We introduce MultiMatch, a novel semisupervised learning (SSL) algorithm combining the paradigms of co-training and consistency regularization with pseudo-labeling.At its core, MultiMatch features a pseudo-label weighting module designed for selecting and filtering pseudo-labels based on head agreement and model confidence, and weighting them according to the perceived classification difficulty.This novel module enhances and unifies three existing techniques -heads agreement from Multihead Co-training, self-adaptive thresholds from FreeMatch, and Average Pseudo-Margins from MarginMatch -resulting in a holistic approach that improves robustness and performance in SSL settings.Experimental results on benchmark datasets highlight the superior performance of MultiMatch, i.e., Multi-Match achieves state-of-the-art results on 8 out of 10 setups from 5 natural language processing datasets and ranks first according to the Friedman test among 21 methods.Furthermore, Mul-tiMatch demonstrates exceptional robustness in highly imbalanced settings, outperforming the second-best approach by 3.26%, a critical advantage for real-world text classification tasks.Our code is available on GitHub.

Shaona Ghosh, Prasoon Varshney, Makesh Narsimhan Sreedhar, Aishwarya Padmakumar, Traian Rebedea, Jibin Rajan Varghese, Christopher Parisien. Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers). 2025.

One proposed mechanism to improve exploration in reinforcement learning is the use of macro-actions, a form of temporal abstractions over actions. Paradoxically though, in many scenarios the naive addition of macro-actions does not lead to better exploration, but rather the opposite. In this work, we argue that the difficulty stems from the trade-offs between reducing the average number of decisions per episode versus increasing the size of the action space. Namely, one typically treats each potential macro-action as independent and atomic, hence strictly increasing the search space and making typical exploration strategies inefficient. To address this problem we propose a novel regularization term that exploits the relationship between actions and macro-actions to improve the credit assignment mechanism reducing the effective dimension of the action space and therefore improving exploration. The term relies on a similarity matrix that is meta-learned jointly with learning the desired policy. We empirically validate our strategy looking at macro-actions in Atari games, and the StreetFighter II environment. Our results show significant improvements over the Rainbow-DQN baseline in all environments. Additionally, we show that the macro-action similarity is transferable to other environments with similar dynamics. We believe this work is a small but important step towards understanding how the similarity-imposed geometry on the action space can be exploited to improve credit assignment and exploration, therefore making learning more efficient.

While chatbots show promise for large-scale mental health screening, few offer interactive, free-text conversations, limiting their appeal for self-administered screening and impeding the timely detection of mental health issues. This study introduces an AI-based chatbot that allows users to respond to validated screening surveys for mental disorders (PHQ-9, GAD-7, and PCL-5) in a natural, free-text conversation manner with real-time feedback. The study's objectives include evaluating the chatbot's usability and reducing the frequency of response clarifications while accurately interpreting users’ responses. The system was assessed running in hybrid NLU mode (Phase 2; N = 587; Mage = 21.56, SD = 5.56, 67.8% women) after being trained on data collected while running in rule-based mode (Phase 1; N = 274; Mage = 21.86, SD = 5.50). During user-chatbot interactions, the chatbot required clarification only 4.64% of the time. Using the AI NLU model, the chatbot could understand user responses in 85.65% of cases and interpret free-text similarly to human annotators. In terms of usability, the chatbot in hybrid NLU mode was perceived as more engaging, friendly, and easier to use than in the rule-based NLU mode, which may be indirectly attributed to the enhanced autonomy provided by the AI NLU model.

Nikesh Gyawali, Iustin Sirbu, Tiberiu Sosea, Sarthak Khanal, Doina Caragea, Traian Rebedea, Cornelia Caragea. Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2024.

Dialogue policies play a crucial role in developing task-oriented dialogue systems, yet their development and maintenance are challenging and typically require substantial effort from experts in dialogue modeling.While in many situations, large amounts of conversational data are available for the task at hand, people lack an effective solution able to extract dialogue policies from this data.In this paper, we address this gap by first illustrating how Large Language Models (LLMs) can be instrumental in extracting dialogue policies from datasets, through the conversion of conversations into a unified intermediate representation consisting of canonical forms.We then propose a novel method for generating dialogue policies utilizing a controllable and interpretable graphbased methodology.By combining canonical forms across conversations into a flow network, we find that running graph traversal algorithms helps in extracting dialogue flows.These flows are a better representation of the underlying interactions than flows extracted by prompting LLMs.Our technique focuses on giving conversation designers greater control, offering a productivity tool to improve the process of developing dialogue policies.1 bot "Hello, how can I help you today?" intent: express greeting and offer to help user "I received this pair of boots but they are scuffed" intent: inform product received is scuffed user "I want to return and exchange them for a new pair" intent: request to return and exchange product bot "No worries, can I have your full name?"intent: ask for full name user "Alessandro Phoenix" intent: provide full name bot "And your username, email, and order ID?" intent: ask for username, email, and order id ... bot "Well then have a nice day!" intent: say goodbye bot "Hello, how can I help you today?" intent: express greeting and offer to help user "I received this pair of

Competitive programming presents challenges for students seeking to enhance programming and algorithmic skills. This research introduces a system that efficiently matches problem statements to editorials that describe the solution, helping students find relevant learning resources. The main component of this system is our learning-to-rank model, which achieves a P@1 score of 0.93, indicating its proficiency in identifying the most relevant editorial for a specific problem statement. While our model is smaller in scale compared to general models like GPT-4, it distinguishes itself with comparable results and notable computational efficiency. Additionally, we have developed a new dataset of 1550 competitive programming problem statements and their editorials. Integrated into a competitive programming platform, it has the potential to evolve into an adaptive learning system, customizing paths based on individual user performance. Our code and data are public at https://github.com/DinuGeorge0019/MatchingProblemStatementsToEditorialsInCP.

In competitive programming, understanding a problem often requires more than just the official solution. Users typically turn to comments in the contest’s thread for additional insights. These comments often contain irrelevant information, necessitating the manual identification of relevant ones. This paper introduces CommentThreadFilter, a system designed to classify comments as either Relevant or Irrelevant for the main thread post. Leveraging base models like BERT, RoBERTa, and SciBERT, we evaluate the system’s performance on the newly created CFComments dataset. The dataset is the first of its kind, comprising 19 labelled comment threads in the competitive programming domain, manually annotated by two experts, alongside 1131 unlabeled comment threads. The proposed models, combined with a weak augmentation on the text, achieve an F1-score of 86%, outperforming the 77% F1-score obtained using gpt-3.5-turbo. By effectively filtering comments as Relevant or Irrelevant, our system enhances the user’s ability to gain valuable insights and better comprehend the underlying problem.

During natural disasters, people often use social media platforms, such as Twitter, to post information about casualties and damage produced by disasters. This information can help relief authorities gain situational awareness in nearly real time, and enable them to quickly distribute resources where most needed. However, annotating data for this purpose can be burdensome, subjective and expensive. In this paper, we investigate how to leverage the copious amounts of unlabeled data generated on social media by disaster eyewitnesses and affected individuals during disaster events. To this end, we propose a semi-supervised learning approach to improve the performance of neural models on several multimodal disaster tweet classification tasks. Our approach shows significant improvements, obtaining up to 7.7% improvements in F-1 in low-data regimes and 1.9% when using the entire training data. We make our code and data publicly available at https://github.com/iustinsirbu13/multimodal-ssl-for-disaster-tweet-classification.

Natural language processing has had a significant growth in recent years, two key factors being the increase of processing capabilities leading to very large models and the availability of larger standardized datasets. However, not every specific task has a fair amount of available data and using more data does not always lead to better results. Therefore, we also need to focus on obtaining good results even when small datasets are provided. For this reason, we propose a model for question answering, called ZeroQA, which ranked first place when submitted in two popular leaderboards for answering multiple-choice science questions: ARC Easy and ARC Challenge. Our ZeroQA model uses no transformers fine-tuned on the ARC training dataset, and relies mainly on transfer learning using a mixture of experts. We also propose methods of selecting relevant subsets from the ARC datasets for training and we use them to analyze how our model performs with less, but well chosen training datasets.

Running large-scale pre-trained language models in computationally constrained environments remains a challenging problem yet to be addressed, while transfer learning from these models has become prevalent in Natural Language Processing tasks. Several solutions, including knowledge distillation, network quantization, or network pruning have been previously proposed; however, these approaches focus mostly on the English language, thus widening the gap when considering low-resource languages. In this work, we introduce three light and fast versions of distilled BERT models for the Romanian language: Distil-BERT-base-ro, Distil-RoBERT-base, and DistilMulti-BERT-base-ro. The first two models resulted from the individual distillation of knowledge from two base versions of Romanian BERTs available in literature, while the last one was obtained by distilling their ensemble. To our knowledge, this is the first attempt to create publicly available Romanian distilled BERT models, which were thoroughly evaluated on five tasks: part-of-speech tagging, named entity recognition, sentiment analysis, semantic textual similarity, and dialect identification. Our experimental results argue that the three distilled models offer performance comparable to their teachers, while being twice as fast on a GPU and ~35% smaller. In addition, we further test the similarity between the predictions of our students versus their teachers by measuring their label and probability loyalty, together with regression loyalty - a new metric introduced in this work.

We propose a novel solution for assigning labels to topic models by using multiple weak labelers.The method leverages generative transformers to learn accurate representations of the most important topic terms and candidate labels.This is achieved by fine-tuning pretrained BART models on a large number of potential labels generated by state of the art nonneural models for topic labeling, enriched with different techniques.The proposed BART-TL model is able to generate valuable and novel labels in a weakly-supervised manner and can be improved by adding other weak labelers or distant supervision on similar tasks.

Radu Cristian Alexandru Iacob, Florin Brad, Elena-Simona Apostol, Ciprian-Octavian Truică, Ionel Alexandru Hosu, Traian Rebedea. Proceedings of the 28th International Conference on Computational Linguistics. 2020.

George Sebastian Pirtoaca, Traian Rebedea, Stefan Ruseti. Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP). 2019.

Question answering is one of the most important and difficult applications at the border of information retrieval and natural language processing, especially when we talk about complex questions which require some form of inference to determine the correct answer. In this paper, we present a two-step method that combines information retrieval techniques optimized for question answering with deep learning models for natural language inference in order to tackle the multiple-choice question answering problem. In the first stage, each question-answer pair is fed into an information retrieval engine to find relevant candidate contexts that serve as the underlying knowledge for the inference models. In the second stage, deep learning architectures are used to predict if a candidate answer can be inferred from the context extracted in the first stage. We deploy multiple deep learning architectures pre-trained on different datasets in order to capture semantic features and to enlarge the scope of the questions we can answer correctly. As it will be described, each dataset used for training the inference models has particular characteristics that can be exploited. In the end, all these solvers are combined in an ensemble model to predict the correct answer. This proposed two-step model outperforms the best retrieval-based solver by over 3% in absolute accuracy. Moreover, the model can answer both simple, factoid questions and more complex questions that require reasoning or inference.

We propose a sketch-based two-step neural model for generating structured queries (SQL) based on a user’s request in natural language. The sketch is obtained by using placeholders for specific entities in the SQL query, such as column names, table names, aliases and variables, in a process similar to semantic parsing. The first step is to apply a sequence-to-sequence (SEQ2SEQ) model to determine the most probable SQL sketch based on the request in natural language. Then, a second network designed as a dual-encoder SEQ2SEQ model using both the text query and the previously obtained sketch is employed to generate the final SQL query. Our approach shows improvements over previous approaches on two recent large datasets (WikiSQL and SENLIDB) suitable for data-driven solutions for natural language interfaces for databases.

Recent advances in neural code generation have incorporated syntax to improve the generation of the target code based on the user's request in natural language. We adapt the model of [1] to the Natural Language Interface to Databases (NLIDB) problem by taking into account the database schema. We evaluate our model on the recently introduced WIKISQL and SENLIDB datasets. Our results show that the syntax-guided model outperforms a simple sequence-to-sequence (SEQ2SEQ) baseline on WIKISQL, but has trouble with the SENLIDB dataset due to its complexity.

This paper presents a research on using Word2Vec for determining implicit links in multi-participant Computer-Supported Collaborative Learning chat conversations. Word2Vec is a powerful and one of the newest Natural Language Processing semantic models used for computing text cohesion and similarity between documents. This research considers cohesion scores in terms of the strength of the semantic relations established between two utterances, the higher the score, the stronger the similarity between two utterances. An implicit link is established based on cohesion to the most similar previous utterance, within an imposed window. Three similarity formulas were used to compute the cohesion score: an unnormalized score, a normalized score with distance and Mihalcea's formula. Our corpus of conversations incorporated explicit references provided by authors, which were used for validation. A window of 5 utterances and a 1-minute time frame provided the highest detection rate both for exact matching and matching of a block of continuous utterances belonging to the same speaker. Moreover, the unnormalized score correctly identified the largest number of implicit links.

Progress in natural language interfaces to databases (NLIDB) has been slow mainly due to linguistic issues (such as language ambiguity) and domain portability. Moreover, the lack of a large corpus to be used as a standard benchmark has made data-driven approaches difficult to develop and compare. In this paper, we revisit the problem of NLIDBs and recast it as a sequence translation problem. To this end, we introduce a large dataset extracted from the Stack Exchange Data Explorer website, which can be used for training neural natural language interfaces for databases. We also report encouraging baseline results on a smaller manually annotated test corpus, obtained using an attention-based sequence-to-sequence neural network.

Progress in statistical paraphrase generation has been hindered for a long time by the lack of large monolingual parallel corpora.In this paper, we adapt the neural machine translation approach to paraphrase generation and perform transfer learning from the closely related task of entailment generation.We evaluate the model on the Microsoft Research Paraphrase (MSRP) corpus and show that the model is able to generate sentences that capture part of the original meaning, but fails to pick up on important words or to show large lexical variation.

In recent years, there have been several advancements in question answering systems. These were achieved both due to the availability of a greater number of datasets, some of them significantly larger in size than any of the existing corpora, and to the recent advancements in deep learning for text classification. In this paper, we explore the improvements achieved by employing neural networks using the features computed by a string kernel for sentence/answer selection. We have validated this approach using two different standard corpora used as benchmarks in question answering and we have found a significant improvement over string kernels and other unsupervised methods for sentence selection.

Abstract This paper presents a system developed for detecting sexual predators in online chat conversations using a two-stage classification and behavioral features. A sexual predator is defined as a person who tries to obtain sexual favors in a predatory manner, usually with underage people. The proposed approach uses several text categorization methods and empirical behavioral features developed especially for the task at hand. After investigating various approaches for solving the sexual predator identification problem, we have found that a two-stage classifier achieves the best results. In the first stage, we employ a Support Vector Machine classifier to distinguish conversations having suspicious content from safe online discussions. This is useful as most chat conversations in real life do not contain a sexual predator, therefore it can be viewed as a filtering phase that enables the actual detection of predators to be done only for suspicious chats that contain a sexual predator with a very high degree. In the second stage, we detect which of the users in a suspicious discussion is an actual predator using a Random Forest classifier. The system was tested on the corpus provided by the PAN 2012 workshop organizers and the results are encouraging because, as far as we know, our solution outperforms all previous approaches developed for solving this task.

Many historic periods (or events) are remembered by slogans, expressions or words that are strongly linked to them. Educated people are also able to determine whether a particular word or expression is related to a specific period in human history. The present paper aims to establish correlations between significant historic periods (or events) and the texts written in that period. In order to achieve this, we have developed a system that automatically links words (and topics discovered using Latent Dirichlet Allocation) to periods of time in the recent history. For this analysis to be relevant and conclusive, it must be undertaken on a representative set of texts written throughout history. To this end, instead of relying on manually selected texts, the Google Books Ngram corpus has been chosen as a basis for the analysis. Although it provides only word n-gram statistics for the texts written in a given year, the resulting time series can be used to provide insights about the most important periods and events in recent history, by automatically linking them with specific keywords or even LDA topics.

The paper presents a system for helping the analysis of Computer-Supported Collaborative Learning chat (instant messenger) sessions, starting from a polyphonic model of the discourse inspired by the dialogistics of Bakhtin. Some theoretical basics of the model are presented, followed by implementation details and validation results.