Digital Commons Network^™

Opinion Mining On Small And Noisy Samples Of Health-Related Texts, John Cardiff, Liliya Akhtyamova, Mikhail Alexandrov, Oleksiy Koshulko

Conference Papers

The topic of people’s health has always attracted the attention of public and private structures, the patients themselves and, therefore, researchers.

Social networks provide an immense amount of data for analysis of health- related issues; however it is not always the case that researchers have enough

data to build sophisticated models. In the paper, we artificially create this lim- itation to test performance and stability of different popular algorithms on small

samples of texts. There are two specificities in this research apart from the size of a sample: (a) here, instead of usual 5-star classification, we use combined classes reflecting …

Extracting Drug-Drug Interactions With Character-Level And Dependency-Based Embeddings, John Cardiff, Liliya Akhtyamova

Conference Papers

The DDI track of TAC-2018 challenge addresses the problem of an information retrieval of drug-drug interactions on structured product labeling documents with discontinuous and overlapping entities. In this paper, we present our participation for event extraction subtask (Task 1). We used a supervised long-short-term memory (LSTM) network with conditional random fields decoding (LSTM-CRF) approach with an automatic exploring of words and characters features. Additional dependency-based information was integrated into word embeddings to allow better word representation. Our system performed with above median score.

Building Classifiers With Gmdh For Health Social Networks (Bd Askapatient), John Cardiff, Liliya Akhtyamova, Mikhail Alexandrov

Conference Papers

Health social media offer useful data for patients and doctors concerning both various medicines and treatments. Usually, these data are accompanied by their assessments in 5- star scale. But such a detail classification has small usefulness because patients and doctors, first of all, want to know about negative cases and to study in detail the extreme ones. In the paper we build classifiers of texts just for these cases using combined classes as negative, all others and worst, satisfactory, best. For this, we study possibilities of different GMDH-based algorithms and compare them with the results of other methods. The selection …

Classifying Misogynistic Tweets Using A Blended Model: The Ami Shared Task In Ibereval 2018, John Cardiff, Elena Shushkevich

Conference Papers

This article describes a possible solution for Automatic Misogyny Identification (AMI) Shared Task at IBEREVAL-2018. The proposed technique is based on combining several simpler classifiers into one more complex blended model, which classified the data taking into account the probabilities of belonging to classes calculated by simpler models. We used the Logistic Regression, Naive Bayes, and SVM classifiers. The experimental results show that blended model works better than simpler models for all three type of classification, for both binomial classification (Misogyny Identifivation, Target Classification) and multinomial classification (Misogynistic Behavior).

Misogyny Detection And Classification In English Tweets: The Experience Of The Itt Team, John Cardiff, Elena Shushkevich

Conference Papers

The problem of online misogyny and women-based offending has become increasingly widespread, and the automatic detection of such messages is an urgent priority. In this paper, we present an approach based on an ensemble of Logistic Regression, Support Vector Machines, and Naïve Bayes models for the detection of misogyny in texts extracted from the Twitter platform. Our method has been presented in the framework of the participation in the Automatic Misogyny Identification (AMI) Shared Task in the EVALITA 2018 evaluation campaign.