Event coreference resolution (ECR) is a critical task in information extraction of natural language processing, aiming to identify and link event mentions across multiple documents. Despite recent progress, existing datasets for ECR primarily focus on within-document event coreference and English text, lacking cross-document ECR datasets for multiple languages beyond English. To address this issue, this work presents the first multiligual dataset for cross-document ECR, called MCECR (Multilingual Cross-Document Event Coreference Resolution), that manually annotates a diverse collection of documents for event mentions and coreference in five languages, i.e., English, Spanish, Hindi, Turkish, and Ukrainian. Using sampled articles from Wikinews over various topics as the seeds, our dataset fetches related news articles from the Google search engine to increase the number of non-singleton event clusters. In total, we annotate 5,802 news articles, providing a substantial and varied dataset for multilingual ECR in both within-document and cross-document scenarios. Extensive analysis of the proposed dataset reveals the challenging nature of multilingual event coreference resolution tasks, promoting MCECR as a strong benchmark dataset for future research in this area.

Answering questions within business and finance requires reasoning, precision, and a wide-breadth of technical knowledge. Together, these requirements make this domain difficult for large language models (LLMs). We introduce BizBench, a benchmark for evaluating models’ ability to reason about realistic financial problems. BizBench comprises eight quantitative reasoning tasks, focusing on question-answering (QA) over financial data via program synthesis. We include three financially-themed code-generation tasks from newly collected and augmented QA data. Additionally, we isolate the reasoning capabilities required for financial QA: reading comprehension of financial text and tables for extracting intermediate values, and understanding financial concepts and formulas needed to calculate complex solutions. Collectively, these tasks evaluate a model’s financial background knowledge, ability to parse financial documents, and capacity to solve problems with code. We conduct an in-depth evaluation of open-source and commercial LLMs, comparing and contrasting the behavior of code-focused and language-focused models. We demonstrate that the current bottleneck in performance is due to LLMs’ limited business and financial understanding, highlighting the value of a challenging benchmark for quantitative reasoning within this domain.

For large language models (LLMs) to be effective in the financial domain – where each decision can have a significant impact – it is necessary to investigate realistic tasks and data. Financial professionals often interact with documents spanning hundreds of pages, but most financial research datasets only deal with short excerpts from these documents. To address this, we introduce a long-document financial QA task. We augment 7,437 questions from the existing FinQA dataset with full-document context, extending the average context length from under 700 words in FinQA to 123k words in DocFinQA. We conduct extensive experiments over retrieval-based QA pipelines and long-context language models. Based on our experiments, DocFinQA proves a significant challenge for even state-of-the-art systems. We also provide a case study on a subset of the longest documents in DocFinQA and find that models particularly struggle with these documents. Addressing these challenges may have a wide-reaching impact across applications where specificity and long-range contexts are critical, like gene sequences and legal document contract analysis. DocFinQA dataset is publicly accessible.

A key technology for large language models (LLMs) involves instruction tuning that helps align the models’ responses with human expectations to realize impressive learning abilities. Two major approaches for instruction tuning characterize supervised fine-tuning (SFT) and reinforcement learning from human feedback (RLHF), which are applied to produce the best commercial LLMs. To improve the accessibility of LLMs, various instruction-tuned open-source LLMs have also been introduced recently. However, existing open-source LLMs have only been instruction-tuned for English and a few popular languages, thus hindering their accessibility to many other languages in the world. In addition, SFT has been used as the only approach to instruction-tune open-source LLMs for multiple languages. This has left a significant gap for fine-tuned LLMs based on RLHF in diverse languages and raised important questions on how RLHF can boost the performance of multilingual instruction tuning. To overcome this issue, we present Okapi, the first system with instruction-tuned LLMs based on RLHF for multiple languages. Okapi introduces instruction and response-ranked data in 26 diverse languages to facilitate the experiments and development of future multilingual LLM research. We also present benchmark datasets to enable the evaluation of generative LLMs in multiple languages. Our experiments demonstrate the advantages of RLHF for multilingual instruction over SFT for different base models and datasets. Our framework with created resources, fine-tuned LLMs, interaction scripts are released at https://github.com/nlp-uoregon/Okapi. A demo video to show our framework can also be found at: https://youtu.be/QFV2fkPwvi0.

Over the last few years, large language models (LLMs) have emerged as the most important breakthroughs in natural language processing (NLP) that fundamentally transform research and developments in the field. ChatGPT represents one of the most exciting LLM systems developed recently to showcase impressive skills for language generation and highly attract public attention. Among various exciting applications discovered for ChatGPT in English, the model can process and generate texts for multiple languages due to its multilingual training data. Given the broad adoption of ChatGPT for English in different problems and areas, a natural question is whether ChatGPT can also be applied effectively for other languages or it is necessary to develop more language-specific technologies. The answer to this question requires a thorough evaluation of ChatGPT over multiple tasks with diverse languages and large datasets (i.e., beyond reported anecdotes), which is still missing or limited in current research. Our work aims to fill this gap for the evaluation of ChatGPT and similar LLMs to provide more comprehensive information for multilingual NLP applications. In particular, we evaluate ChatGPT on 7 different tasks, covering 37 diverse languages with high, medium, low, and extremely low resources. Compared to the performance of previous models, our extensive experiments demonstrate the worse performance of ChatGPT for different NLP tasks and languages, calling for further research to develop better models and understanding for multilingual learning.

Given the increasing number of livestreaming videos, automatic speech recognition and post-processing for livestreaming video transcripts are crucial for efficient data management as well as knowledge mining. A key step in this process is punctuation restoration which restores fundamental text structures such as phrase and sentence boundaries from the video transcripts. This work presents a new human-annotated corpus, called BehancePR, for punctuation restoration in livestreaming video transcripts. Our experiments on BehancePR demonstrate the challenges of punctuation restoration for this domain. Furthermore, we show that popular natural language processing toolkits like Stanford Stanza, Spacy, and Trankit underperform on detecting sentence boundary on non-punctuated transcripts of livestreaming videos. The dataset is publicly accessible at http://github.com/nlp-uoregon/behancepr.

Suicide is a serious problem in every society. Understanding life events of a potential patient is essential for successful suicide-risk assessment and prevention. In this work, we focus on the Event Detection (ED) task to identify event trigger words of suicide-related events in public posts of discussion forums. In particular, we introduce SuicideED: a new dataset for the ED task that features seven suicidal event types to comprehensively capture suicide actions and ideation, and general risk and protective factors. Our experiments with current state-of-the-art ED systems suggest that this domain poses meaningful challenges as there is significant room for improvement of ED models. We will release SuicideED to support future research in this important area.

Subevent Relation Extraction (SRE) is a task in Information Extraction that aims to recognize spatial and temporal containment relations between event mentions in text. Recent methods have utilized pre-trained language models to represent input texts for SRE. However, a key issue in existing SRE methods is the employment of sequential order of words in texts to feed into representation learning methods, thus unable to explicitly focus on important context words and their interactions to enhance representations. In this work, we introduce a new method for SRE that learns to induce effective graph structures for input texts to boost representation learning. Our method features a word alignment framework with dependency paths and optimal transport to identify important context words to form effective graph structures for SRE. In addition, to enable SRE research on non-English languages, we present a new multilingual SRE dataset for five typologically different languages. Extensive experiments reveal the state-of-the-art performance for our method on different datasets and languages.

We describe Symlink, a SemEval shared task of extracting mathematical symbols and their descriptions from LaTeX source of scientific documents. This is a new task in SemEval 2022, which attracted 180 individual registrations and 59 final submissions from 7 participant teams. We expect the data developed for this task and the findings reported to be valuable for the scientific knowledge extraction and automated knowledge base construction communities. The data used in this task is publicly accessible at https://github.com/nlp-oregon/symlink.

Livestreaming videos have become an effective broadcasting method for both video sharing and educational purposes. However, livestreaming videos contain a considerable amount of off-topic content (i.e., up to 50%) which introduces significant noises and data load to downstream applications. This paper presents BehanceCC, a new human-annotated benchmark dataset for off-topic detection (also called chitchat detection) in livestreaming video transcripts. In addition to describing the challenges of the dataset, our extensive experiments of various baselines reveal the complexity of chitchat detection for livestreaming videos and suggest potential future research directions for this task. The dataset will be made publicly available to foster research in this area.

Question-Answer (QA) is one of the effective methods for storing knowledge which can be used for future retrieval. As such, identifying mentions of questions and their answers in text is necessary for a knowledge construction and retrieval systems. In the literature, QA identification has been well studied in the NLP community. However, most of the prior works are restricted to formal written documents such as papers or websites. As such, Questions and Answers that are presented in informal/noisy documents have not been adequately studied. One of the domains that can significantly benefit from QA identification is the domain of livestreaming video transcripts that involve abundant QA pairs to provide valuable knowledge for future users and services. Since video transcripts are often transcribed automatically for scale, they are prone to errors. Combined with the informal nature of discussion in a video, prior QA identification systems might not be able to perform well in this domain. To enable comprehensive research in this domain, we present a large-scale QA identification dataset annotated by human over transcripts of 500 hours of streamed videos. We employ Behance.net to collect the videos and their automatically obtained transcripts. Furthermore, we conduct extensive analysis on the annotated dataset to understand the complexity of QA identification for livestreaming video transcripts. Our experiments show that the annotated dataset presents unique challenges for existing methods and more research is necessary to explore more effective methods. The dataset and the models developed in this work will be publicly released for future research.

Cross-Lingual Event Detection (CLED) models are capable of performing the Event Detection (ED) task in multiple languages. Such models are trained using data from a source language and then evaluated on data from a distinct target language. Training is usually performed in the standard supervised setting with labeled data available in the source language. The Few-Shot Learning (FSL) paradigm is yet to be explored for CLED despite its inherent advantage of allowing models to better generalize to unseen event types. As such, in this work, we study the CLED task under an FSL setting. Our contribution is threefold: first, we introduce a novel FSL classification method based on Optimal Transport (OT); second, we present a novel regularization term to incorporate the global distance between the support and query sets; and third, we adapt our approach to the cross-lingual setting by exploiting the alignment between source and target data. Our experiments on three, syntactically-different, target languages show the applicability of our approach and its effectiveness at improving the cross-lingual performance of few-shot models for event detection.

Event Detection (ED) aims to recognize mentions of events (i.e., event triggers) and their types in text. Recently, several ED datasets in various domains have been proposed. However, the major limitation of these resources is the lack of enough training data for individual event types which hinders the efficient training of data-hungry deep learning models. To overcome this issue, we propose to exploit the powerful pre-trained language model GPT-2 to generate training samples for ED. To prevent the noises inevitable in automatically generated data from hampering training process, we propose to exploit a teacher-student architecture in which the teacher is supposed to learn anchor knowledge from the original data. The student is then trained on combination of the original and GPT-generated data while being led by the anchor knowledge from the teacher. Optimal transport is introduced to facilitate the anchor knowledge-based guidance between the two networks. We evaluate the proposed model on multiple ED benchmark datasets, gaining consistent improvement and establishing state-of-the-art results for ED.

We address the sampling bias and outlier issues in few-shot learning for event detection, a subtask of information extraction. We propose to model the relations between training tasks in episodic few-shot learning by introducing cross-task prototypes. We further propose to enforce prediction consistency among classifiers across tasks to make the model more robust to outliers. Our extensive experiment shows a consistent improvement on three few-shot learning datasets. The findings suggest that our model is more robust when labeled data of novel event types is limited. The source code is available at http://github.com/laiviet/fsl-proact.