Korea, Republic of

This paper describes the approaches to the SMM4H 2022 Shared Tasks that were taken by our team for tasks 1 and 6. 
Task 6 was the “Classification of tweets which indicate self-reported COVID-19 vaccination status (in English)”. The best test F1 score was 0.82 using a CT-BERT model, which exceeded the median test F1 score of 0.77, and was close to the 0.83 F1 score of the SMM4H baseline model. 
Task 1 was described as the “Classification, detection and normalization of Adverse Events (AE) mentions in tweets (in English)”. We undertook task 1a, and with a RoBERTa-base model achieved an F1 Score of 0.61 on test data, which exceeded the mean test F1 for the task of 0.56.


COLING 2022

RoBERTa, GPT-2 and Sampling - An interesting concoction

personal health mentions

vaccination status detection

adverse drug events detection

social media text mining

[![](https://assets.underline.io/uploads/markdown_image/1/image/81a3c0317d24f663b49b996875024d45.png)](https://aclanthology.org/volumes/2022.coling-1/)



### THE CONFERENCE WE KNOW AND WE WANT.
**COLING**, the International Conference on Computational Linguistics, is one of the premier conferences for the natural language processing and computational linguistics.

First established in 1965, the biennial COLING conference is held in diverse parts of the globe and attracts participants from both top-ranked research centers and emerging countries. Today, the most important developments in our field are taking place not only in universities and academic research institutes but also in industrial research departments including tech-startups. COLING provides opportunities for all these communities to showcase their exciting discovery.

In fall of 2022, COLING will be held in Gyeongju in a hybrid format. All participants can either present at the venue site or join virtually. As more people get vaccinated, we are happy to provide safer environments for our colleagues. We believe that COLING 2022 will be one of the conferences, free from the pandemic. The hybrid format gives presenters and sponsors a valuable opportunity to promote their companies in both an online and in-person venue. For the first time in a long time, customers can interact with their sponsor's products first-hand. The online venue, too, gives sponsors the chance to network with those unable to attend the in-person session.

The hybrid format ultimately lends itself to greater exposure for our sponsors: COLING2022 will let you reach more potential partners and customers than ever before!<br><br><br>

To gain access to this event page you are required to register.<br> Find more information and pay the registration fee on event the organizer’s website **[https://coling2022.org/reg](https://coling2022.org/reg)**

COLING, the International Conference on Computational Linguistics, is one of the premier conferences for the natural language processing and computational linguistics.

workshop paper

Despite the availability of state-of-the-art models for standard NER tasks, their performance degrades on biomedical data due to OOV entities and the challenges encountered in specialized domains.
We use Flair-NER framework to investigate the effectiveness of various contextual and static embeddings for NER on Spanish tweets, in particular, to capture complex disease mentions.

Catchme if you can! Capturing complex diseasementions in tweets

Social media data is extensively used for various informed decision-making applications in diverse domains like e-business, healthcare, policy making and more. As part of the Shared Task Social Media Mining for Health - SMM4H'22 Task-1a is to classify tweets reporting Adverse Drug Events (ADE). Task 1b is to detect the adverse spans and Task 1c is to normalize ADE mentions with MedDRA terms from tweets data. We present an adaptive learner mechanism for the foundation model to identify Adverse Drug Event (ADE) tweets. For the detected ADE tweets, a pipeline consisting of a pre-trained question-answering model followed by a fuzzy matching algorithm was leveraged for the span extraction and normalization tasks.

Adaptive Learning Model for Effective Detection, Extraction and Normalization of Adverse Drug Events from Social Media Data

This paper describes our approach for 11 classification tasks (Task1a, Task2a, Task2b, Task3a, Task3b, Task4, Task5, Task6, Task7, Task8 and Task9) from Social Media Mining for Health (SMM4H) 2022 Shared Tasks. We developed a classification model that incorporated rdrop to augment data and avoid overfitting, poly loss and focal loss to alleviate sample imbalance, and pseudo labels to improve model performance. The results of our submissions are over or equal to the median scores in almost all tasks. In addition, our model achieved the highest score in Task4, with a higher 7.8% and 5.3% F1-score than the median scores in Task2b and
Task3a respectively.

Improved BERT-based classification models with Rdrop and PolyLoss

We present our results for the shared tasks 2, 4 and 9 of the SMM4H Workshop at COLING 2022 achieved by succesfully fine-tuning pre-trained language models on the downstream tasks. We identify the occurence of code-switching in the test data for task 2 as a possible source of considerable performance degradation on the test set scores. We successfully exploit structural linguistic similarities in the datasets of tasks 4 and 9 for training on joined datasets, scoring first in task 9 and on par with SOTA in task 4.


Fraunhofer FKIE

Automation of social network data assessment is one of the classic challenges of natural language processing. During the COVID-19 pandemic, mining people's stances from their public messages have become crucial regarding the understanding of attitudes towards health orders. In this paper, the authors propose the transformer-based predictive model allowing to effectively classify the presence of stance and premise in the Twitter texts.

Transformer-based classification of premise in tweets related to COVID-19

This paper describes the system used to predict stances towards health orders and to detect premises in Tweets as part of the Social Media Mining for Health 2022 (#SMM4H) shared task. It takes advantage of GPT-2 to generate new labeled data samples which are used together with pre-labeled and unlabeled data to fine-tune an ensemble of GAN-BERT models. First experiments on the validation set yielded good results, although it also revealed that the proposed architecture is more suited for sentiment analysis.
The system achieved a score of 0.4258 for the stance and 0.3581 for the premise detection on the test set.

Learning to Predict Stances and Premisesin Tweets related to COVID-19 Health Orders Using Generative Models

With the Surge in COVID-19, the number of social media postings related to the vaccine has grown, specifically tracing the confirmed reports by the users regarding the COVID-19 vaccine dose termed as Vaccine Surveillance. To mitigate this research problem, we present our novel ensembled approach for self-reporting COVID-19 vaccination status tweets into two labels, namely Vaccine Chatter and Self Report. We utilize state-of-the-art models, namely BERT, RoBERTa, and XLNet. We also present an empirical analysis of the results to present how well the tweet was able to classify and report.

An Ensembles Approach for self-reporting of COVID-19 Vaccination Status Tweets

This paper presents our submission for the Shared Task-2 of classification of stance and premise in tweets about health mandates re- lated to COVID-19 at the Social Media Mining for Health 2022. There have been a plethora of tweets about people expressing their opinions on the COVID-19 epidemic since it first emerged. The shared task emphasizes finding the level of cooperation within the mandates for their stance towards the health orders of the pandemic. Overall the shared subjects the participants to propose system’s that can efficiently perform 1) Stance Detection, which focuses on determining the author’s point of view in
the text. 2) Premise Classification, which indicates whether or not the text has arguments. Through this paper we propose an orchestra- tion of multiple transformer based encoders to derive the output for stance and premise clas- sification. Our best model achieves a F1 score of 0.771 for Premise Classification and an aggregate macro-F1 score of 0.661 for Stance Detection

An Ensembles Approach for Classification of Stance and Premise in COVID-19 Related Tweets

This paper describes the models developed by the AILAB-Udine team for the SMM4H’22 Shared Task. We explored the limits of Transformer based models on text classification, entity extraction and entity normalization, tackling
Tasks 1, 2, 5, 6 and 10. The main takeaways we got from participating in different
tasks are: the overwhelming positive effects of combining different architectures when using ensemble learning, and the great potential of generative models for term normalization.

Limits of Transformers and BERT Ensembles

This paper presents our submission for the SMM4H 2022-Shared Task on the classification of self-reported intimate partner violence on Twitter (in English). The goal of this task was to accurately determine if the contents of a given tweet demonstrated someone reporting their own experience with intimate partner violence. The submitted system is an ensemble of five RoBERTa models each weighted by their respective F1-scores on the validation data-set. This system performed 13% better than the baseline and was the best performing system overall for this shared task.

Premium content

RoBERTa, GPT-2 and Sampling - An interesting concoction

Downloads

Next from COLING 2022

Catchme if you can! Capturing complex diseasementions in tweets

Similar lecture

The SocialDisNER shared task on detection of disease mentions in health-relevant content from social media: methods, evaluation, guidelines andcorpora