Expectation-Regulated Neural Model for Event Mention Extraction

Abstract

We tackle the task of extracting tweets that mention a specific event from all tweets that contain relevant keywords, for which the main challenges include unbalanced positive and negative cases, and the unavailability of manually labeled training data. Existing methods leverage a few manually given seed events and large unlabeled tweets to train a classifier, by using expectation regularization training with discrete ngram features. We propose a LSTM-based neural model that learns tweet-level features automatically. Compared with discrete ngram features, the neural model can potentially capture non-local dependencies and deep semantic information, which are more effective for disambiguating subtle semantic differences between true event mentions and false cases that use similar wording patterns. Results on both tweets and forum posts show that our neural model is more effective compared with a state-of-the-art discrete baseline.

Detecting Cybersecurity Events from Noisy Short Text

“…Roy et al (2017) proposed a method to learn domain-specific word embeddings for sparse cyber security text. Prior art in this direction (Ritter et al, 2015;Chang et al, 2016) focuses on extracting events and in particular predicting the events' posterior given the presence of particular words. Le Sceller et al (2017); Tonon et al (2017) focus on detecting cyber security events from Twitter.…”

“…Recently, Ritter et al (2015) introduced the first study to extract event mentions from a raw Twitter stream for event categories DDoS attacks, data breaches, and account hijacking. Chang et al (2016) proposed an LSTM based approach which learns tweet level features automatically to extract events from tweet mentions. Lately, Le Sceller et al (2017) proposed a model to detect cyber security events in Twitter which uses a taxonomy and a set of seed keywords to retrieve relevant tweets.…”

Detecting Denial-of-Service Attacks from Social Media Text: Applying NLP to Computer Security

“…The main thread in this area is the learning model from Ritter et al (Ritter et al, 2015) and follow-on work (Kergl, 2015;Chang et al, 2016 collected tweets that contain the word 'DDoS', and then collected a set of known network attack days. The known days provided a training set from which they trained this weakly-supervised classifier on the 'DDoS' tweets.…”

“…The main thread in this area is the learning model from Ritter et al (Ritter et al, 2015) and follow-on work (Kergl, 2015;Chang et al, 2016). They proposed a weakly supervised learner to identify cybersecurity events from Twitter.…”

TweeTime : A Minimally Supervised Method for Recognizing and Normalizing Time Expressions in Twitter

“…1 Our code and data are publicly available at https:// github.com/jeniyat/TweeTime. Social media especially contains time-sensitive information and requires accurate temporal analysis, for example, for detecting real-time cybersecurity events (Ritter et al, 2015;Chang et al, 2016), disease outbreaks (Kanhabua et al, 2012) and extracting personal information (Schwartz et al, 2015). However, most work on social media simply uses generic temporal resolvers and therefore suffers from suboptimal performance.…”