Chain of thought (CoT) has proven useful for problems requiring complex reasoning. Many of these problems are both textual and multimodal. Given the inputs in different modalities, a model generates a rationale and then uses it to answer a question. Because of the hallucination issue, the generated soft negative rationales with high textual quality but illogical semantics do not always help improve answer accuracy. This study proposes a rationale generation method using soft negative sampling (SNSE-CoT) to mitigate hallucinations in multimodal CoT. Five methods were applied to generate soft negative samples that shared highly similar text but had different semantics from the original. Bidirectional margin loss (BML) was applied to introduce them into the traditional contrastive learning framework that involves only positive and negative samples. Extensive experiments on the ScienceQA dataset demonstrated the effectiveness of the proposed method. Code and data are released at https://github.com/zgMin/SNSE-CoT.

This paper describes our participation in the SemEval-2024 Task 1, “Semantic Textual Relatedness for African and Asian Languages.” This task detects the degree of semantic relatedness between pairs of sentences. Our approach is to take out the sentence pairs of each instance to construct a new sentence as the prompt template, use MASK to predict the correlation between the two sentences, use the BERT pre-training model to process and calculate the text sequence, and use the synonym replacement method in text data augmentation to expand the size of the data set. We participate in English in track A, which uses a supervised approach, and the Spearman Correlation on the test set is 0.809.

This paper describes our work for SemEval-2024 Task 5: The Legal Argument Reasoning Task in Civil Procedure. After analyzing the task requirements and the training dataset, we used data augmentation, adopted the large model GPT for summary generation, and added supervised contrastive learning to the basic BERT model. Our system achieved an F1 score of 0.551, ranking 14th in the competition leaderboard. Our system achieves an F1 score improvement of 0.1241 over the official baseline model.

This paper describes the system and the resulting model submitted by our team “PCJ” to the SemEval-2023 Task 10 sub-task A contest. In this task, we need to test the English text content in the posts to determine whether there is sexism, which involves emotional text classification. Our submission system utilizes methods based on RoBERTa, SimCSE-RoBERTa pre-training models, and model ensemble to classify and train on datasets provided by the organizers. In the final assessment, our submission achieved a macro average F1 score of 0.8449, ranking 28th out of 84 teams in Task A.

This paper introduces our method in the system for SemEval 2023 Task 2: MultiCoNER II Multilingual Complex Named Entity Recognition, Track 9-Chinese. This task focuses on detecting fine-grained named entities whose data set has a fine-grained taxonomy of 36 NE classes, representing a realistic challenge for NER. In this task, we need to identify entity boundaries and category labels for the six identified categories. We use BERT embedding to represent each character in the original sentence and train CRF-Rdrop to predict named entity categories using the data set provided by the organizer. Our best submission, with a macro average F1 score of 0.5657, ranked 15th out of 22 teams.

This paper introduces the related work and the results of Team Sapphire’s system for SemEval-2022 Task 4: Patronizing and Condescending Language Detection. We only participated in subtask 1. The task goal is to judge whether a news text contains PCL. This task can be considered as a task of binary classification of news texts. In this binary classification task, the BERT-base model is adopted as the pre-trained model used to represent textual information in vector form and encode it. Capsule networks is adopted to extract features from the encoded vectors. The official evaluation metric for subtask 1 is the F1 score over the positive class. Finally, our system’s submitted prediction results on test set achieved the score of 0.5187.

This paper introduces the result of Team Grenzlinie’s experiment in SemEval-2021 task 7: HaHackathon: Detecting and Rating Humor and Offense. This task has two subtasks. Subtask1 includes the humor detection task, the humor rating prediction task, and the humor controversy detection task. Subtask2 is an offensive rating prediction task. Detection task is a binary classification task, and the rating prediction task is a regression task between 0 to 5. 0 means the task is not humorous or not offensive, 5 means the task is very humorous or very offensive. For all the tasks, this paper chooses RoBERTa as the pre-trained model. In classification tasks, Bi-LSTM and adversarial training are adopted. In the regression task, the Bi-LSTM is also adopted. And then we propose a new approach named compare method. Finally, our system achieves an F1-score of 95.05% in the humor detection task, F1-score of 61.74% in the humor controversy detection task, 0.6143 RMSE in humor rating task, 0.4761 RMSE in the offensive rating task on the test datasets.

In this paper, we propose a method of fusing sentence information and word frequency information for the SemEval 2021 Task 1-Lexical Complexity Prediction (LCP) shared task. In our system, the sentence information comes from the RoBERTa model, and the word frequency information comes from the Tf-Idf algorithm. Use Inception block as a shared layer to learn sentence and word frequency information We described the implementation of our best system and discussed our methods and experiments in the task. The shared task is divided into two sub-tasks. The goal of the two sub-tasks is to predict the complexity of a predetermined word. The shared task is divided into two subtasks. The goal of the two subtasks is to predict the complexity of a predetermined word. The evaluation index of the task is the Pearson correlation coefficient. Our best performance system has Pearson correlation coefficients of 0.7434 and 0.8000 in the single-token subtask test set and the multi-token subtask test set, respectively.

This paper introduces the system description of the hub team, which explains the related work and experimental results of our team’s participation in SemEval 2021 Task 2: Multilingual and Cross-lingual Word-in-Context Disambiguation (MCL-WiC). The data of this shared task is mainly some cross-language or multi-language sentence pair corpus. The languages covered in the corpus include English, Chinese, French, Russian, and Arabic. The task goal is to judge whether the same words in these sentence pairs have the same meaning in the sentence. This can be seen as a task of binary classification of sentence pairs. What we need to do is to use our method to determine as accurately as possible the meaning of the words in a sentence pair are the same or different. The model used by our team is mainly composed of RoBERTa and Tf-Idf algorithms. The result evaluation index of task submission is the F1 score. We only participated in the English language task. The final score of the test set prediction results submitted by our team was 84.60.

This article introduces the system description of the hub team, which explains the related work and experimental results of our team’s participation in SemEval 2021 Task 5: Toxic Spans Detection. The data for this shared task comes from some posts on the Internet. The task goal is to identify the toxic content contained in these text data. We need to find the span of the toxic text in the text data as accurately as possible. In the same post, the toxic text may be one paragraph or multiple paragraphs. Our team uses a classification scheme based on word-level to accomplish this task. The system we used to submit the results is ALBERT+BILSTM+CRF. The result evaluation index of the task submission is the F1 score, and the final score of the prediction result of the test set submitted by our team is 0.6640226029.

Introducing common sense to natural language understanding systems has received increasing research attention. To facilitate the researches on common sense reasoning, the SemEval-2020 Task 4 Commonsense Validation and Explanation(ComVE) is proposed. We participate in sub-task A and try various methods including traditional machine learning methods, deep learning methods, and also recent pre-trained language models. Finally, we concatenate the original output of BERT and the output vector of BERT hidden layer state to obtain more abundant semantic information features, and obtain competitive results. Our model achieves an accuracy of 0.8510 in the final test data and ranks 25th among all the teams.

We participate in the classification tasks of SemEval-2020 Task: Subtask1: Detecting counterfactual statements of semeval-2020 task5(Detecting Counterfactuals). This paper examines different approaches and models towards detecting counterfactual statements classification. We choose the Bert model. However, the output of Bert is not a good summary of semantic information, so in order to obtain more abundant semantic information features, we modify the upper layer structure of Bert. Finally, our system achieves an accuracy of 88.90% and F1 score of 86.30% by hard voting, which ranks 6th on the final leader board of the in subtask 1 competition.

This paper reports the zyy1510 team’s work in the International Workshop on Semantic Evaluation (SemEval-2020) shared task on Sentiment analysis for Code-Mixed (Hindi-English, English-Spanish) Social Media Text. The purpose of this task is to determine the polarity of the text, dividing it into one of the three labels positive, negative and neutral. To achieve this goal, we propose an ensemble model of word n-grams-based Multinomial Naive Bayes (MNB) and sub-word level representations in LSTM (Sub-word LSTM) to identify the sentiments of code-mixed data of Hindi-English and English-Spanish. This ensemble model combines the advantage of rich sequential patterns and the intermediate features after convolution from the LSTM model, and the polarity of keywords from the MNB model to obtain the final sentiment score. We have tested our system on Hindi-English and English-Spanish code-mixed social media data sets released for the task. Our model achieves the F1 score of 0.647 in the Hindi-English task and 0.682 in the English-Spanish task, respectively.

We only participated in the first subtask, and a neural sequence model was used to perform the sequence tagging task. We investigated the effects of different markup strategies on model performance. Bert that performed very well in NLP was used as a feature extractor.

This article describes the system submitted to SemEval 2020 Task 12: OffensEval 2020. This task aims to identify and classify offensive languages in different languages on social media. We only participate in the English part of subtask A, which aims to identify offensive languages in English. To solve this task, we propose a BERT model system based on the transform mechanism, and use the maximum self-ensemble to improve model performance. Our model achieved a macro F1 score of 0.913(ranked 13/82) in subtask A.

Social media used for health applications usually contains a large amount of data posted by users, which brings various challenges to NLP, such as spoken language, spelling errors, novel/creative phrases, etc. In this paper, we describe our system submitted to SMM4H 2020: Social Media Mining for Health Applications Shared Task which consists of five sub-tasks. We participate in subtask 1, subtask 2-English, and subtask 5. Our final submitted approach is an ensemble of various fine-tuned transformer-based models. We illustrate that these approaches perform well in imbalanced datasets (For example, the class ratio is 1:10 in subtask 2), but our model performance is not good in extremely imbalanced datasets (For example, the class ratio is 1:400 in subtask 1). Finally, in subtask 1, our result is lower than the average score, in subtask 2-English, our result is higher than the average score, and in subtask 5, our result achieves the highest score. The code is available online.

This paper describes our system designed for SemEval 2019 Task 5 “Shared Task on Multilingual Detection of Hate”.We only participate in subtask-A in English. To address this task, we present a stacked BiGRU model based on a capsule network system. In or- der to convert the tweets into corresponding vector representations and input them into the neural network, we use the fastText tools to get word representations. Then, the sentence representation is enriched by stacked Bidirectional Gated Recurrent Units (BiGRUs) and used as the input of capsule network. Our system achieves an average F1-score of 0.546 and ranks 3rd in the subtask-A in English.

This paper describes the system submitted to SemEval 2019 Task 6: OffensEval 2019. The task aims to identify and categorize offensive language in social media, we only participate in Sub-task A, which aims to identify offensive language. In order to address this task, we propose a system based on a K-max pooling convolutional neural network model, and use an argument for averaging as a valid meta-embedding technique to get a metaembedding. Finally, we also use a cyclic learning rate policy to improve model performance. Our model achieves a Macro F1-score of 0.802 (ranked 9/103) in the Sub-task A.

In this paper we describe a deep-learning system that competed as SemEval 2019 Task 9-SubTask A: Suggestion Mining from Online Reviews and Forums. We use Word2Vec to learn the distributed representations from sentences. This system is composed of a Stacked Bidirectional Long-Short Memory Network (SBiLSTM) for enriching word representations before and after the sequence relationship with context. We perform an ensemble to improve the effectiveness of our model. Our official submission results achieve an F1-score 0.5659.

We participated in the BioNLP 2019 Open Shared Tasks: binary relation extraction of SeeDev task. The model was constructed us- ing convolutional neural networks (CNN) and long short term memory networks (LSTM). The full text information and context information were collected using the advantages of CNN and LSTM. The model consisted of two main modules: distributed semantic representation construction, such as word embedding, distance embedding and entity type embed- ding; and CNN-LSTM model. The F1 value of our participated task on the test data set of all types was 0.342. We achieved the second highest in the task. The results showed that our proposed method performed effectively in the binary relation extraction.

We perform the LSTM and BiLSTM model for the emotion intensity prediction. We only join the third subtask in Task 1:Affect in Tweets. Our system rank 6th among all the teams.

We firstly use GloVe to learn the distributed representations automatically from the instance, question and answer triples. Then an attentionbased Bidirectional LSTM (BiLSTM) model is used to encode the triples. We also perform a simple ensemble method to improve the effectiveness of our model. The system we developed obtains an encouraging result on this task. It achieves the accuracy 0.7472 on the test set. We rank 5th according to the official ranking.

Machine Comprehension of text is a typical Natural Language Processing task which remains an elusive challenge. This paper is to solve the task 11 of SemEval-2018, Machine Comprehension using Commonsense Knowledge task. We use deep learning model to solve the problem. We build distributed word embedding of text, question and answering respectively instead of manually extracting features by linguistic tools. Meanwhile, we use a series of frameworks such as CNN model, LSTM model, LSTM with attention model and biLSTM with attention model for processing word vector. Experiments demonstrate the superior performance of biLSTM with attention framework compared to other models. We also delete high frequency words and combine word vector and data augmentation methods, achieved a certain effect. The approach we proposed rank 6th in official results, with accuracy rate of 0.7437 in test dataset.

This paper describes the system submitted to SemEval-2018 Task 12 (The Argument Reasoning Comprehension Task). Enabling a computer to understand a text so that it can answer comprehension questions is still a challenging goal of NLP. We propose a Bidirectional LSTM (BiLSTM) model that reads two sentences separated by a delimiter to determine which warrant is correct. We extend this model with a neural attention mechanism that encourages the model to make reasoning over the given claims and reasons. Officially released results show that our system ranks 6th among 22 submissions to this task.

Reasoning is a crucial part of natural language argumentation. In order to comprehend an argument, we have to reconstruct and analyze its reasoning. In this task, given a natural language argument with a reason and a claim, the goal is to choose the correct implicit reasoning from two options, in order to form a reasonable structure of (Reason, Warrant, Claim). Our approach is to build distributed word embedding of reason, warrant and claim respectively, meanwhile, we use a series of frameworks such as CNN model, LSTM model, GRU with attention model and biLSTM with attention model for processing word vector. Finally, ensemble mechanism is used to integrate the results of each framework to improve the final accuracy. Experiments demonstrate superior performance of ensemble mechanism compared to each separate framework. We are the 11th in official results, the final model can reach a 0.568 accuracy rate on the test dataset.

Sentiment analysis is one of the central issues in Natural Language Processing and has become more and more important in many fields. Typical sentiment analysis classifies the sentiment of sentences into several discrete classes (e.g.,positive or negative). In this paper we describe our deep learning system(combining GRU and SVM) to solve both two-, three- and five-tweet polarity classifications. We first trained a gated recurrent neural network using pre-trained word embeddings, then we extracted features from GRU layer and input these features into support vector machine to fulfill both the classification and quantification subtasks. The proposed approach achieved 37th, 19th, and 14rd places in subtasks A, B and C, respectively.

In this paper, we perform convolutional neural networks (CNN) to learn the joint representations of question-answer pairs first, then use the joint representations as the inputs of the long short-term memory (LSTM) with attention to learn the answer sequence of a question for labeling the matching quality of each answer. We also incorporating external knowledge by training Word2Vec on Flashcards data, thus we get more compact embedding. Experimental results show that our method achieves better or comparable performance compared with the baseline system. The proposed approach achieves the accuracy of 0.39, 0.42 in English valid set, test set, respectively.