The scientific study of hate speech, from a computer science point of view, is recent. This survey organizes and describes the current state of the field, providing a structured overview of previous approaches, including core algorithms, methods, and main features used. This work also discusses the complexity of the concept of hate speech, defined in many platforms and contexts, and provides a unifying definition. This area has an unquestionable potential for societal impact, particularly in online communities and digital media platforms. The development and systematization of shared resources, such as guidelines, annotated datasets in multiple languages, and algorithms, is a crucial step in advancing the automatic detection of hate speech.

Over the past years, the amount of online offensive speech has been growing steadily. To successfully cope with it, machine learning is applied. However, ML-based techniques require sufficiently large annotated datasets. In the last years, different datasets were published, mainly for English. In this paper, we present a new dataset for Portuguese, which has not been in focus so far. The dataset is composed of 5,668 tweets. For its annotation, we defined two different schemes used by annotators with different levels of expertise. First, non-experts annotated the tweets with binary labels ('hate' vs. 'no-hate'). Then, expert annotators classified the tweets following a fine-grained hierarchical multiple label scheme with 81 hate speech categories in total. The inter-annotator agreement varied from category to category, which reflects the insight that some types of hate speech are more subtle than others and that their detection depends on personal perception. The hierarchical annotation scheme is the main contribution of the presented work, as it facilitates the identification of different types of hate speech and their intersections. To demonstrate the usefulness of our dataset, we carried a baseline classification experiment with pre-trained word embeddings and LSTM on the binary classified data, with a state-of-the-art outcome.

No abstract

a b s t r a c tIn the area of Information Retrieval, the task of automatic text summarization usually assumes a static underlying collection of documents, disregarding the temporal dimension of each document. However, in real world settings, collections and individual documents rarely stay unchanged over time. The World Wide Web is a prime example of a collection where information changes both frequently and significantly over time, with documents being added, modified or just deleted at different times. In this context, previous work addressing the summarization of web documents has simply discarded the dynamic nature of the web, considering only the latest published version of each individual document. This paper proposes and addresses a new challenge -the automatic summarization of changes in dynamic text collections. In standard text summarization, retrieval techniques present a summary to the user by capturing the major points expressed in the most recent version of an entire document in a condensed form. In this new task, the goal is to obtain a summary that describes the most significant changes made to a document during a given period. In other words, the idea is to have a summary of the revisions made to a document over a specific period of time. This paper proposes different approaches to generate summaries using extractive summarization techniques. First, individual terms are scored and then this information is used to rank and select sentences to produce the final summary. A system based on Latent Dirichlet Allocation model (LDA) is used to find the hidden topic structures of changes. The purpose of using the LDA model is to identify separate topics where the changed terms from each topic are likely to carry at least one significant change. The different approaches are then compared with the previous work in this area. A collection of articles from Wikipedia, including their revision history, is used to evaluate the proposed system. For each article, a temporal interval and a reference summary from the article's content are selected manually. The articles and intervals in which a significant event occurred are carefully selected. The summaries produced by each of the approaches are evaluated comparatively to the manual summaries using ROUGE metrics. It is observed that the approach using the LDA model outperforms all the other approaches. Statistical tests reveal that the differences in ROUGE scores for the LDA-based approach is statistically significant at 99% over baseline.

Time has been successfully used as a feature in web information retrieval tasks. In this context, estimating a document's inception date or last update date is a necessary task. Classic approaches have used HTTP header fields to estimate a document's last update time. The main problem with this approach is that it is applicable to a small part of web documents. In this work, we evaluate an alternative strategy based on a document's neighborhood. Using a random sample containing 10,000 URLs from the Yahoo! Directory, we study each document's links and media assets to determine its age. If we only consider isolated documents, we are able to date 52% of them. Including the document's neighborhood, we are able to estimate the date of more than 86% of the same sample. Also, we find that estimates differ significantly according to the type of neighbors used. The most reliable estimates are based on the document's media assets, while the worst estimates are based on incoming links. These results are experimentally evaluated with a real world application using different datasets.

No abstract

Purpose The quality of consumer-oriented health information on the web has been defined and evaluated in several studies. Usually it is based on evaluation criteria identified by the researchers and, so far, there is no agreed standard for the quality indicators to use. Based on such indicators, tools have been developed to evaluate the quality of web information. The HONcode is one of such tools. The purpose of this paper is to investigate the influence of web document features on their quality, using HONcode as ground truth, with the aim of finding whether it is possible to predict the quality of a document using its characteristics. Design/methodology/approach The present work uses a set of health documents and analyzes how their characteristics (e.g. web domain, last update, type, mention of places of treatment and prevention strategies) are associated with their quality. Based on these features, statistical models are built which predict whether health-related web documents have certification-level quality. Multivariate analysis is performed, using classification to estimate the probability of a document having quality given its characteristics. This approach tells us which predictors are important. Three types of full and reduced logistic regression models are built and evaluated. The first one includes every feature, without any exclusion, the second one disregards the Utilization Review Accreditation Commission variable, due to it being a quality indicator, and the third one excludes the variables related to the HONcode principles, which might also be indicators of quality. The reduced models were built with the aim to see whether they reach similar results with a smaller number of features. Findings The prediction models have high accuracy, even without including the characteristics of Health on the Net code principles in the models. The most informative prediction model considers characteristics that can be assessed automatically (e.g. split content, type, process of revision and place of treatment). It has an accuracy of 89 percent. Originality/value This paper proposes models that automatically predict whether a document has quality or not. Some of the used features (e.g. prevention, prognosis or treatment) have not yet been explicitly considered in this context. The findings of the present study may be used by search engines to promote high-quality documents. This will improve health information retrieval and may contribute to reduce the problems caused by inaccurate information.

In real-world information retrieval systems, the underlying document collection is rarely stable or definite. This work is focused on the study of signals extracted from the content of documents at different points in time for the purpose of weighting individual terms in a document. The basic idea behind our proposals is that terms that have existed for a longer time in a document should have a greater weight. We propose four term weighting functions that use each document's history to estimate a current term score. To evaluate this thesis, we conduct three independent experiments using a collection of documents sampled from Wikipedia. In the first experiment we use data from Wikipedia to judge each set of terms. In a second experiment we use an external collection of tags from a popular social bookmarking service as a gold standard. In the third experiment, we crowdsource user judgments to collect feedback on term preference. Across all experiments results consistently support our thesis. We show that temporally aware measures, specifically the proposed revision term frequency and revision term frequency span, outperform a term weighting measure based on raw term frequency alone.