We investigate the potential of rule-based systems for the task of ofensive text detection in English and German, and demonstrate their efectiveness in low-resource setings, as an alternative or addition to transfer learning across tasks and languages. Task definitions and annotation guidelines used by existing datasets show great variety, hence state-of-the-art machine learning models do not transfer well across datasets or languages. Furthermore, such systems lack explainability and pose a critical risk of unintended bias. We present simple rule systems based on semantic graphs for classifying ofensive text in two languages and provide both quantitative and qualitative comparison of their performance with deep learning models on 5 datasets across multiple languages and shared tasks.
1. Introduction with many overlapping definitions of categories such as toxicity, hate speech, profanity etc. Datasets are constructed using diferent sets of class definitions corresponding to diferent annotation instructions, and machine learning models that learn paterns of one dataset may perform poorly on another. Modern deep learning models also ofer litle or no explainability of their decisions, and their potential for unintended bias reduces their applicability in real-world scenarios such as automatic content moderation. In this paper we present a rule-based approach, a semi-automatic method for constructing paterns over Abstract Meaning Representations (AMR graphs) built from input text, and evaluate its potential as an alternative to machine learning for ofensive text detection using five datasets of English and German social media text. Our quantitative analysis compares the rule-based method to both monolingual and multilingual deep learning models trained on data from nEvelop-O (G. Recski) tial in low-resource setings as an alternative or addition to transfer learning. Our qualitative analysis examines the decisions made by each system on samples of 100100 texts from both languages and provides a subjective CIKM’22: Advances in Interpretable Machine Learning and Artificial
is subject of growing interest, also as part of the broader research area of explainable artificial intelligence (xAI).
Datasets As pointed out already in a 2017 survey [1], Deep learning models are considered black boxes in most the definition of ofensive text varies greatly across datasets, applications and eforts to interpret them are generally which makes the portability of deep learning models for limited to feature weight visualizations with limited validofensive text detection a hard problem. Annual shared ity (see e.g. [30], [31], and [32] for the controversy about tasks on hate speech detection and related tasks may use using atention weights as explanation). Yet even the similar definitions year after year, but there is great vari- more mature methods for interpreting neural networks ation when moving from one shared task to another and (e.g. LIME [33]) do not ofer the kind of transparency models that achieve high quantitative results on their of ML models that would allow developers to customize targeted test set don’t generalize well (see [2] for a re- their functionality the way a domain expert can update a cent survey). In this paper we shall experiment on yearly traditional rule system. In this work we experiment with datasets from two tasks that both use the same labeling a rudimentary method for semi-automatic, human-in-thescheme for ofensive text, HASOC [ 3] and GermEval [4]. loop (HITL) learning of simple rule systems over semantic Both challenges define a binary classification of social graphs. Recent approaches to automatic learning of rule media texts (Tweets or Facebook comments) into the of- systems for NLP tasks range from the learning of first fensive and non-ofensive classes, and a fine-grained clas- order logic formulae over semantic representations ussification of the ofensive category into the subclasses ing neural networks [34] and integer programming [35] abusive, insulting, and profane. A detailed description to the training of probabilistic grammars over semanof these tasks and datasets will be given in Section 3. tic graphs [36]. Human-in-the-loop (HITL) approaches hTe OLID and SOLID datasets of SemEval 2019 [ 5] and involve generating rule candidates to be reviewed by 2020 [6] use task definitions similar to GermEval. Other experts, e.g. by extracting textual paterns [ 37] or semanwidely used datasets with a narrower scope include the tic structures [38]. Rule-based approaches are also often data provided by the TRAC [7, 8] and HatEval [9] shared combined with ML methods, e.g. by incorporating lexical tasks. TRAC contains English, Hindi and Bangla data features into DL architectures [39, 40] or voting between from Twiter and Facebook and annotation focuses on rule-based and ML systems [41, 42, 43]. the categories aggression and misogyny, the HatEval task is concerned with hate speech directed at immigrants or women in English and Spanish Twiter data. 3. Data rAeplyporonadcishterisbutMioonsatl steyxsttermepsrefosrenotfeantisoivnes,tienxctluddetinecgtiboonth IanndthHisAsSecOtiConshwaereidnttraosdkusc,ewdhaictahseatrseftrhoembtahseisGoeframllEovuarl static [10] and contextual embeddings [8, 11]. As in many quantitative experiments in Section 5 and our qualitative popular text classification tasks, the most widely used analysis in Section 7. We choose two recent tasks that use neural language models are based on the Transformer ar- identical labeling schemes and also have one language in chitecture [12], and in particular BERT-based models [13] common (German) allowing us to perform various crossare the basis of the state of the art machine learning sys- dataset experiments. Our experiments involve datasets tems for most datasets, including the best-performing sys- in German and English only, these are the two languages tems on GermEval2021 [14], GermEval2019 [15], HASOC for which we are able to build rule systems and also 2020 English [16] and HASOC 2020 German [8, 17]. Top perform qualitative analysis (see Section 7) in addition to systems enhance quantitative performance by optimiz- quantitative results, allowing us to investigate the ability ing metaparameters such as maximum sentence length of both ML and rule-based models to transfer between or number of training epochs [18, 19], by training on tasks as well as languages. joint subtask labels [20] or utilizing multiple Transformer hTe GermEval shared task was organized in 2018 [ 44], based models to counteract the small dataset sizes [14], 2019 [45], and 2021 [4]. German Twiter posts were annoby pretraining on additional hate speech corpora [21], tated for the 2018 and 2019 challenges, the 2021 task used training jointly on diferent corpora [ 8], or by using ad- comments from a news-related Facebook group. The versarial learning [22]. Further deep learning methods 2018 and 2019 Twiter datasets consist of posts from 100 used in ofensive text detection include LSTMs [ 23, 24], user timelines and is limited to tweets in German that CNNs [25, 26], or both [27], sentence embeddings [28], are not retweets, do not contain URLs, and contain at and ensembles of multiple machine learning models [27, least 5 alphabetic tokens. The dataset is not a random 29]. sample of posts meeting these criteria, users were heuristically selected to ensure a high ratio of ofensive tweets (further details on this selection were not given), then the dataset was debiased using additional tweets with Explainability and rule learning hTe interpretability of NLP models and the explainability of their decisions non-ofensive words that were observed to be overrepre- the semantics of each sentence. For English texts we use sented in ofensive posts, such as Merkel or Flüchtlinge a pretrained Transformer-based AMR parser [50] and the ‘refugees’. The 2021 edition of Germeval featured a collec- amrlib2 library, for German we construct AMRs from tion of comments from the Facebook page of a German text using a multilingual, transition-based system [51] via political talk show. The 2021 training data was collected the amr-eager-multilingual3 library. A rule system between January and June of 2019, while the test set is for a task consists of lists of paterns over graph reprefrom between September and December of 2020. The sentations of text for each possible class, and a text is dataset has been anonymized to comply with Facebook’s predicted to belong to a given class if at least one patern guidelines for publishing data. The datasets from 2018 in the class’s list of paterns matches the corresponding and 2019 categorize the ofensive texts further into three graph. Graphs must be directed and can be edge- and/or categories, profanity, insult, and abuse and defines ofen- node-labeled. Individual paterns are directed graphs sive text as the union of these categories, this is identical whose edge and node labels may be strings or regular to the definition used at HASOC. The 2021 dataset does expressions (regexes) defining sets of possible labels, and not contain such fine-grained labels and defines ofensive a graph patern with regexes for labels defines the set of texts as the union of screaming, vulgar language, insults, all graphs whose corresponding node and edge labels are sarcasm, discrimination, discrediting, and accusation of matched by those regexes. Paterns can also be negated lying. and a conjunction of paterns used as a single rule, a comhTe Hate Speech and Ofensive Content Identification plete rule system can therefore be considered as a single in English and Indo-Aryan Languages (HASOC) shared boolean statement in disjunctive normal form (DNF) of task was inspired by GermEval and OfensEval and was boolean predicates corresponding to graph paterns, in organized in 2019 [46], 2020 [47], and 2021 [3]. The this regard the method is similar to the approach of [35] dataset from 2019 contained tweets and Facebook com- and [34] (see Section 2). ments in English, Hindi and German. Ofensive posts To construct rule systems eficiently, POTATO implewere selected based on keywords and hashtags, and de- ments a form of human-in-the-loop (HITL) learning. For biased similarly to the process described by GermEval each training dataset we consider all AMR graphs and organizers. From 2020 datasets were selected by training generate a list of frequently occurring subgraphs with at a Support Vector Machine classifier (SVM) on a collection most 2 edges, then rank them based on their importance of hate speech datasets and using this classifier to select for the classification task. For this we use subgraphs as the tweets to be annotated for the dataset. Following features to train a decision tree on the dataset using the the definition of the 2019 and 2020 GermEval challenges, sklearn library and then rank these features based on each HASOC task distinguishes between three types of their Gini coeficient. The maximum size of subgraphs ofensive text, those displaying profanity ( PRFN), ofense is a free parameter of the system but must be kept low (OFFN), or hate (HATE). The binary classification of ofen- to limit the search space. We thus obtain a ranked list of sive texts considers the union of these three categories, relevant graph paterns that we can use to construct our and both our quantitative experiments in Section 5 and rule systems manually. We shall describe the individual our qualitative analysis in Section 7 are concerned with rule systems built for our experiments in Section 5. this task only.
uQantitative evaluation is performed using 5 datasets.
In our quantitative experiments as well as in our error For English we train models using the three datasets analysis we compare the performance of standard deep from the 2019-2021 editions of the HASOC shared task, learning models with rule-based systems that define sets for German we use the 2021 GermEval dataset (the trainof paterns over AMR graphs built from the texts of posts ing portion of which is from earlier editions of GermEval) to be classified. For the DL models we use standard archi- and the 2020 HASOC corpus (see Section 3 for details on tectures without modification, technical details will be each dataset). We train standard BERT-based classifiers described along with the experimental setup in Section 5. on each dataset and compare them with rule systems we
Our rule-based solutions are built using POTATO1 [48], built manually. We investigate the ability of models to a framework that enables the rapid construction of graph- transfer between tasks by evaluating each of them on the based rule systems and has recently been used for text test sets of all other datasets as well. We also atempt classification in multiple domains and languages. Input transfer learning between English and German data, by text is parsed into Abstract Meaning Representations training models using multilingual BERT on datasets (AMR, [49]), directed graphs of concepts representing 1https://github.com/adaamko/POTATO 2https://amrlib.readthedocs.io/en/latest/ 3https://github.com/mdtux89/amr-eager-multilingual from one language and evaluating them on the other lan- only used for quantitative evaluation, but not for HITL guage. Finally, we also measure the contribution of our learning or manual analysis. rule-based system to DL models by evaluating the union In each of the 5 rule systems the rules with the highest of their predicted positive labels, i.e. by considering the yield are those that consist of a single node, i.e. that strategy of classifying a text as ofensive if at least one refer to the presence of a single word in the text. The of multiple models would classify it as such. In this sec- majority of these words are in themselves profane and/or tion we provide details of our deep learning experiments, insulting. In English rule systems top keywords include followed by an overview of our rule systems built from asshole, stupid, bitch, shit, fuck as well as useless and diseach dataset using the method in Section 4. Results and grace. In German rule sets the top words that trigger the discussion follow in Section 6. ofensive label in themselves also include ficken ‘fuck’, porno, hurensohn ‘son of a bitch’, arsch ‘ass’ and scheiße Deep learning models For training BERT-based mod- ‘shit’. Rules with multiple nodes typically serve to sepaels we preprocess text data by replacing emoticons with rate ofensive and non-ofensive occurrences of a word. their textual representation using the emoji Python li- For example, the word shame is present in over 200 ofbrary, then removing hashtag symbols and substituting fensive posts of the English HASOC 2021 dataset, but currencies and urls with special tags using the regex- as a keyword rule it would also yield 43 false positives. based library clean-text4. Finally, we use our own regu- Using a patern over AMR graphs we can filter occurlar expressions for masking usernames, media tags, and rences of the word by the object (ARG1) of shame and moderators, by replacing each with the [USER] tag. For construct the rule shame −−1−−−→ (media|person|publicaboth languages we fine-tune a language specific pre- tion|they|you|party|have|government), which yields only 8 trained BERT model (bert-base-german-cased for Ger- false positives for 103 true positives. Another example of man and bert-base-uncased for English) as well as the paterns over multiple nodes are rules covering negation. multilingual model (bert-base-multilingual-cased). For example, in the rule system based on the GermEval On each dataset we then train one model with the language- specific BERT and one with multilingual BERT. Each of 2021 training set, the rule normal −−−−−−→ − matches all the 6 datasets consists of a train and test portion. For posts where the word normal is negated, such as in the selecting training metaparameters we further divide the sentence Das ist doch nicht mehr normal! ‘That’s just not train portions of each dataset into into train and valida- normal anymore!’. The complete rule lists built from each tion sets, using a 3:1 ratio, then for the final experiments of the 5 datasets is available from our repository. we train our models using the full training datasets and evaluate them on the test sets. For each dataset we train 6. Results a neural network with a single linear classification head on top of BERT. Hyper-parameters are set based on per- The shared tasks we focus on each evaluate classifiers formance on the validation set. We use Adam optimizer by measuring precision, recall, and F1-score on both the with a weight decay value of 10−5 and initial learning ofensive and non-ofensive class, and systems are ranked rate of 10−5. We use the balanced weighted loss func- based on the macro-average F-score. tion of sklearn,5 to compensate for unbalanced labels, as HASOC organizers argue that using macro-average suggested by [52]. We set batch size to 8 and train each F1-score counteracts class imbalance [46]. We follow model for 10 epochs, then determine the optimal number this practice in our evaluation, especially since many of of iterations based on their F-score on the validation set. the top participating systems do not publish scores for individual classes. Our main results on the test portions Rule based system For building and applying our of each of the 5 corpora is presented in Table 1. On each AMR-based rule systems we parse all text with language- dataset we evaluate DL models trained on data from the specific text-to-AMR parsers (see Section 4 for details). same task, on data from the other task of the same lanhTe only preprocessing step we apply is the replacement guage, on all data in the language, or on all data from the of emoticons, as described in the previous paragraph. other language (using multilingual BERT). Additionally We build rule systems based on each of the 5 training we evaluate our dataset-specific rule systems and the datasets (HASOC 2019-2021 for English, GermEval 2021 pairwise unions of various systems. We also present the and HASOC 2020 for German). Rule systems were built scores of the top-performing system for each dataset. semi-automatically by the authors, based only on the hTe DL models trained on data from the same task training portions of each dataset, test sets were excluded achieve the best results. These models are typically within from the process entirely and even validation sets were a few percentage points of the best models, and are not improved significantly with the addition of the rule system. Rule systems achieve the highest precision values 4hTe dependencies and BERT models are noted in our repository 5https://scikit-learn.org/ 1 2 0 2 l a v E m re G E D Macro avg
R on each dataset, which is by design and at the expense of recall. The efect of rules as an enhancement is considerable in the case of the transfer learning scenarios, both between tasks and languages. Since rules are generally high-precision, most models’ performance is improved by considering their union with the task-specific rule system. (taking the union of two or more binary classifiers means classifying a text as ofensive if at least one of the models classifies it as such). This efect can be observed the presence of profanity alone warrants the ofensive on both German and English datasets. On the German label. The posts FPen1*‡ and FPen2*†, which have been HASOC dataset, where the EN-multi model is in itself predicted as ofensive by several of our models and conmore than 20 points below the F-score on the ofensive tain words such as fuck and bitch, are annotated as nonclass achieved by the model trained on the training data ofensive. One might atribute these annotations to the corresponding to the test set (DE-HASOC), but adding la- lack of hostile intent in these posts, but this would be bels predicted by the rule-based system closes almost half in sharp contrast with FNen22† and FNen23†, which of this gap, raising F-score from 52.9 to 61.9. On the 2019 contain the same words, also lack any ofensive content, English HASOC dataset the efect is similar, rules close but are nevertheless annotated as ofensive (and profane about half of the performance gap between German and in particular).
English models. This efect shows the potential of simple hTe German sample, taken from the GermEval dataset rule systems in low-resource scenarios where training containing longer Facebook comments, also contained data is only available for other languages and/or for other several instances of sarcasm, which typically resulted in tasks/genres. On some datasets, our rule systems work false negative predictions such as FNde4*†‡ and FNde5*†‡. well as standalone solutions as well. In case of the 2020 Finally, the English sample contained several examples English dataset our rules achieve 83.7 F-score on the of- of data error, such as the inclusion of non-English text fensive class, compared to 90.3 of the best DL system. (FNen3†‡) or encoding issues (FNen13†‡). We believe that in real-world applications, e.g. automatic content moderation, such a system may be preferred de- ID Text spite its lower performance, due to its transparency and the fact that its precision is above 95%.
FNen14†‡ How many people you planning to shag in September? — one person.
the rest are a bonus https://t.co/FcS1FpxSvE FNde1*†‡ @USER solch sinnfreie Beiträge…
FPde2* Schauspielen kann er nicht. Und inzwischen meint er, Ahnung von
Allem zu haben. Schlimm dieser Typ FPde4* @USER…äh, Verzeihung! Fangen Sie doch einfach mal bei sich
selbst, mit Ihren unnützen Motorrädern, an! FNen1*†‡ @timesofindia How dare they call it Indian variant when they dint call it a #wuhanvirus or #chinesevirus?? India should file a legal case against WHO and China in international court.
FNen2*†‡ Sad reality of Indian news channels. A minute by minute coverage of elections while a common man struggles to find #covid treatment essentials. Useless News channels. #COVIDSecondWaveInIndia #CoronaPandemic #IndiaCovidCrisis #COVID19India #IndiaChoked #aajtak #zeenews #ABPnews In this section we perform manual error analysis on samples of 100 posts each of the 2021 datasets for each language (GermEval for German and HASOC for English).
Samples were selected randomly and classified by each of the models described and evaluated in previous sections. Here we provide an overview of errors made by FPen1*‡ miya four creeps into every thought i have what the fuck each model and cite selected examples. The quantitative FPen2*† @imtillyherron Happy MF birthday to my fave bitch out there!! results on this sample are noted in the README of our thhaavnektoyowuofrorryaalwboauytswbehiantgoYtOheUrsamndigfhort sshayowthinagnkmyeotuhaftorI sbheoinugldmn’yt repository. Errors made by our models are grouped into motivation, my idol who radiates nothing but positive energ what we consider to be typical error classes, but we note FFNNeenn2223†† BwiotcuhldIydoounefudcikd msoem?-ucahshto—daIdykI’wmhtoireadsh is? So you gotta tell me that such a categorization is subjective and is made solely lol https://t.co/I0Jj7LNEho for the purpose of discussion and presentation of the re- FNde4*†‡ @USER Sie sind Hellseher? sults of our manual analysis. The examples we refer to in FNde5*†‡ Oh…die Frau hat eine Glaskugel ? Ist ja interessant. our discussion below are presented in Table 2, a full list FNen3†‡ @ANI Naa desh ko corona se bachaya Naa WB elections jeeta itna of errors made by each of the systems as well as quanti- wcahmatpaIitghnoiungghkt.e#bRaeasdigSnePrMiomusolydiModi is big failure for India than tative evaluation of each classifier on the two samples is FNen13†‡ Windy says oh ya hoor sir… No long in. Shattered. Got myself a available in our repository. wðŸe˜e®pðaŸrt¤¦tðimŸe»jâo€b. â3™da‚ïy¸s aðŸm¤o£nTthh.inFkirsItâ€d™ayl.l 1g2ivheomurassheilft. aBo9l/l1o0ckthse hTe largest error class consists of false negative predic- day though. What an absolute fuking stonker eh 😎🔥🙌 tions that are clearly ofensive and some models failed to Table 2 detect them as such. These include e.g. the profanity in Sample texts misclassified by any of our systems, grouped FNen14†‡ or the insult in FNde1*†‡. by error type. Text IDs indicate false positive (FP) or false
Another major group consists of posts on controver- negative (FN) and the models that made the false prediction. sial/sensitive topics whose status as ofensive/non-ofensive * denotes the language specific BERT model, † refers to the is influenced by both form and content and is also proba- multilingual BERT model, ‡ marks the rule-based system. bly controversial. False positive predictions in this group include texts that express strong negative opinions in a relatively civil way (FPde2*, FPde4*), while false negatives are those that may have been annotated as ofensive References because of their tone (FNen1*†‡, FNen2*†‡).
Ground truth annotations are inconsistent about whether [1] A. Schmidt, M. Wiegand, A survey on hate speech detection using natural language processing, in: Proceedings of the Fifth International Workshop on training across diferent classification tasks, in: Natural Language Processing for Social Media, As- FIRE, CEUR, Hyderabad, India, 2020, pp. 823–828. sociation for Computational Linguistics, Valencia, URL: http://ceur-ws.org/Vol-2826/T10-3.pdf. Spain, 2017, pp. 1–10. URL: https://aclanthology. [9] V. Basile, C. Bosco, E. Fersini, D. Nozza, V. Pati, F. M. org/W17-1101. doi:10.18653/v1/W17- 1101. Rangel Pardo, P. Rosso, M. Sanguineti, SemEval[2] W. Yin, A. Zubiaga, Towards generalisable 2019 task 5: Multilingual detection of hate speech hate speech detection: a review on obstacles against immigrants and women in Twiter, in: Proand solutions, PeerJ Computer Science 7 (2021). ceedings of the 13th International Workshop on SeURL: https://peerj.com/articles/cs-598/. doi:https: mantic Evaluation, Association for Computational //doi.org/10.7717/peerj- cs.598. Linguistics, Minneapolis, Minnesota, USA, 2019, [3] T. Mandl, S. Modha, G. K. Shahi, H. Madhu, S. Sa- pp. 54–63. URL: https://aclanthology.org/S19-2007. tapara, P. Majumder, J. Schäfer, T. Ranasinghe, doi:10.18653/v1/S19- 2007.
M. Zampieri, D. Nandini, A. K. Jaiswal, Overview [10] P. Chiril, F. Benamara Zitoune, V. Moriceau, of the HASOC subtrack at FIRE 2021: Hate speech M. Coulomb-Gully, A. Kumar, Multilingual and and ofensive content identification in English and multitarget hate speech detection in tweets, in: Indo-Aryan languages, in: Working Notes of FIRE Actes de la Conférence sur le Traitement Automa2021 - Forum for Information Retrieval Evalua- tique des Langues Naturelles (TALN) PFIA 2019. Voltion, FIRE 2021, Association for Computing Ma- ume II : Articles courts, ATALA, Toulouse, France, chinery, New York, NY, USA, 2021, pp. 1–3. URL: 2019, pp. 351–360. URL: https://aclanthology.org/ https://doi.org/10.1145/3503162.3503176. 2019.jeptalnrecital-court.21. [4] J. Risch, A. Stoll, L. Wilms, M. Wiegand, Overview [11] T. Ranasinghe, M. Zampieri, Multilingual ofensive of the GermEval 2021 shared task on the identifi- language identification with cross-lingual embedcation of toxic, engaging, and fact-claiming com- dings, in: Proceedings of the 2020 Conference on ments, in: Proceedings of the GermEval 2021 Empirical Methods in Natural Language ProcessShared Task on the Identification of Toxic, En- ing (EMNLP), Association for Computational Lingaging, and Fact-Claiming Comments co-located guistics, Online, 2020, pp. 5838–5844. URL: https: with KONVENS, Association for Computational //aclanthology.org/2020.emnlp-main.470. doi:10. Linguistics, Düsseldorf, Germany, 2021, pp. 1–12. 18653/v1/2020.emnlp- main.470. URL: https://aclanthology.org/2021.germeval-1.1. [12] A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, doi:10.48415/2021/fhw5- x128. L. Jones, A. N. Gomez, L. u. Kaiser, I. Polosukhin, [5] M. Zampieri, S. Malmasi, P. Nakov, S. Rosenthal, Atention is all you need, in: I. Guyon, U. V.
N. Farra, R. Kumar, SemEval-2019 task 6: Identify- Luxburg, S. Bengio, H. Wallach, R. Fergus, ing and categorizing ofensive language in social S. Vishwanathan, R. Garnet (Eds.), Advances media (OfensEval), in: Proceedings of the 13th in Neural Information Processing Systems 30, International Workshop on Semantic Evaluation, Curran Associates, Inc., Long Beach, CA, USA, Association for Computational Linguistics, Min- 2017, pp. 5998–6008. URL: http://papers.nips. neapolis, Minnesota, USA, 2019, pp. 75–86. URL: cc/paper/7181-attention-is-all-you-need.pdf. https://aclanthology.org/S19-2010. doi:10.18653/ arXiv:1706.03762.
v1/S19- 2010. [13] J. Devlin, M.-W. Chang, K. Lee, K. Toutanova, BERT: [6] M. Zampieri, P. Nakov, S. Rosenthal, P. Atanasova, Pre-training of deep bidirectional transformers for G. Karadzhov, H. Mubarak, L. Derczynski, Z. Pite- language understanding, in: Proc. of NAACL, nis, Çağrı Çöltekin, Semeval-2020 task 12: Multi- Association for Computational Linguistics, Minlingual ofensive language identification in social neapolis, Minnesota, 2019, pp. 4171–4186. URL: media (OfensEval 2020), in: Proceedings of the https://aclanthology.org/N19-1423. doi:10.18653/ Fourteenth Workshop on Semantic Evaluation, In- v1/N19- 1423. ternational Commitee for Computational Linguis- [14] T. Bornheim, N. Grieger, S. Bialonski, FHAC at tics, Barcelona (online), 2020, pp. 1425–1447. URL: GermEval 2021: Identifying German toxic, engaghttps://aclanthology.org/2020.semeval-1.188. ing, and fact-claiming comments with ensemble [7] ws-2018-trolling, Proceedings of the First Work- learning, in: Proceedings of the GermEval 2021 shop on Trolling, Aggression and Cyberbullying Workshop on the Identification of Toxic, Engag(TRAC-2018), Association for Computational Lin- ing, and Fact-Claiming Comments, Association for guistics, Santa Fe, New Mexico, USA, 2018. URL: Computational Linguistics, Heinrich Heine Univerhttps://aclanthology.org/W18-4400. sity Düsseldorf, Germany, 2021, pp. 105–111. URL: [8] R. Kumar, B. Lahiri, A. K. Ojha, A. Bansal, https://aclanthology.org/2021.germeval-1.16.
ComMA@FIRE 2020: Exploring multilingual joint [15] A. Paraschiv, D.-C. Cercel, UPB at GermEval-2019 task 2: BERT-based ofensive language classifica- org/2020.emnlp-main.606. doi:10.18653/v1/2020. tion of german Tweets, in: Proceedings of the 15th emnlp-main.606.
Conference on Natural Language Processing (KON- [23] B. Wang, Y. Ding, S. Liu, , X. Zhou, YNU_wb at VENS 2019), German Society for Computational HASOC 2019: Ordered neurons LSTM with atenLinguistics & Language Technology, Erlangen, Ger- tion for identifying hate speech and ofensive lanmany, 2019, pp. 398–404. guage, in: FIRE (Working Notes), CEUR, Kolkata, [16] N. Ghanghor, R. Ponnusamy, P. K. Kumare- India, 2019, pp. 191–198. URL: http://ceur-ws.org/ san, R. Priyadharshini, S. Thavareesan, B. R. Vol-2517/T3-2.pdf.
Chakravarthi, IIITK@LT-EDI-EACL2021: Hope [24] A. Mishra, S. Saumya, A. Kumar, speech detection for equality, diversity, and inclu- IIIT_DWD@HASOC 2020: Identifying ofension in Tamil , Malayalam and English, in: Pro- sive content in Indo-European languages, in: FIRE, ceedings of the First Workshop on Language Tech- CEUR, Hyderabad, India, 2020, pp. 139–144. URL: nology for Equality, Diversity and Inclusion, Asso- http://ceur-ws.org/Vol-2826/T2-5.pdf. ciation for Computational Linguistics, Kyiv, 2021, [25] B. Gambäck, U. K. Sikdar, Using convolutional neupp. 197–203. URL: https://aclanthology.org/2021. ral networks to classify hate-speech, in: Proceedltedi-1.30. ings of the First Workshop on Abusive Language [17] S. Dowlagar, R. Mamidi, Hasocone@fire-hasoc2020: Online, Association for Computational LinguisUsing bert and multilingual bert models for hate tics, Vancouver, BC, Canada, 2017, pp. 85–90. URL: speech detection, 2021. URL: https://arxiv.org/pdf/ https://aclanthology.org/W17-3013. doi:10.18653/ 2101.09007.pdf. arXiv:2101.09007. v1/W17-3013. [18] K. Kumari, J. Singh, AI_ML_NIT_Patna @HASOC [26] J. H. Park, P. Fung, One-step and two-step clas2020: BERT models for hate speech identification sification for abusive language detection on Twitin Indo-European languages, in: FIRE, CEUR, ter, in: Proceedings of the First Workshop on AbuHyderabad, India, 2020, pp. 319–324. URL: http: sive Language Online, Association for Computa//ceur-ws.org/Vol-2826/T2-29.pdf. tional Linguistics, Vancouver, BC, Canada, 2017, [19] P. Liu, W. Li, L. Zou, NULI at SemEval-2019 task pp. 41–45. URL: https://aclanthology.org/W17-3006. 6: Transfer learning for ofensive language detec- doi:10.18653/v1/W17-3006. tion using bidirectional transformers, in: Proceed- [27] P. Badjatiya, S. Gupta, M. Gupta, V. Varma, Deep ings of the 13th International Workshop on Se- learning for hate speech detection in tweets, in: mantic Evaluation, Association for Computational Proceedings of the 26th International Conference Linguistics, Minneapolis, Minnesota, USA, 2019, on World Wide Web Companion, WWW ’17 Compp. 87–91. URL: https://aclanthology.org/S19-2011. panion, International World Wide Web Conferdoi:10.18653/v1/S19-2011. ences Steering Commitee, Republic and Canton [20] S. Mishra, S. Mishra, 3Idiots at HASOC 2019: Fine- of Geneva, CHE, 2017, p. 759–760. URL: https: tuning transformer neural networks for hate speech //doi.org/10.1145/3041021.3054223. doi:10.1145/ identification in Indo-European languages, in: FIRE 3041021.3054223. (Working Notes), CEUR, Kolkata, India, 2019, pp. [28] V. Indurthi, B. Syed, M. Shrivastava, N. Chakravar208–213. URL: http://ceur-ws.org/Vol-2517/T3-4. tula, M. Gupta, V. Varma, FERMI at SemEval-2019 pdf. task 5: Using sentence embeddings to identify hate [21] T. Caselli, V. Basile, J. Mitrović, M. Granitzer, Hate- speech against immigrants and women in Twiter, BERT: Retraining BERT for abusive language detec- in: Proceedings of the 13th International Worktion in english, in: Proceedings of the 5th Work- shop on Semantic Evaluation, Association for Comshop on Online Abuse and Harms, volume Pro- putational Linguistics, Minneapolis, Minnesota, ceedings of the 5th Workshop on Online Abuse USA, 2019, pp. 70–74. URL: https://aclanthology. and Harms (WOAH 2021), Association for Compu- org/S19-2009. doi:10.18653/v1/S19-2009. tational Linguistics, Bangkok, Thailand, 2021, pp. [29] A. Nikolov, V. Radivchev, Nikolov-radivchev at 17–25. URL: https://aclanthology.org/2021.woah-1. SemEval-2019 task 6: Ofensive Tweet classifica3. doi:10.18653/v1/2021.woah-1.3. tion with BERT and ensembles, in: Proceedings [22] T. Tran, Y. Hu, C. Hu, K. Yen, F. Tan, K. Lee, of the 13th International Workshop on Semantic S. Park, HABERTOR: An eficient and efec- Evaluation, Association for Computational Lintive deep hatespeech detector, in: Proceedings guistics, Minneapolis, Minnesota, USA, 2019, pp. of the 2020 Conference on Empirical Methods 691–695. URL: https://aclanthology.org/S19-2123. in Natural Language Processing (EMNLP), As- doi:10.18653/v1/S19-2123. sociation for Computational Linguistics, Online, [30] S. Serrano, N. A. Smith, Is Atention Interpretable?, 2020, pp. 7486–7502. URL: https://aclanthology. in: Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, Associ- https://aclanthology.org/K19-2006. doi:10.18653/ ation for Computational Linguistics, Florence, Italy, v1/K19- 2006. 2019, pp. 2931–2951. URL: https://aclanthology.org/ [37] P. Lertvitayakumjorn, L. Choshen, E. Shnarch, P19-1282. doi:10.18653/v1/P19- 1282. F. Toni, GrASP: A library for extracting and [31] S. Wiegrefe, Y. Pinter, Atention is not not Expla- exploring human-interpretable textual patnation, in: Proceedings of the 2019 Conference on terns, htps://arxiv.org/abs/2104.03958, 2021. Empirical Methods in Natural Language Processing arXiv:2104.03958. and the 9th International Joint Conference on Nat- [38] P. Sen, Y. Li, E. Kandogan, Y. Yang, W. Lasecki, ural Language Processing (EMNLP-IJCNLP), Asso- HEIDL: Learning linguistic expressions with deep ciation for Computational Linguistics, Hong Kong, learning and human-in-the-loop, in: ProceedChina, 2019, pp. 11–20. URL: https://aclanthology. ings of the 57th Annual Meeting of the Assoorg/D19-1002. doi:10.18653/v1/D19- 1002. ciation for Computational Linguistics: System [32] S. Jain, B. C. Wallace, Atention is not Explana- Demonstrations, Association for Computational tion, in: Proceedings of the 2019 Conference of Linguistics, Florence, Italy, 2019, pp. 135–140. the North American Chapter of the Association URL: https://www.aclweb.org/anthology/P19-3023. for Computational Linguistics: Human Language doi:10.18653/v1/P19- 3023.
Technologies, Volume 1 (Long and Short Papers), [39] A. Koufakou, E. W. Pamungkas, V. Basile, V. Pati, Association for Computational Linguistics, Min- HurtBERT: Incorporating lexical features with neapolis, Minnesota, 2019, pp. 3543–3556. URL: BERT for the detection of abusive language, in: https://aclanthology.org/N19-1357. doi:10.18653/ Proceedings of the Fourth Workshop on Online v1/N19- 1357. Abuse and Harms, Association for Computational [33] M. T. Ribeiro, S. Singh, C. Guestrin, ”Why Should Linguistics, Online, 2020, pp. 34–43. URL: https: I Trust You?”: Explaining the Predictions of Any //aclanthology.org/2020.alw-1.5. doi:10.18653/v1/ Classifier, in: Proceedings of the 22nd ACM 2020.alw- 1.5.
SIGKDD International Conference on Knowledge [40] E. W. Pamungkas, V. Pati, Cross-domain and crossDiscovery and Data Mining, KDD ’16, Association lingual abusive language detection: A hybrid apfor Computing Machinery, New York, NY, USA, proach with deep learning and a multilingual lex2016, p. 1135–1144. URL: https://doi.org/10.1145/ icon, in: Proceedings of the 57th Annual Meet2939672.2939778. doi:10.1145/2939672.2939778. ing of the Association for Computational Linguis[34] P. Sen, M. Danilevsky, Y. Li, S. Brahma, M. Boehm, tics: Student Research Workshop, Association for L. Chiticariu, R. Krishnamurthy, Learning explain- Computational Linguistics, Florence, Italy, 2019, pp. able linguistic expressions with neural inductive 363–370. URL: https://aclanthology.org/P19-2051. logic programming for sentence classification, in: doi:10.18653/v1/P19- 2051.
Proceedings of the 2020 Conference on Empirical [41] A. Razavi, D. Inkpen, S. Uritsky, S. Matwin, OfMethods in Natural Language Processing (EMNLP), fensive language detection using multi-level classiAssociation for Computational Linguistics, On- ifcation, in: Advances in Artificial Intelligence, line, 2020, pp. 4211–4221. URL: https://www. Springer Berlin Heidelberg, Berlin, Heidelberg, aclweb.org/anthology/2020.emnlp-main.345. 2010, pp. 16–27. URL: http://www.eiti.uottawa.ca/ doi:10.18653/v1/2020.emnlp- main.345. ~diana/publications/Flame_Final.pdf. doi:10.1007/ [35] S. Dash, O. Gunluk, D. Wei, Boolean decision 978- 3- 642- 13059- 5_5.
rules via column generation, in: S. Bengio, [42] K. Gémes, G. Recski, TUW-Inf at GermEval2021: H. Wallach, H. Larochelle, K. Grauman, N. Cesa- Rule-based and hybrid methods for detecting toxic, Bianchi, R. Garnet (Eds.), Advances in Neu- engaging, and fact-claiming comments, in: Proceedral Information Processing Systems, volume 31, ings of the GermEval 2021 Workshop on the IdentifiCurran Associates, Inc., Montréal, Canada, 2018. cation of Toxic, Engaging, and Fact-Claiming ComURL: https://proceedings.neurips.cc/paper/2018/ ments, Association for Computational Linguistics, file/743394beff4b1282ba735e5e3723ed74-Paper.pdf. Heinrich Heine University Düsseldorf, Germany, [36] L. Donatelli, M. Fowlie, J. Groschwitz, A. Koller, 2021, pp. 69–75. URL: https://aclanthology.org/2021.
M. Lindemann, M. Mina, P. Weißenhorn, Saar- germeval-1.10. land at MRP 2019: Compositional parsing across [43] K. Gémes, A. Kovács, M. Reichel, G. Recski, Ofall graphbanks, in: Proceedings of the Shared fensive text detection on English Twiter with Task on Cross-Framework Meaning Representa- deep learning models and rule-based systems, in: tion Parsing at the 2019 Conference on Natural FIRE 2021 Working Notes, CEUR, Gandhinagar, InLanguage Learning, Association for Computational dia, 2021, pp. 283–296. URL: http://ceur-ws.org/ Linguistics, Hong Kong, 2019, pp. 66–75. URL: Vol-3159/T1-29.pdf. [44] M. Wiegand, M. Siegel, J. Ruppenhofer, Overview W13-2322.
of the GermEval 2018 shared task on the identifi- [50] C. Rafel, N. Shazeer, A. Roberts, K. Lee, S. Narang, cation of ofensive language, in: Proceedings of M. Matena, Y. Zhou, W. Li, P. J. Liu, Exploring the GermEval 2018, 14th Conference on Natural Lan- limits of transfer learning with a unified text-to-text guage Processing (KONVENS 2018), Vienna, Aus- transformer, 2020. arXiv:1910.10683. tria – September 21, 2018, Austrian Academy of Sci- [51] M. Damonte, S. B. Cohen, Cross-lingual Abences, Vienna, Austria, 2018, pp. 1–10. URL: https: stract Meaning Representation parsing, in: Pro//nbn-resolving.org/urn:nbn:de:bsz:mh39-84935. ceedings of the 2018 Conference of the North [45] J. Struß, M. Siegel, J. Ruppenhofer, M. Wie- American Chapter of the Association for Compugand, M. Klenner, Overview of GermEval tational Linguistics: Human Language Technolotask 2, 2019 shared task on the identification gies, Volume 1 (Long Papers), Association for Comof ofensive language, in: Preliminary pro- putational Linguistics, New Orleans, Louisiana, ceedings of the 15th Conference on Natural 2018, pp. 1146–1155. URL: https://aclanthology.org/ Language Processing (KONVENS 2019), October N18-1104. doi:10.18653/v1/N18-1104. 9 – 11, 2019 at Friedrich-Alexander-Universität [52] G. King, L. Zeng, Logistic regression in Erlangen-Nürnberg, German Society for Com- rare events data, Political Analysis 9 (2001) putational Linguistics & Language Technology 137–163. URL: https://www.cambridge. und Friedrich-Alexander-Universität Erlangen- org/core/journals/political-analysis/article/ Nürnberg, München, Germany, 2019, pp. 352–363. logistic-regression-in-rare-events-data/ URL: https://corpora.linguistik.uni-erlangen. 1E09F0F36F89DF12A823130FDF0DA462. de/data/konvens/proceedings/papers/germeval/ doi:10.1093/oxfordjournals.pan.a004868.
GermEvalSharedTask2019Iggsa.pdf. [46] T. Mandl, S. Modha, P. Majumder, D. Patel, M. Dave,
M. Chintak, A. Patel, Overview of the HASOC Track at FIRE 2019: Hate speech and ofensive content identification in Indo-European languages, in: Proceedings of the 11th Forum for Information Retrieval Evaluation, FIRE ’19, Association for Computing Machinery, New York, NY, USA, 2019, p. 14–17. URL: https://doi.org/10.1145/ 3368567.3368584. doi:10.1145/3368567.3368584. [47] T. Mandl, S. Modha, A. K. M, B. R. Chakravarthi,
Overview of the HASOC Track at FIRE 2020: Hate speech and ofensive language identification in Tamil, Malayalam, Hindi, English and German, in: Forum for Information Retrieval Evaluation, FIRE 2020, Association for Computing Machinery, New York, NY, USA, 2020, p. 29–32. URL: https: //doi.org/10.1145/3441501.3441517. doi:10.1145/ 3441501.3441517. [48] Á. Kovács, K. Gémes, E. Iklódi, G. Recski, Potato: explainable information extraction framework, in: Proceedings of the 31st ACM International Conference on Information and Knowledge Management, CIKM ’22, Association for Computing Machinery, 2022, p. 4897–4901. URL: https://doi.org/10.1145/ 3511808.3557196. doi:10.1145/3511808.3557196. [49] L. Banarescu, C. Bonial, S. Cai, M. Georgescu,
K. Grifit, U. Hermjakob, K. Knight, P. Koehn, M. Palmer, N. Schneider, Abstract Meaning Representation for sembanking, in: Proceedings of the 7th Linguistic Annotation Workshop and Interoperability with Discourse, Association for Computational Linguistics, Sofia, Bulgaria, 2013, pp. 178–186. URL: https://www.aclweb.org/anthology/