Croatia

Recent research has shown that language models exploit `artifacts&#39; in benchmarks to solve tasks, rather than truly learning them, leading to inflated model performance. In pursuit of creating better benchmarks, we propose VAIDA, a novel benchmark creation paradigm for NLP, that focuses on guiding crowdworkers, an under-explored facet of addressing benchmark idiosyncrasies. VAIDA facilitates sample correction by providing realtime visual feedback and recommendations to improve sample quality. Our approach is domain, model, task, and metric agnostic, and constitutes a paradigm shift for robust, validated, and dynamic benchmark creation via human-and-metric-in-the-loop workflows. We evaluate via expert review and a user study with NASA TLX. We find that VAIDA decreases effort, frustration, mental, and temporal demands of crowdworkers and analysts, simultaneously increasing the performance of both user groups with a 45.8\% decrease in the level of artifacts in created samples. As a by product of our user study, we observe that created samples are adversarial across models, leading to decreases of 31.3\% (BERT), 22.5\% (RoBERTa), 14.98\% (GPT-3 fewshot) in performance.

EACL 2023

Real-Time Visual Feedback to Guide Benchmark Creation: A Human-and-Metric-in-the-Loop Workflow

crowdworker education

visualization

bias

Recent research has shown that language models exploit `artifacts' in benchmarks to solve tasks, rather than truly learning them, leading to inflated model performance. In pursuit of creating better benchmarks, we propose VAIDA, a novel benchmark creation paradigm for NLP, that focuses on guiding crowdworkers, an under-explored facet of addressing benchmark idiosyncrasies. VAIDA facilitates sample correction by providing realtime visual feedback and recommendations to improve sample quality. Our approach is domain, model, task, and metric agnostic, and constitutes a paradigm shift for robust, validated, and dynamic benchmark creation via human-and-metric-in-the-loop workflows. We evaluate via expert review and a user study with NASA TLX. We find that VAIDA decreases effort, frustration, mental, and temporal demands of crowdworkers and analysts, simultaneously increasing the performance of both user groups with a 45.8\% decrease in the level of artifacts in created samples. As a by product of our user study, we observe that created samples are adversarial across models, leading to decreases of 31.3\% (BERT), 22.5\% (RoBERTa), 14.98\% (GPT-3 fewshot) in performance.

poster

### Welcome to the 17th Conference of the European Chapter of the Association for Computational Linguistics (EACL) 
Continuing its mission of expanding and involving the science community of all European countries, EACL had selected the Ukrainian community for the 16th EACL, which took place online due to the COVID pandemic. Unfortunately, the ongoing war made the organisation in Kyiv impossible. Considering the importance of physical interaction among researchers, especially after the restrictions imposed by the COVID pandemic, in addition to an online mode, the conference will be held in Dubrovnik, Croatia, from 2 to 6 of May, 2023. The original aim of strengthening the connection with the Ukrainian community will not change as our program will feature a dedicated session and a workshop to highlight work on Ukrainian language technologies. As the flagship European conference in the field of computational linguistics, EACL welcomes European and international researchers covering a broad spectrum of research areas that are concerned with computational approaches to natural language.

You need to register for the conference in order to access this site. Please visit https://2023.eacl.org/registration for more information.

As the flagship European conference in the field of computational linguistics, EACL welcomes European and international researchers covering a broad spectrum of research areas that are concerned with computational approaches to natural language.

The semantic code search is to find code snippets from the collection of candidate code snippets with respect to a user query that describes functionality. Recent work on code search proposes data augmentation of queries for contrastive learning. This data augmentation approach modifies random words in queries. When a user web query for searching code snippet is too brief, the important word that represents the search intent of the query could be undesirably modified. A code snippet has informative components such as function name and documentation that describe its functionality. We propose to utilize these code components to identify important words and preserve them in the data augmentation step. We present KeyDAC (Keyword-based Data Augmentation for Contrastive learning) that identifies important words for code search from queries and code components based on term matching. KeyDAC augments query-code pairs while preserving keywords, and then leverages generated training instances for contrastive learning. We use KeyDAC to fine-tune various pre-trained language models and evaluate the performance of code search and code question answering via CoSQA and WebQueryTest. The experimental results confirm that KeyDAC substantially outperforms the current state-of-the-art performance, and achieves the new state-of-the-arts for both tasks.

Contrastive Learning with Keyword-based Data Augmentation for Code Search and Code Question Answering

Social commonsense contains many human biases due to social and cultural influence (Sap et al., 2020; Emelin et al., 2020). We focus on identifying cultural biases in data, specifically causal assumptions and commonsense implications, that strongly influence model decisions for a variety of tasks designed for social impact. This enables us to examine data for bias by making explicit the causal (if-then, inferential) relations in social commonsense knowledge used for decision making, furthering interpretable commonsense reasoning from a dataset perspective. We apply our methods on 2 social tasks: emotion detection and perceived value detection. We identify influential social commonsense knowledge to explain model behavior in the following ways. First, we augment large-scale language models with social knowledge and show improvements for the tasks, indicating the implicit assumptions a model requires to be successful on each dataset. Second, we identify influential events in the datasets by using social knowledge to cluster data and demonstrate the influence that these events have on model behavior via leave-K-out experiments. This allows us to gain a dataset-level understanding of the events and causal commonsense relationships that strongly influence predictions. We then analyze these relationships to detect influential cultural bias in each dataset. Finally, we use our influential event identification for detecting mislabeled examples and improve training and performance through their removal. We support our findings with manual analysis.

Social Commonsense for Explanation and Cultural Bias Discovery

Bias-measuring datasets play a critical role in detecting biased behavior of language models and in evaluating progress of bias mitigation methods. In this work, we focus on evaluating gender bias through coreference resolution, where previous datasets are either hand-crafted or fail to reliably measure an explicitly defined bias. To overcome these shortcomings, we propose a novel method to collect diverse, natural, and minimally distant text pairs via counterfactual generation, and construct Counter-GAP, an annotated dataset consisting of 4008 instances grouped into 1002 quadruples. We further identify a bias cancellation problem in previous group-level metrics on Counter-GAP, and propose to use the difference between inconsistency across genders and within genders to measure bias at a quadruple level. Our results show that four pre-trained language models are significantly more inconsistent across different
gender groups than within each group, and that a name-based counterfactual data augmentation method is more effective to mitigate such bias than an anonymization-based method.

Counter-GAP: Counterfactual Bias Evaluation through Gendered Ambiguous Pronouns

Although the curse of multilinguality significantly restricts the language abilities of multilingual models in monolingual settings, researchers now still have to rely on multilingual models to develop state-of-the-art systems in Vietnamese Machine Reading Comprehension. This difficulty in researching is because of the limited number of high-quality works in developing Vietnamese language models. In order to encourage more work in this research field, we present a comprehensive analysis of language weaknesses and strengths of current Vietnamese monolingual models using the downstream task of Machine Reading Comprehension. From the analysis results, we suggest new directions for developing Vietnamese language models. Besides this main contribution, we also successfully reveal the existence of artifacts in Vietnamese Machine Reading Comprehension benchmarks and suggest an urgent need for new high-quality benchmarks to track the progress of Vietnamese Machine Reading Comprehension. Moreover, we also introduced a minor but valuable modification to the process of annotating unanswerable questions for Machine Reading Comprehension from previous work. Our proposed modification helps improve the quality of unanswerable questions to a higher level of difficulty for Machine Reading Comprehension systems to solve.

Revealing Weaknesses of Vietnamese Language Models Through Unanswerable Questions in Machine Reading Comprehension

Social chatbots have gained immense popularity, and their appeal lies not just in their capacity to respond to the diverse requests from users, but also in the ability to develop an emotional connection with users. To further develop and promote social chatbots, we need to concentrate on increasing user interaction and
take into account both the intellectual and emotional quotient in the conversational agents. In this paper, we propose a multi-task framework that jointly identifies the emotion of a given dialogue and generates response in accordance to the identified emotion. We employ a BERT based network for creating an empathetic system and use a mixed objective function that trains the end-to-end network with both the
classification and generation loss. Experimental results show that our proposed framework outperforms current state-of-the-art models.

A Unified Framework for Emotion Identification and Generation in Dialogues

Automatic error type classification is an important process in both learner corpora creation and evaluation of large-scale grammatical error correction systems. 
Rule-based classifier approaches such as ERRANT have been widely used to classify edits between correct-erroneous sentence pairs into predefined error categories.
However, the used error categories are far from being universal yielding many language specific variants of ERRANT.
In this paper, we discuss the applicability of the previously introduced grammatical error types to an agglutinative language, Turkish. We suggest changes on current error categories and discuss a hierarchical structure to better suit the inflectional and derivational properties of this morphologically highly rich language. We also introduce ERRANT-TR, the first automatic error type classification toolkit for Turkish. ERRANT-TR currently uses a rule-based error type classification pipeline which relies on word level morphological information. Due to unavailability of learner corpora in Turkish, the proposed system is evaluated on a small set of 106 annotated sentences and its performance is measured as 77.04\% $F_{0.5}$ score. 
The next step is to use ERRANT-TR for the development of a Turkish learner corpus. The code will be made publicly available.

Towards Automatic Grammatical Error Type Classification for Turkish

Automatically generated emotion arcs, which capture how an individual or a population feels over time, are widely used in industry and research. However, there is little research on evaluating the accuracy of these generated arcs. This is partly due to the difficulty of establishing the true (gold) emotion arc. Our work, for the first time, systematically and quantitatively evaluates automatically generated emotion arcs. We also compare two common ways of generating emotion arcs: Machine-Learning (ML) models and Lexicon-Only (LexO) methods. Using 36 diverse datasets, we systematically study factors that impact the quality of generated arcs. We show that despite being markedly poor at instance level, LexO methods are highly accurate at generating emotion arcs by aggregating information from hundreds of instances. These findings have wide-spread implications for both commercial development as well as research in fields such as psychology, public health, digital humanities, and social sciences, as LexO methods are simple, interpretable, have low carbon footprint, and do not require domain-specific training data or extensive programming expertise.

Frustratingly Easy Sentiment Analysis of Text Streams: Generating High-Quality Emotion Arcs Using Emotion Lexicons

Open-retrieval question answering systems are generally trained and tested on large datasets in well-established domains. However, low-resource settings such as new and emerging domains would especially benefit from reliable question answering systems. Furthermore, multilingual and cross-lingual resources in emergent domains are scarce, leading to few or no such systems.
In this paper, we demonstrate a cross-lingual open-retrieval question answering system for the emergent domain of COVID-19.
Our system adopts a corpus of scientific articles to ensure that retrieved documents are reliable. To address the scarcity of cross-lingual training data in emergent domains, we present a method utilizing automatic translation, alignment, and filtering to produce English-to-all datasets.
We show that a deep semantic retriever greatly benefits from training on our English-to-all data and significantly outperforms a BM25 baseline in the cross-lingual setting.
We illustrate the capabilities of our system with examples and release all code necessary to train and deploy such a system.

Addressing Issues of Cross-Linguality in Open-Retrieval Question Answering Systems For Emergent Domains

In recent years, COVID-19 has impacted all aspects of human life. As a result, numerous publications relating to this disease have been issued. Due to the massive volume of publications, some retrieval systems have been developed to provide researchers with useful information. In these systems, lexical searching methods are widely used, which raises many issues related to acronyms, synonyms, and rare keywords. In this paper, we present a hybrid relation retrieval system, CovRelex-SE, based on embeddings to provide high-quality search results.

CovRelex-SE: Adding Semantic Information for Relation Search via Sequence Embedding

The development of poetry generation system mainly focuses on enhancing the capacity of generation model. However, the demands of customization and polishing are generally ignored, which highly reduces the scope of application. In this work, we present Yu Sheng, a web-based poetry generation system that is featured a human-in-loop generation framework, providing various customization options for users with different backgrounds to engage in the process of poetry composition. To this end, we propose two methods and train the models that can perform constrained generation and fine-grained polishing. The automatic and human evaluation results show that our system has a strong ability to generate and polish poetry compared to other vanilla models. Our system is publicly accessible at: https://yusheng.cis.um.edu.mo.

Premium content

Real-Time Visual Feedback to Guide Benchmark Creation: A Human-and-Metric-in-the-Loop Workflow

Downloads

Next from EACL 2023

Contrastive Learning with Keyword-based Data Augmentation for Code Search and Code Question Answering