English. Sources, in the form of selected Facebook pages, can be used as indicators of hate-rich content. Polarized distributed representations created over such content prove superior to generic embeddings in the task of hate speech detection. The same content seems to carry a too weak signal to proxy silver labels in a distant supervised setting. However, this signal is stronger than gold labels which come from a different distribution, leading to re-think the process of annotation in the context of highly subjective judgments.
Hate speech is “the use of aggressive, hatred or
offensive language, targeting a specific group of
people sharing a common trait: their gender,
ethnic group, race, religion, sexual orientation, or
disability” (Merriam-Webster’s collegiate dictionary,
1999). The phenomenon is widely spread on-line,
and Italian Social Media is definitely not an
exception
The Italian NLP community is active on this
front
Introduced as a new take on data annotation
2The EVALITA 2018 haspeede task addresses this issue by setting the task in a cross-genre fashion. 2012). This reduces the risk of adding extra bias (see also point (iii) about limitation in the previous paragraph), modulo the choices related to which proxies should be considered.
cial take on distant supervision where we use as proxies the sources where the content is published on-line rather than any hint in the content itself. Through a battery of experiments on hate speech detection in Italian we show that this approach yields meaningful representations and an increase in performance over the use of generic representations. Contextually, we show the limitations of silver labels, but also of gold labels that come from a different dataset with respect to the evaluation set. 2
Our approach is based on previous studies on
on-line communities showing that communities
tend to reinforce themselves, enhancing “filter
bubbles” effects, decreasing diversity, distorting
information, and polarizing socio-political
opinions
Building on this principle, we scraped data from social media communities on Facebook, acquiring what we call source-driven representations. The data is indeed used in two ways in the context of Hate Speech detection, namely: i.) to generate (potentially) polarized word embeddings to be used in a variety of models, comparing it to more standard generic embeddings (Section 3); and ii.) as training data for a supervised machine learning classifier, combining and comparing it with manually labeled data (Section 4). 3
Polarized embeddings are representations built on a corpus which is not randomly representative of the Italian language, rather collected with a specific bias. In this context, we use data scraped from Facebook pages (communities) in order to create hate-rich embeddings.
Data acquisition We selected a set of publicly available Facebook pages that may promote or be the target of hate speech, such as pages known for promoting nationalism (Italia Patria Mia), controversies (Dagospia, La Zanzara - Radio 24), hate against migrants and other minorities (La Fabbrica Del Degrado, Il Redpillatore, Cloroformio), support for women and LGBT rights (NON UNA DI MENO, LGBT News Italia). Using the Facebook API, we downloaded the comments to posts as they are the text portions most likely to express hate, collecting a total of over 1M comments for almost 13M tokens (Table 1).
Page Name Matteo Salvini NON UNA DI MENO LGBT News Italia Italia Patria Mia Dagospia La Fabbrica Del Degrado Boom. Friendzoned.
Cloroformio
Il Redpillatore
Sesso Droga e Pastorizia
PSDM
Cara, sei femminista - Returned
Se solo avrei studiato
La Zanzara - Radio 24
Total
Making Embeddings We built distributed
representations over the acquired data. The
embeddings have been generated with the word2vec 3
skip-gram model
These hate-rich embeddings are used in
models for hate speech detection. For comparison,
we also use larger, generic embeddings that were
trained on the Italian Wikipedia (more than 300M
tokens)4 using GloVe
4http://hlt.isti.cnr.it/ wordembeddings/
5https://nlp.stanford.edu/projects/ glove/ and “trans”. For the former, it is interesting to see how the polarized embeddings return more hateleaning words compared to the generic embeddings. For the latter, in addition to hateful epithets, we also see how these embeddings capture the correct semantic field, while the generic ones do not. Classification To test the contribution of our embeddings, we used them in two different classifiers, comparing them to alternative distributed representations.
First, we built a Convolutional Neural
Network (CNN), using the implementation of
Second, and for further comparison, we also
built a simple Linear Support Vector Machine
(SVM), using the LinearSVC scikit learn
implementation
We train and test our models using the manually labelled data provided in the context of the EVALITA 2018 task on Hate Speech Detection (haspeede) 6. The released training/development set comprises 3000 Facebook comments and 3000 tweets. The proportion of hateful content in this dataset is 39%, with 46% in the Facebook portion, and 32% in Twitter. We train on 80% of haspeede (4800 instances), and test on the remaining 20%. We report precision, recall, and F-score per class, averaged over ten random train/test splits. To assess general performance, we use macro F-score rather than micro F-score as the classifier’s accuracy on the minority class is particularly important. This is also reported as the average of the ten different runs. Results The results in Table 3 show that despite our embeddings being almost 25 times smaller than the generic ones, they yield a substantially better performance both in the CNN model and in the SVM classifier. In the former, they are also more informative than the representations obtained on-the-fly from the training data. In the latter, the contribution of embeddings in general appears though rather marginal on top of a more standard SVM model based on n-gram tf-idf information, and the difference according to which representation is used is not significant. Finally, it is interesting to note that the polarized embeddings cover 55% of the tokens in the training data (vs. only 45% of the generic ones, in spite of the substantial size difference between the two. 4
In a more standard distantly supervised setting, modulo proxing labels via sources rather than specific keywords/emojis, we also used the scraped text as training data directly. Because we approximate labels with sources, and we had collected data from supposedly hate-rich pages, for the current experimental settings we balanced the data by 6http://www.di.unito.it/˜tutreeb/ haspeede-evalita18/index.html F MACRO F .749 .760 .786 .728 .750 .806 .807 .802 evaluation, we use the same settings as the experiments in Section 3, by picking a random test set out of the haspeede dataset ten times, and reporting averaged results.
Results From Table 4 we can make the following observations: (i) training on silver labels lets us detect hate speech better than a most-frequentlabel baseline (macro F=:383); (ii) however, in this context, training on small amounts of gold data is substantially more accurate than training on large amounts of distantly supervised data (:807 vs :464); (iii) adding even small amounts of silver data to gold decreases performance (:792 vs :807)8; (iv) also adding more gold data decreases performance, even more so than adding an equal amount of silver data, if the manually labeled data comes from a different dataset (thus created with different guidelines, and in this case with a different hate/non-hate distribution). Performance goes up as expected when adding more data from the same dataset (:814 vs :807). 5
We exploited distant supervision to automatically obtain representations from Facebook-scraped content in two forms. First, we generated polarized, hate-rich distributed representations which proved superior to larger, generic embeddings when used both in a CNN and an SVM model for hate speech detection. Second, we used the scraped data as training material directly, proxing scraping Facebook comments from an Italian news agency (i.e. ANSA), assuming it conveys neutral content rather than polarized.
As for the distribution of labels, we followed the proportion of the Facebook portion of the haspeede dataset (46% of hateful content, and the rest non-polarized). We proxy labels according to sources, and under the above presumed proportions, we selected a total of 100,000 comments.
For comparison, and in combination, we also
used gold data. In addition to the previously
mentioned 6000 instances from the haspeede task,
we used the Turin dataset, a collection of
990 manually labelled tweets concerning the topic
of immigration, religion and Roma7
We trained an SVM classifier with the best settings as observed in Section 3 (tf-idf and and polarised embeddings) using different training sets, combining gold and silver data (see Table 4). For 7The Romani, Romany, or Roma are an ethnic group of traditionally itinerant people who originated in northern India and are nowadays subject to ethnic discrimination. 8We also experimented with adding progressively larger batches of silver data to gold (2K, 3K, 5K, etc.), but this yielded a steady decrease in performance. labels (hate vs non-hate) with the sources where the data was coming from (Facebook pages). This did not prove as a successful alternative nor complementary strategy to using gold data, though performance above baseline indicates some signal is present. Importantly, though, our experiments also suggest that gold data is not better than silver data if it comes from a different dataset. This highlights a crucial aspect related to the creation of manually labeled datasets, especially in the highly subjective area of hate speech and affective computing in general, where different guidelines and different annotators clearly introduce large biases and discrepancies across datasets.
All considered, we believe that obtaining data in a distant, more ecological way should be further pursued and refined. How to better exploit the information that comes from polarized embeddings in combination with other features is also left to future work.
The authors want to thank the EVALITA 2018 Hate Speech Detection (HaSpeeDe) task organizers for allowing us to use their datasets.