RiSER: Learning Better Representations for Richly Structured Emails

Kocayusufoglu, Furkan; Sheng, Ying; Vo, Nguyen; Wendt, James B.; Zhao, Qian; Tata, Sandeep; Najork, Marc

doi:10.1145/3308558.3313720

Cited by 12 publications

(12 citation statements)

References 31 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Topic classification categorizes Web pages based on their topic or subject (e.g., whether a page is about "news" or a "movie") for topic-specific search engines and Web content management [22]. Approaches to topic classification range from using only the URL [1,11], to using the page content [5,27], to including the structure as well as the content [14].…”

Section: Web Page Classificationmentioning

confidence: 99%

“…Thus, if we train a model on a subset of pages from a repository R and then use pages from R in our test data, it will be too easy for the model to identify them. To prevent the effects of memorization [14], we ensured that pages from the same host were either all in the training or all in the test set.…”

Section: Balance In Test Vs Training Setmentioning

confidence: 99%

“…As an alternative to classification based only on content, we also considered RiSER [14], a model that incorporates both the structure and content of pages to classify machinegenerated emails into verticals (hotel, bill, etc.). Such emails are typically short, and have rich HTML structure.…”

Section: Vertical Classifier (Content-and Structure-based)mentioning

confidence: 99%

See 2 more Smart Citations

Dataset or Not? A Study on the Veracity of Semantic Markup for Dataset Pages

Alrashed

Paparas

Benjelloun

et al. 2021

The Semantic Web – ISWC 2021

Self Cite

View full text Add to dashboard Cite

Semantic markup, such as , allows providers on the Web to describe content using a shared controlled vocabulary. This markup is invaluable in enabling a broad range of applications, from vertical search engines, to rich snippets in search results, to actions on emails, to many others. In this paper, we focus on semantic markup for datasets, specifically in the context of developing a vertical search engine for datasets on the Web, Google’s Dataset Search. Dataset Search relies on to identify pages that describe datasets. While was the core enabling technology for this vertical search, we also discovered that we need to address the following problem: pages from 61% of internet hosts that provide markup do not actually describe datasets. We analyze the veracity of dataset markup for Dataset Search’s Web-scale corpus and categorize pages where this markup is not reliable. We then propose a way to drastically increase the quality of the dataset metadata corpus by developing a deep neural-network classifier that identifies whether or not a page with markup is a dataset page. Our classifier achieves 96.7% recall at the 95% precision point. This level of precision enables Dataset Search to circumvent the noise in semantic markup and to use the metadata to provide high quality results to users.

show abstract

Section: Web Page Classificationmentioning

confidence: 99%

Section: Balance In Test Vs Training Setmentioning

confidence: 99%

See 1 more Smart Citation

Dataset or Not? A Study on the Veracity of Semantic Markup for Dataset Pages

Alrashed

Paparas

Benjelloun

et al. 2021

The Semantic Web – ISWC 2021

Self Cite

View full text Add to dashboard Cite

show abstract

“…Multi-modal extraction: The incorporation of visual information into IE was proposed by Aumann et al (2006), who attempted to learn a fitness function to calculate the visual similarity of a document to one in its training set to extract elements like headlines and authors. Other recent approaches that attempt to address the layout structure of documents are CharGrid (Katti et al, 2018), which represents a document as a two-dimensional grid of characters, RiSER, an extraction technique targeted at templated emails (Kocayusufoglu et al, 2019), and that by Liu et al (2018), which presents an RNN method for learning DOM-tree rules. However, none of these address the OpenIE setting, which requires understanding the relationship between different text fields on the page.…”

Section: Related Workmentioning

confidence: 99%

ZeroShotCeres: Zero-Shot Relation Extraction from Semi-Structured Webpages

Lockard

Shiralkar

Dong

et al. 2020

Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics

View full text Add to dashboard Cite

In many documents, such as semi-structured webpages, textual semantics are augmented with additional information conveyed using visual elements including layout, font size, and color. Prior work on information extraction from semi-structured websites has required learning an extraction model specific to a given template via either manually labeled or distantly supervised data from that template. In this work, we propose a solution for "zero-shot" open-domain relation extraction from webpages with a previously unseen template, including from websites with little overlap with existing sources of knowledge for distant supervision and websites in entirely new subject verticals. Our model uses a graph neural network-based approach to build a rich representation of text fields on a webpage and the relationships between them, enabling generalization to new templates. Experiments show this approach provides a 31% F1 gain over a baseline for zero-shot extraction in a new subject vertical.

show abstract

“…With the increase in machine-generated emails, recent studies have shifted their focus away from dialogs and towards parsing and categorizing (Aberdeen et al, 2010;Zhang et al, 2017) or threading notifications (Ailon et al, 2013), as well as automated template induction (Proskurnia et al, 2017;Castro et al, 2018;Kocayusufoglu et al, 2019).…”

Section: Related Workmentioning

confidence: 99%

Crawling and Preprocessing Mailing Lists At Scale for Dialog Analysis

Bevendorff

Khatib

Potthast

et al. 2020

Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics

View full text Add to dashboard Cite

This paper introduces the Webis Gmane EmailCorpus 2019, the largest publicly available and fully preprocessed email corpus to date. We crawled more than 153 million emails from 14,699 mailing lists and segmented them into semantically consistent components using a new neural segmentation model. With 96% accuracy on 15 classes of email segments, our model achieves state-of-the-art performance while being more efficient to train than previous ones. All data, code, and trained models are made freely available alongside the paper. 1

show abstract

RiSER: Learning Better Representations for Richly Structured Emails

Cited by 12 publications

References 31 publications

Dataset or Not? A Study on the Veracity of Semantic Markup for Dataset Pages

Dataset or Not? A Study on the Veracity of Semantic Markup for Dataset Pages

ZeroShotCeres: Zero-Shot Relation Extraction from Semi-Structured Webpages

Crawling and Preprocessing Mailing Lists At Scale for Dialog Analysis

Contact Info

Product

Resources

About