Topic-driven Ensemble for Online Advertising Generation

Proceedings of the 28th International Conference on Computational Linguistics, pages 2273–2283 Barcelona, Spain (Online), December 8-13, 2020 2273 Topic-driven Ensemble for Online Advertising Generation Egor Nevezhin ITMO University egornevezhin@itmo.ru Nikolay Butakov ITMO University alipoov.nb@gmail.com Maria Khodorchenko ITMO University mariyaxod@yandex.ru Maxim Petrov ITMO University djvipmax@gmail.com Denis Nasonov ITMO University denis.nasonov@gmail.com Abstract Online advertising is one of the most widespread ways to reach and increase a target audience for those selling products. Usually having a form of a banner, advertising engages users into visiting a corresponding webpage. Professional generation of banners requires creative and writing skills and a basic understanding of target products. The great variety of goods presented in the online market enforce professionals to spend more and more time creating new advertisements different from existing ones. In this paper, we propose a neural network-based approach for the automatic generation of online advertising using texts from given webpages as sources. The important part of the approach is training on open data available online, which allows avoiding costly procedures of manual labeling. Collected open data consist of multiple subdomains with high data heterogeneity. The subdomains belong to different topics and vary in used vocabularies, phrases, styles that lead to reduced quality in adverts generation. We try to solve the problem of identifying existed subdomains and proposing a new ensemble approach based on exploiting multiple instances of a seq2seq model. Our experimental study on a dataset in the Russian language shows that our approach can significantly improve the quality of adverts generation. 1 Introduction The task of abstractive summarization assumes getting short but still a semantically relevant represen- tation of some text. The summarization helps advertisement to catch user attention, get him or her interested, and finally to engage in the full text reading. This is especially true for online advertising when it flashes in front of eyes on most website in the Internet. Crafting such short texts for online advertising requires a level of mastering and some degree of subject understanding. Humans doing this job have to read at least the base text, which describes products or services and then to spend additional time thinking of suitable describing phrase. That may be very time-consuming and thus a good option for automation. Modern natural language processing methods and techniques allow solving the task up to some degree of quality while still leaving space for improvement. The most advanced methods are based on neu- ral network seq2seq models (Gehrmann et al., 2018) and consume large manually labeled datasets for supervised learning. Due to being so popular and widespread, online advertising gives us an opportunity to obtain labeled data (in this particular case, a pair of the source text and banner text) using only data crawling techniques significantly decreasing human efforts needed to build a sufficient dataset for the task. At the same time, ollected open data may consist of multiple subdomains with high data heterogeneity. Moreover, what data will come from the crawling depends on advertising providers and their strategies adopted to the end-user. …

Content truncated to 500 words