We address the task of human action representation and show how the approach to generating word representations based on co-occurrence can be adapted to generate human action representations by analyzing their co-occurrence in videos. To this end, we formalize the new task of human action co-occurrence identification in online videos, i.e., determine whether two human actions are likely to co-occur in the same interval of time.We create and make publicly available the Co-Act (Action Co-occurrence) dataset, consisting of a large graph of ~12k co-occurring pairs of visual actions and their corresponding video clips. We describe graph link prediction models that leverage visual and textual information to automatically infer if two actions are co-occurring.We show that graphs are particularly well suited to capture relations between human actions, and the learned graph representations are effective for our task and capture novel and relevant information across different data domains.
This paper presents a language model trained from scratch exclusively on a brand new corpus consisting of about 6 GiB of Uruguayan newspaper text. We trained the model for 30 days on a single Nvidia P100 using the RoBERTa-base architecture but with considerably fewer parameters than other standard RoBERTa models. We evaluated the model on two NLP tasks and found that it outperforms BETO, the widely used Spanish BERT pre-trained model. We also compared our model on the masked-word prediction task with two popular multilingual BERT-based models, Multilingual BERT and XLM-RoBERTa, obtaining outstanding results on sentences from the Uruguayan press domain. Our experiments show that training a language model on a domain-specific corpus can significantly improve performance even when the model is smaller and was trained with significantly less data than more standard pre-trained models.
Recent progress in large language models (LLMs) has enabled the deployment of many generative NLP applications. At the same time, it has also led to a misleading public discourse that “it’s all been solved.” Not surprisingly, this has, in turn, made many NLP researchers – especially those at the beginning of their careers – worry about what NLP research area they should focus on. Has it all been solved, or what remaining questions can we work on regardless of LLMs? To address this question, this paper compiles NLP research directions rich for exploration. We identify fourteen different research areas encompassing 45 research directions that require new research and are not directly solvable by LLMs. While we identify many research areas, many others exist; we do not cover areas currently addressed by LLMs, but where LLMs lag behind in performance or those focused on LLM development. We welcome suggestions for other research directions to include: https://bit.ly/nlp-era-llm.
Joint vision-language models have shown great performance over a diverse set of tasks. However, little is known about their limitations, as the high dimensional space learned by these models makes it difficult to identify semantic errors. Recent work has addressed this problem by designing highly controlled probing task benchmarks. Our paper introduces a more scalable solution that relies on already annotated benchmarks. Our method consists of extracting a large set of diverse features from a vision-language benchmark and measuring their correlation with the output of the target model. We confirm previous findings that CLIP behaves like a bag of words model and performs better with nouns and verbs; we also uncover novel insights such as CLIP getting confused by concrete words. Our framework is available at https://github.com/MichiganNLP/Scalable-VLM-Probing and can be used with other multimodal models and benchmarks.
We propose fill-in-the-blanks as a video understanding evaluation framework and introduce FIBER – a novel dataset consisting of 28,000 videos and descriptions in support of this evaluation framework. The fill-in-the-blanks setting tests a model’s understanding of a video by requiring it to predict a masked noun phrase in the caption of the video, given the video and the surrounding text. The FIBER benchmark does not share the weaknesses of the current state-of-the-art language-informed video understanding tasks, namely: (1) video question answering using multiple-choice questions, where models perform relatively well because they exploit linguistic biases in the task formulation, thus making our framework challenging for the current state-of-the-art systems to solve; and (2) video captioning, which relies on an open-ended evaluation framework that is often inaccurate because system answers may be perceived as incorrect if they differ in form from the ground truth. The FIBER dataset and our code are available at https://lit.eecs.umich.edu/fiber/.
Existing video understanding datasets mostly focus on human interactions, with little attention being paid to the “in the wild” settings, where the videos are recorded outdoors. We propose WILDQA, a video understanding dataset of videos recorded in outside settings. In addition to video question answering (Video QA), we also introduce the new task of identifying visual support for a given question and answer (Video Evidence Selection). Through evaluations using a wide range of baseline models, we show that WILDQA poses new challenges to the vision and language research communities. The dataset is available at https: //lit.eecs.umich.edu/wildqa/.
We aim to automatically identify human action reasons in online videos. We focus on the widespread genre of lifestyle vlogs, in which people perform actions while verbally describing them. We introduce and make publicly available the WhyAct dataset, consisting of 1,077 visual actions manually annotated with their reasons. We describe a multimodal model that leverages visual and textual information to automatically infer the reasons corresponding to an action presented in the video.
We introduce LifeQA, a benchmark dataset for video question answering that focuses on day-to-day real-life situations. Current video question answering datasets consist of movies and TV shows. However, it is well-known that these visual domains are not representative of our day-to-day lives. Movies and TV shows, for example, benefit from professional camera movements, clean editing, crisp audio recordings, and scripted dialog between professional actors. While these domains provide a large amount of data for training models, their properties make them unsuitable for testing real-life question answering systems. Our dataset, by contrast, consists of video clips that represent only real-life scenarios. We collect 275 such video clips and over 2.3k multiple-choice questions. In this paper, we analyze the challenging but realistic aspects of LifeQA, and we apply several state-of-the-art video question answering models to provide benchmarks for future research. The full dataset is publicly available at https://lit.eecs.umich.edu/lifeqa/.
This paper presents the development of a corpus of 30,000 Spanish tweets that were crowd-annotated with humor value and funniness score. The corpus contains approximately 38.6% of humorous tweets with an average score of 2.04 in a scale from 1 to 5 for the humorous tweets. The corpus has been used in an automatic humor recognition and analysis competition, obtaining encouraging results from the participants.
Sarcasm is often expressed through several verbal and non-verbal cues, e.g., a change of tone, overemphasis in a word, a drawn-out syllable, or a straight looking face. Most of the recent work in sarcasm detection has been carried out on textual data. In this paper, we argue that incorporating multimodal cues can improve the automatic classification of sarcasm. As a first step towards enabling the development of multimodal approaches for sarcasm detection, we propose a new sarcasm dataset, Multimodal Sarcasm Detection Dataset (MUStARD), compiled from popular TV shows. MUStARD consists of audiovisual utterances annotated with sarcasm labels. Each utterance is accompanied by its context of historical utterances in the dialogue, which provides additional information on the scenario where the utterance occurs. Our initial results show that the use of multimodal information can reduce the relative error rate of sarcasm detection by up to 12.9% in F-score when compared to the use of individual modalities. The full dataset is publicly available for use at https://github.com/soujanyaporia/MUStARD.
Computational Humor involves several tasks, such as humor recognition, humor generation, and humor scoring, for which it is useful to have human-curated data. In this work we present a corpus of 27,000 tweets written in Spanish and crowd-annotated by their humor value and funniness score, with about four annotations per tweet, tagged by 1,300 people over the Internet. It is equally divided between tweets coming from humorous and non-humorous accounts. The inter-annotator agreement Krippendorff’s alpha value is 0.5710. The dataset is available for general usage and can serve as a basis for humor detection and as a first step to tackle subjectivity.
False friends are words in two languages that look or sound similar, but have different meanings. They are a common source of confusion among language learners. Methods to detect them automatically do exist, however they make use of large aligned bilingual corpora, which are hard to find and expensive to build, or encounter problems dealing with infrequent words. In this work we propose a high coverage method that uses word vector representations to build a false friends classifier for any pair of languages, which we apply to the particular case of Spanish and Portuguese. The required resources are a large corpus for each language and a small bilingual lexicon for the pair.