gained a widespread popularity by fueling the creative minds of entrepreneurs. Not only has it democratized the funding of startups, it has also bridged the gap between the venture capitalists and the entrepreneurs by providing a plethora of opportunities for people seeking to invest in new business ventures. Nonetheless, despite the huge success of the crowdfunding platforms, not every project reaches its funding goal. One of the main reasons for a project's failure is the difficulty in establishing a linkage between it's founders and those investors who are interested in funding such projects. A potential solution to this problem is to develop recommendation systems that suggest suitable projects to crowdfunding investors by capturing their interests. In this paper, we explore Kickstarter, a popular reward-based crowdfunding platform. Being a highly heterogeneous platform, Kickstarter is fuelled by a dynamic community of people who constantly interact with each other before investing in projects. Therefore, the decision to invest in a project depends not only on the preference of individuals, but also on the influence of groups that a person belongs and the on-going status of the projects. In this paper, we propose a probabilistic recommendation model, called CrowdRec, that recommends Kickstarter projects to a group of investors by incorporating the on-going status of projects, the personal preference of individual members, and the collective preference of the group . Using a comprehensive dataset of over 40K crowdfunding groups and 5K projects, we show that our model is effective in recommending projects to groups of Kickstarter users.

Automatic detection of tweets that provide Location-specific information will be extremely useful in conveying geo-location based knowledge to the users. However, there is a significant challenge in retrieving such tweets due to the sparsity of geo-tag information, the short textual nature of tweets, and the lack of pre-defined set of topics. In this paper, we develop a novel framework to identify and summarize tweets that are specific to a location. First, we propose a weighting scheme called Location Centric Word Co-occurrence (LCWC) that uses the content of the tweets and the network information of the twitterers to identify tweets that are location-specific. We evaluate the proposed model using a set of annotated tweets and compare the performance with other weighting schemes studied in the literature. This paper reports three key findings: (a) top trending tweets from a location are poor descriptors of location-specific tweets, (b) ranking tweets purely based on users' geo-location cannot ascertain the location specificity of tweets, and (c) users' network information plays an important role in determining the location-specific characteristics of the tweets. Finally, we train a topic model based on Latent Dirichlet Allocation (LDA) using a large collection of local news database and tweet-based Urls to predict the topics from the location-specific tweets and present them using an interactive web-based interface.