Open-domain KeyPhrase Extraction (KPE) aims to extract keyphrases from documents without domain or quality restrictions, e.g., web pages with variant domains and qualities. Recently, neural methods have shown promising results in many KPE tasks due to their powerful capacity for modeling contextual semantics of the given documents. However, we empirically show that most neural KPE methods prefer to extract keyphrases with good phraseness, such as short and entity-style n-grams, instead of globally informative keyphrases from open-domain documents. This paper presents JointKPE, an open-domain KPE architecture built on pre-trained language models, which can capture both local phraseness and global informativeness when extracting keyphrases. JointKPE learns to rank keyphrases by estimating their informativeness in the entire document and is jointly trained on the keyphrase chunking task to guarantee the phraseness of keyphrase candidates. Experiments on two large KPE datasets with diverse domains, OpenKP and KP20k, demonstrate the effectiveness of JointKPE on different pre-trained variants in open-domain scenarios. Further analyses reveal the significant advantages of JointKPE in predicting long and non-entity keyphrases, which are challenging for previous neural KPE methods. Our code is publicly available at https://github.com/thunlp/BERT-KPE.
Multimedia content in social media platforms provides significant information during disaster events. The types of information shared include reports of injured or deceased people, infrastructure damage, and missing or found people, among others. Although many studies have shown the usefulness of both text and image content for disaster response purposes, the research has been mostly focused on analyzing only the text modality in the past. In this paper, we propose to use both text and image modalities of social media data to learn a joint representation using state-of-the-art deep learning techniques. Specifically, we utilize convolutional neural networks to define a multimodal deep learning architecture with a modality-agnostic shared representation. Extensive experiments on real-world disaster datasets show that the proposed multimodal architecture yields better performance than models trained using a single modality (e.g., either text or image).
Time-critical analysis of social media streams is important for humanitarian organizations for planing rapid response during disasters. The \textit{crisis informatics} research community has developed several techniques and systems for processing and classifying big crisis-related data posted on social media. However, due to the dispersed nature of the datasets used in the literature (e.g., for training models), it is not possible to compare the results and measure the progress made towards building better models for crisis informatics tasks. In this work, we attempt to bridge this gap by combining various existing crisis-related datasets. We consolidate eight human-annotated datasets and provide 166.1k and 141.5k tweets for \textit{informativeness} and \textit{humanitarian} classification tasks, respectively. We believe that the consolidated dataset will help train more sophisticated models. Moreover, we provide benchmarks for both binary and multiclass classification tasks using several deep learning architecrures including, CNN, fastText, and transformers. We make the dataset and scripts available at: https://crisisnlp.qcri.org/crisis_datasets_benchmarks.html
Standard informativeness measures used to evaluate Automatic Text Summarization mostly rely on n-gram overlapping between the automatic summary and the reference summaries. These measures differ from the metric they use (cosine, ROUGE, Kullback-Leibler, Logarithm Similarity, etc.) and the bag of terms they consider (single words, word n-grams, entities, nuggets, etc.). Recent word embedding approaches offer a continuous alternative to discrete approaches based on the presence/absence of a text unit. Informativeness measures have been extended to Focus Information Retrieval evaluation involving a user's information need represented by short queries. In particular for the task of CLEF-INEX Tweet Contextualization, tweet contents have been considered as queries. In this paper we define the concept of Interestingness as a generalization of Informativeness, whereby the information need is diverse and formalized as an unknown set of implicit queries. We then study the ability of state of the art Informativeness measures to cope with this generalization. Lately we show that with this new framework, standard word embeddings outperforms discrete measures only on uni-grams, however bi-grams seems to be a key point of interestingness evaluation. Lastly we prove that the CLEF-INEX Tweet Contextualization 2012 Logarithm Similarity measure provides best results.
Most existing work on automated fact checking is concerned with predicting the veracity of claims based on metadata, social network spread, language used in claims, and, more recently, evidence supporting or denying claims. A crucial piece of the puzzle that is still missing is to understand how to automate the most elaborate part of the process -- generating justifications for verdicts on claims. This paper provides the first study of how these explanations can be generated automatically based on available claim context, and how this task can be modelled jointly with veracity prediction. Our results indicate that optimising both objectives at the same time, rather than training them separately, improves the performance of a fact checking system. The results of a manual evaluation further suggest that the informativeness, coverage and overall quality of the generated explanations are also improved in the multi-task model.
Multi-Sentence Compression (MSC) aims to generate a short sentence with the key information from a cluster of similar sentences. MSC enables summarization and question-answering systems to generate outputs combining fully formed sentences from one or several documents. This paper describes an Integer Linear Programming method for MSC using a vertex-labeled graph to select different keywords, with the goal of generating more informative sentences while maintaining their grammaticality. Our system is of good quality and outperforms the state of the art for evaluations led on news datasets in three languages: French, Portuguese and Spanish. We led both automatic and manual evaluations to determine the informativeness and the grammaticality of compressions for each dataset. In additional tests, which take advantage of the fact that the length of compressions can be modulated, we still improve ROUGE scores with shorter output sentences.
Supervised deep learning requires a large amount of training samples with annotations (e.g. label class for classification task, pixel- or voxel-wised label map for segmentation tasks), which are expensive and time-consuming to obtain. During the training of a deep neural network, the annotated samples are fed into the network in a mini-batch way, where they are often regarded of equal importance. However, some of the samples may become less informative during training, as the magnitude of the gradient start to vanish for these samples. In the meantime, other samples of higher utility or hardness may be more demanded for the training process to proceed and require more exploitation. To address the challenges of expensive annotations and loss of sample informativeness, here we propose a novel training framework which adaptively selects informative samples that are fed to the training process. The adaptive selection or sampling is performed based on a hardness-aware strategy in the latent space constructed by a generative model. To evaluate the proposed training framework, we perform experiments on three different datasets, including MNIST and CIFAR-10 for image classification task and a medical image dataset IVUS for biophysical simulation task. On all three datasets, the proposed framework outperforms a random sampling method, which demonstrates the effectiveness of proposed framework.
In many real-world machine learning applications, unlabeled samples are easy to obtain, but it is expensive and/or time-consuming to label them. Active learning is a common approach for reducing this data labeling effort. It optimally selects the best few samples to label, so that a better machine learning model can be trained from the same number of labeled samples. This paper considers active learning for regression (ALR) problems. Three essential criteria -- informativeness, representativeness, and diversity -- have been proposed for ALR. However, very few approaches in the literature have considered all three of them simultaneously. We propose three new ALR approaches, with different strategies for integrating the three criteria. Extensive experiments on 12 datasets in various domains demonstrated their effectiveness.
Neural conversational models learn to generate responses by taking into account the dialog history. These models are typically optimized over the query-response pairs with a maximum likelihood estimation objective. However, the query-response tuples are naturally loosely coupled, and there exist multiple responses that can respond to a given query, which leads the conversational model learning burdensome. Besides, the general dull response problem is even worsened when the model is confronted with meaningless response training instances. Intuitively, a high-quality response not only responds to the given query but also links up to the future conversations, in this paper, we leverage the query-response-future turn triples to induce the generated responses that consider both the given context and the future conversations. To facilitate the modeling of these triples, we further propose a novel encoder-decoder based generative adversarial learning framework, Posterior Generative Adversarial Network (Posterior-GAN), which consists of a forward and a backward generative discriminator to cooperatively encourage the generated response to be informative and coherent by two complementary assessment perspectives. Experimental results demonstrate that our method effectively boosts the informativeness and coherence of the generated response on both automatic and human evaluation, which verifies the advantages of considering two assessment perspectives.
We propose a new framework for reasoning about information in complex systems. Our foundation is based on a variational extension of Shannon's information theory that takes into account the modeling power and computational constraints of the observer. The resulting \emph{predictive $\mathcal{V}$-information} encompasses mutual information and other notions of informativeness such as the coefficient of determination. Unlike Shannon's mutual information and in violation of the data processing inequality, $\mathcal{V}$-information can be created through computation. This is consistent with deep neural networks extracting hierarchies of progressively more informative features in representation learning. Additionally, we show that by incorporating computational constraints, $\mathcal{V}$-information can be reliably estimated from data even in high dimensions with PAC-style guarantees. Empirically, we demonstrate predictive $\mathcal{V}$-information is more effective than mutual information for structure learning and fair representation learning.
Sliced-Wasserstein distance (SW) and its variant, Max Sliced-Wasserstein distance (Max-SW), have been used widely in the recent years due to their fast computation and scalability even when the probability measures lie in a very high dimensional space. However, SW requires many unnecessary projection samples to approximate its value while Max-SW only uses the most important projection, which ignores the information of other useful directions. In order to account for these weaknesses, we propose a novel distance, named Distributional Sliced-Wasserstein distance (DSW), that finds an optimal distribution over projections that can balance between exploring distinctive projecting directions and the informativeness of projections themselves. We show that the DSW is a generalization of Max-SW, and it can be computed efficiently by searching for the optimal push-forward measure over a set of probability measures over the unit sphere satisfying certain regularizing constraints that favor distinct directions. Finally, we conduct extensive experiments with large-scale datasets to demonstrate the favorable performances of the proposed distances over the previous sliced-based distances in generative modeling applications.
Learning the preferences of a human improves the quality of the interaction with the human. The number of queries available to learn preferences maybe limited especially when interacting with a human, and so active learning is a must. One approach to active learning is to use uncertainty sampling to decide the informativeness of a query. In this paper, we propose a modification to uncertainty sampling which uses the expected output value to help speed up learning of preferences. We compare our approach with the uncertainty sampling baseline, as well as conduct an ablation study to test the validity of each component of our approach.
In this work, we take a closer look at the evaluation of two families of methods for enriching information from knowledge graphs: Link Prediction and Entity Alignment. In the current experimental setting, multiple different scores are employed to assess different aspects of model performance. We analyze the informativeness of these evaluation measures and identify several shortcomings. In particular, we demonstrate that all existing scores can hardly be used to compare results across different datasets. Moreover, we demonstrate that varying size of the test size automatically has impact on the performance of the same model based on commonly used metrics for the Entity Alignment task. We show that this leads to various problems in the interpretation of results, which may support misleading conclusions. Therefore, we propose adjustments to the evaluation and demonstrate empirically how this supports a fair, comparable, and interpretable assessment of model performance. Our code is available at https://github.com/mberr/rank-based-evaluation.
Sentence compression is the task of compressing a long sentence into a short one by deleting redundant words. In sequence-to-sequence (Seq2Seq) based models, the decoder unidirectionally decides to retain or delete words. Thus, it cannot usually explicitly capture the relationships between decoded words and unseen words that will be decoded in the future time steps. Therefore, to avoid generating ungrammatical sentences, the decoder sometimes drops important words in compressing sentences. To solve this problem, we propose a novel Seq2Seq model, syntactically look-ahead attention network (SLAHAN), that can generate informative summaries by explicitly tracking both dependency parent and child words during decoding and capturing important words that will be decoded in the future. The results of the automatic evaluation on the Google sentence compression dataset showed that SLAHAN achieved the best kept-token-based-F1, ROUGE-1, ROUGE-2 and ROUGE-L scores of 85.5, 79.3, 71.3 and 79.1, respectively. SLAHAN also improved the summarization performance on longer sentences. Furthermore, in the human evaluation, SLAHAN improved informativeness without losing readability.
Set functions are functions (or signals) indexed by the powerset (set of all subsets) of a finite set N. They are fundamental and ubiquitous in many application domains and have been used, for example, to formally describe or quantify loss functions for semantic image segmentation, the informativeness of sensors in sensor networks the utility of sets of items in recommender systems, cooperative games in game theory, or bidders in combinatorial auctions. In particular, the subclass of submodular functions occurs in many optimization and machine learning problems. In this paper, we derive discrete-set signal processing (SP), a novel shift-invariant linear signal processing framework for set functions. Discrete-set SP considers different notions of shift obtained from set union and difference operations. For each shift it provides associated notions of shift-invariant filters, convolution, Fourier transform, and frequency response. We provide intuition for our framework using the concept of generalized coverage function that we define, identify multivariate mutual information as a special case of a discrete-set spectrum, and motivate frequency ordering. Our work brings a new set of tools for analyzing and processing set functions, and, in particular, for dealing with their exponential nature. We show two prototypical applications and experiments: compression in submodular function optimization and sampling for preference elicitation in combinatorial auctions.
In this work, we study the problem of federated learning (FL), where distributed users aim to jointly train a machine learning model with the help of a parameter server (PS). In each iteration of FL, users compute local gradients, followed by transmission of the quantized gradients for subsequent aggregation and model updates at PS. One of the challenges of FL is that of communication overhead due to FL's iterative nature and large model sizes. One recent direction to alleviate communication bottleneck in FL is to let users communicate simultaneously over a multiple access channel (MAC), possibly making better use of the communication resources. In this paper, we consider the problem of FL learning over a MAC. In particular, we focus on the design of digital gradient transmission schemes over a MAC, where gradients at each user are first quantized, and then transmitted over a MAC to be decoded individually at the PS. When designing digital FL schemes over MACs, there are new opportunities to assign different amount of resources (such as rate or bandwidth) to different users based on a) the informativeness of the gradients at each user, and b) the underlying channel conditions. We propose a stochastic gradient quantization scheme, where the quantization parameters are optimized based on the capacity region of the MAC. We show that such channel aware quantization for FL outperforms uniform quantization, particularly when users experience different channel conditions, and when have gradients with varying levels of informativeness.
The automatic summarization of multimedia sources is an important task that facilitates the understanding of an individual by condensing the source while maintaining relevant information. In this paper we focus on audio summarization based on audio features and the probability of distribution divergence. Our method, based on an extractive summarization approach, aims to select the most relevant segments until a time threshold is reached. It takes into account the segment's length, position and informativeness value. Informativeness of each segment is obtained by mapping a set of audio features issued from its Mel-frequency Cepstral Coefficients and their corresponding Jensen-Shannon divergence score. Results over a multi-evaluator scheme shows that our approach provides understandable and informative summaries.
In many real-world machine learning applications, unlabeled data can be easily obtained, but it is very time-consuming and/or expensive to label them. So, it is desirable to be able to select the optimal samples to label, so that a good machine learning model can be trained from a minimum amount of labeled data. Active learning (AL) has been widely used for this purpose. However, most existing AL approaches are supervised: they train an initial model from a small amount of labeled samples, query new samples based on the model, and then update the model iteratively. Few of them have considered the completely unsupervised AL problem, i.e., starting from zero, how to optimally select the very first few samples to label, without knowing any label information at all. This problem is very challenging, as no label information can be utilized. This paper studies unsupervised pool-based AL for linear regression problems. We propose a novel AL approach that considers simultaneously the informativeness, representativeness, and diversity, three essential criteria in AL. Extensive experiments on 14 datasets from various application domains, using three different linear regression models (ridge regression, LASSO, and linear support vector regression), demonstrated the effectiveness of our proposed approach.
This article presents a stochastic corpus-based model for generating natural language text. Our model first encodes dependency relations from training data through a feature set, then concatenates these features to produce a new dependency tree for a given meaning representation, and finally generates a natural language utterance from the produced dependency tree. We test our model on nine domains from tabular, dialogue act and RDF format. Our model outperforms the corpus-based state-of-the-art methods trained on tabular datasets and also achieves comparable results with neural network-based approaches trained on dialogue act, E2E and WebNLG datasets for BLEU and ERR evaluation metrics. Also, by reporting Human Evaluation results, we show that our model produces high-quality utterances in aspects of informativeness and naturalness as well as quality.
The performance of multi-domain image-to-image translation has been significantly improved by recent progress in deep generative models. Existing approaches can use a unified model to achieve translations between all the visual domains. However, their outcomes are far from satisfying when there are large domain variations. In this paper, we reveal that improving the sample selection strategy is an effective solution. To select informative samples, we dynamically estimate sample importance during the training of Generative Adversarial Networks, presenting Informative Sample Mining Network. We theoretically analyze the relationship between the sample importance and the prediction of the global optimal discriminator. Then a practical importance estimation function for general conditions is derived. Furthermore, we propose a novel multi-stage sample training scheme to reduce sample hardness while preserving sample informativeness. Extensive experiments on a wide range of specific image-to-image translation tasks are conducted, and the results demonstrate our superiority over current state-of-the-art methods.