As pointed out by several scholars, current research on hate speech (HS) recognition is characterized by unsystematic data creation strategies and diverging annotation schemata. Subsequently, supervised-learning models tend to generalize poorly to datasets they were not trained on, and the performance of the models trained on datasets labeled using different HS taxonomies cannot be compared. To ease this problem, we propose applying extremely weak supervision that only relies on the class name rather than on class samples from the annotated data. We demonstrate the effectiveness of a state-of-the-art weakly-supervised text classification model in various in-dataset and cross-dataset settings. Furthermore, we conduct an in-depth quantitative and qualitative analysis of the source of poor generalizability of HS classification models.

Ad creatives are ads served to users on a webpage, app, or other digital environments. The demand for compelling ad creatives surges drastically with the ever-increasing popularity of digital marketing. The two most essential elements of (display) ad creatives are the advertising message, such as headlines and description texts, and the visual component, such as images and videos. Traditionally, ad creatives are composed by professional copywriters and creative designers. The process requires significant human effort, limiting the scalability and efficiency of digital ad campaigns. This work introduces AUTOCREATIVE, a novel system to automatically generate ad creatives relying on natural language generation and computer vision techniques. The system generates multiple ad copies (ad headlines/description texts) using a sequence-to-sequence model and selects images most suitable to the generated ad copies based on heuristic-based visual appeal metrics and a text-image retrieval pipeline.

Contextual advertising provides advertisers with the opportunity to target the context which is most relevant to their ads. The large variety of potential topics makes it very challenging to collect training documents to build a supervised classification model or compose expert-written rules in a rule-based classification system. Besides, in fine-grained classification, different categories often overlap or co-occur, making it harder to classify accurately. In this work, we propose wiki2cat, a method to tackle large-scaled fine-grained text classification by tapping on the Wikipedia category graph. The categories in the IAB taxonomy are first mapped to category nodes in the graph. Then the label is propagated across the graph to obtain a list of labeled Wikipedia documents to induce text classifiers. The method is ideal for large-scale classification problems since it does not require any manually-labeled document or hand-curated rules or keywords. The proposed method is benchmarked with various learning-based and keyword-based baselines and yields competitive performance on publicly available datasets and a new dataset containing more than 300 fine-grained categories.