English. Automatic Misogyny Identification (AMI) is a new shared task proposed for the first time at the Evalita 2018 evaluation campaign. The AMI challenge, based on both Italian and English tweets, is distinguished into two subtasks, i.e. Subtask A on misogyny identification and Subtask B about misogynistic behaviour categorization and target classification. Regarding the Italian language, we have received a total of 13 runs for Subtask A and 11 runs for Subtask B. Concerning the English language, we received 26 submissions for Subtask A and 23 runs for Subtask B. The participating systems have been distinguished according to the language, counting 6 teams for Italian and 10 teams for English. We present here an overview of the AMI shared task, the datasets, the evaluation methodology, the results obtained by the participants and a discussion of the methodology adopted by the teams. Finally, we draw some conclusions and discuss future work.
Italiano. Automatic Misogyny Identification (AMI) e` un nuovo shared task proposto per la prima volta nella campagna di valutazione Evalita 2018. La sfida AMI, basata su tweet italiani e inglesi, si distingue in due sottotask ossia Subtask A relativo al riconoscimento della misoginia e Subtask B relativo alla categorizzazione di espressioni misogine e alla classificazione del soggetto target. Per quanto riguarda la lingua italiana, sono stati ricevuti un totale di 13 run per il Subtask A e 11 run per il Subtask B. Per quanto riguarda la lingua inglese, sono stati ricevuti 26 run per il Subtask A e 23 per Subtask B. I sistemi partecipanti sono stati distinti in base alla lingua, raccogliendo un totale di 6 team partecipanti per l’italiano e 10 team per l’inglese. Presentiamo di seguito una sintesi dello shared task AMI, i dataset, la metodologia di valutazione, i risultati ottenuti dai partecipanti e una discussione sulle metodologie adottate dai diversi team. Infine, vengono discusse conclusioni e delineati gli sviluppi futuri.
During the last years, the phenomenon of hate
against women increased exponentially especially
in online environment such as microblogs
The AMI shared task is organized according to two main subtasks:
tem must discriminate misogynistic contents from the non-misogynistic ones. Examples of misogynous and non-misogynous tweets are reported in Table 1.
ognize the targets that can be either specific users or groups of women together with the identification of the type of misogyny against women.
Regarding the misogynistic behaviour, a tweet must be classified as belonging to one of the following categories:
Stereotype & Objectification: a widely held but fixed and oversimplified image or idea of a woman; description of women’s physical appeal and/or comparisons to narrow standards.
Dominance: to assert the superiority of men over women to highlight gender inequality. Derailing: to justify woman abuse, rejecting male responsibility; an attempt to disrupt the conversation in order to redirect women’s conversations on something more comfortable for men.
Sexual Harassment & Threats of Violence: to describe actions as sexual advances, requests for sexual favours, harassment of a sexual nature; intent to physically assert power over women through threats of violence.
Discredit: slurring over women with no other larger intention.
Examples of Misogynistic Behaviours are reported in Table 2.
Concerning the target classification, the main goal is to classify each misogynous tweet as belonging to one of the following two target categories:
Active (individual): the text includes offensive messages purposely sent to a specific target; Passive (generic): it refers to messages posted to many potential receivers (e.g. groups of women).
Examples of targets of misogynous tweets are reported in Table 3. 3
In order to provide training and testing data both for Italian and English, three approaches were employed to collect misogynous text on Twitter: Streaming download using a set of manually defined representative keywords, e.g. bi**h, w**re, c*nt for English and pu****a, tr**a, f**a di legno for Italian; Monitoring of potential victims’ accounts, e.g. gamergate victims and public feminist women; Downloading the history of identified misogynist, i.e. explicitly declared hate against women on their Twitter profiles.
Among all the collected tweets we selected a subset of text querying the database with the copresence of keywords, originating two corpora initially composed of 10000 tweets for each language. In order to label both the Italian and English datasets, we involved a group of 6 experts exploiting the CrowdFlower1 platform for internal use. At the end of the labelling phase, we provided one corpus for Italian and one corpus for English to all the participants. The inter-rater annotator agreement on the English dataset for the fields of “misogynous”, “misogyny category” and “target” 1Now Figure Eight: https://figure-eight.com/ is 0.81, 0.45 and 0.49 respectively, while for the Italian dataset is 0.96, 0.68 and 0.76. Each corpus is distinguished in Training and Test datasets. Regarding the training data, both the Italian and English corpora are composed of 4000 tweets. Concerning the test data, we provided 1000 tweets for each language. The training data has been provided as tab-separated, according to the following fields: id denotes a unique identifier of the tweet. text represents the tweet text. misogynous defines if the tweet is misogynous or not misogynous; it takes values as 1 if the tweet is misogynous, 0 if the tweet is not misogynous. misogyny category denotes the type of misogynistic behaviour; it takes value as: – stereotype: denotes the category “Stereotype & Objectification”; – dominance: denotes the category “Dominance”; – derailing: denotes the category “Derailing”; – sexual harassment: denotes the category “Sexual Harassment & Threats of Violence”; – discredit: denotes the category “Discredit”; – 0 if the tweet is not misogynous. target denotes the subject of the misogynous tweet; it takes value as: – active: denotes a specific target (individual); – passive: denotes potential receivers (generic); – 0 if the tweet is not misogynous.
Concerning the test data, only “id” and “text” have been provided to the participants. Examples of all possible allowed combinations are reported below. Additionally to the field “id”, we report all the combinations of labels to be predicted, i.e. “misogynous”, “misogyny category” and “target”: 0 0 0 1 stereotype active 1 stereotype passive 1 dominance active 1 dominance passive 1 derailing active 1 derailing passive 1 sexual harassment active 1 sexual harassment passive 1 discredit active 1 discredit passive
The label distribution related to the Training and Test datasets is reported in Table 4. While the distribution of labels related to the field “misogynous” is almost balanced (for both languages), the classes related to the other fields are quite unbalanced. Regarding the “misogyny category”, we can distinguish between the two considered languages. In particular, for the Italian language, the most frequent label is related to the category Stereotype & Objectification, while for English the most predominant one is Discredit. Concerning the “target”, the most predominant victims are specific users (active) with a strong imbalanced distribution on the Italian corpus, while it is almost balanced for the English training dataset and strongly imbalanced on the (active) targets for the corresponding test dataset. 4
Considering the distribution of labels of the dataset, we have chosen different evaluation metrics. In particular, we distinguished as follows: Subtask A. Systems have been evaluated on the field “misogynous” using the standard accuracy measure, and ranked accordingly.
Subtask B. Each field to be predicted has been evaluated independently on the other using a Macro F1-score. In particular, the Macro F1-score for the “misogyny category” field has been computed as average of F1-scores obtained for each category (stereotype, dominance, derailing, sexual harassment, discredit), estimating F1(misogyny category). Analogously, the Macro F1-score for the “target” field has been computed as average of F1-scores obtained for each category (active, passive), F1(target). The final ranking of the systems participating to Subtask B was based on the Average Macro F1-score (F1), computed as follows:
F1 = F1(misogyny category)+F1(target) 2 (1)
In order to compare the submitted runs with a baseline model, we provided a benchmark (AMIBASELINE) based on Support Vector Machine trained on a unigram representation of tweets. In particular, we created one training set for each field to be predicted, i.e. “misogynous”, “misogyny category” and “target”, where each tweet has been represented as a bag-of-words (composed of 1000 terms) coupled with the corresponding label. Once the representations have been obtained, Support Vector Machines with linear kernel have been trained, and provided as AMI-BASELINE.
A total of 6 teams for Italian and 10 teams for English from 10 different countries participated in at least one of the two subtasks of AMI. Each team had the chance to submit up to three runs for English and three runs for Italian. Runs could be constrained, where only the provided training data and lexicons were admitted, and unconstrained, where additional data for training were allowed. Table 5 shows an overview of the teams2 reporting their affiliation, their country, the number of submissions for each language and the subtasks they addressed.
Table 6 reports the results for the Misogyny Identification task, which received 13 submissions for Italian and 26 runs for English submitted respectively from 6 and 10 teams. The highest Accuracy has been achieved by bakarov at 0.844 for Italian and by hateminers at 0.704 for English, both in a constrained setting. Most of the systems have shown an improvement with respect to the AMIBASELINE. While the bakarov team submitted only one run based on TF-IDF coupled with Singular Value Decomposition and Boosting classifier, hateminers achieved the highest performance with a run based on vector representation that concatenates sentence embedding, TF-IDF and average word embeddings coupled with a Logistic Regression model.
Table 7 reports the results for the Misogynistic
Behaviour and Target Classification task, which
received 11 submissions by 5 teams for Italian and
23 submissions by 9 teams for English. The
highest Average Macro F1-score has been achieved by
bakarov at 0.501 for Italian (even if the amended
run of CrotoneMilano achieved the highest
effective performance) and by himani at 0.406 for
English, both in a constrained setting. On the
contrary of the previous task, most of the systems have
shown lower performance compared to the
AMIBASELINE. It can be easily noted by looking at
the Average Macro F1-score of all the approaches,
that the problem of recognizing the misogyny
category and the target is more difficult than the
2The teams himani and resham described their systems in
the same report
This is due to the fact that there can be a high overlapping between textual expressions of different misogyny categories, therefore it is highly subjective for an annotator (and consequently for a system) to select a category rather than another one. Regarding the target classification, systems can be easily misled by the presence of mentions that are not the target of the misogynous content.
While for the bakarov team the system for Subtask B is the same one of Subtask A, himani achieved the highest performance on the English language with a run based on a Bag of N-Gram representation coupled with an Ensemble of 5 models for classifying the Misogynistic Behaviour and 2 models for Target Classification. 6
The submitted systems can be compared by taking into consideration the kind of input features that they have considered for representing tweets and the machine learning model that has been used as classification model.
Textual Feature Representation. The systems submitted by the challenge participants’ consider various techniques for representing the tweet contents. Some teams have concentrated the effort on considering a single type of representation, i.e. the team ITT adopted the traditional TF-IDF representation, while bakarov and RCLN proposed systems considering only weighted n-grams at character level for better dealing with misspellings and capturing few stylistic aspects.
Additionally to the traditional textual
feature representation techniques (i.e. bag of
words/characters, n-grams of words/characters
eventually weighted with TF-IDF) several teams
proposed specific lexical features for improving
the input space and consequently the classification
performances. The team of CrotoneMilano
experimented feature abstraction following the
bleaching approach proposed by Goot et al.
Differently from these approaches, StopPropagHate and hateminers teams proposed systems that consider the popular Embeddings techniques both at word and sentence level.
machine learning models, we can distinguish between approaches that work with traditional Support Vector Machines and Logistic Regression, Ensemble Models and finally Deep Learning methods. Following, we report the models adopted by the systems that participated in the AMI shared task, according to the type of the machine learning model that has been adopted: Support Vector Machines have been exploited by 14-exlab by using both linear and RBF kernel, by SB investigating only a radial basis function kernel, and by CrotoneMilano by adopting again a simple linear kernel; Logistic Regression has been used by bakarov and hateminers; Ensemble Models have been adopted by three teams according to different settings, i.e. ITT and himani used a Simple Voting of different classifiers, resham induced a Simple Voting over different input features and RCLN used an Ensemble based on Random Forest; A Deep Learning classifier has been adopted by only one team, i.e StopPropagHate that trained a simple dense neural network.
External Resources Several participants exploited external resources for providing taskspecific lexical features.
The lexicons for addressing AMI for Italian
have been mostly obtained from lists available
online. The team SB used an available specific Italian
lexicon called “Le parole per ferire” built by Tullio
De Mauro3. Starting from this lexicon provided
by De Mauro, the HurtLex multilingual lexicon
has been created
Regarding the English language, both resham
and 14-exlab used the list of swear words from
noswearing dictionary and the sexist slur list
provided by
Finally, pre-trained Word Embeddings have 3https://www.internazionale.it/ opinione/tullio-de-mauro/2016/09/27/ razzismo-parole-ferire
4https://www.parolacce.org/2016/12/ 20/dati-frequenza-turpiloquio/ and https: //it.wikipedia.org/wiki/Turpiloquio_ nella_lingua_italiana 5https://www.noswearing.com/dictionary Rank We presented here a new shared task about Automatic Misogyny Identification on Twitter for Italian and English. By analysing the runs submitted by the participants we can conclude that the problem of misogyny identification has been satisfactorily addressed by all the teams, while the misogynistic behaviour and target classification still remains a challenging problem. Concerning the future work, several issues should be considered to improve the quality of the collected data, especially for capturing those less frequent misogynistic behaviours such as Dominance and Derailing. The problem of hate speech against women will be further addressed in the HatEval shared task at SemEval in English and Spanish tweets6.
6SemEval 2019 Task 5: HatEval: Multilingual Detection of Hate Speech Against Immigrants and Women in Twitter https://competitions.codalab.org/ competitions/19935
The work of the third author was partially funded by the SomEMBED TIN2015-71147-C2-1-P research project (MINECO/FEDER). We thank Maria Anzovino for her initial help in collecting the tweets subsequently used for the labelling phase and the final creation of the Italian and English corpora used for the AMI shared task.