Canada

Intermediate training of pre-trained transformer-based language models on domain-specific data leads to substantial gains for downstream tasks. To increase efficiency and prevent catastrophic forgetting alleviated from full domain-adaptive pre-training, approaches such as adapters have been developed. However, these require additional parameters for each layer, and are criticized for their limited expressiveness. In this work, we introduce TADA, a novel task-agnostic domain adaptation method which is modular, parameter-efficient, and thus, data-efficient. Within TADA, we retrain the embeddings to learn domain-aware input representations and tokenizers for the transformer encoder, while freezing all other parameters of the model. Then, task-specific fine-tuning is performed. We further conduct experiments with meta-embeddings and newly introduced meta-tokenizers, resulting in one model per task in multi-domain use cases. Our broad evaluation in 4 downstream tasks for 14 domains across single- and multi-domain setups and high- and low-resource scenarios reveals that TADA is an effective and efficient alternative to full domain-adaptive pre-training and adapters for domain adaptation, while not introducing additional parameters or complex training steps.

ACL 2023

TADA: Efficient Task-Agnostic Domain Adaptation for Transformers

task-agnostic

multi-domain

domain adaptation

### Welcome to ACL 2023, the 61st Annual Meeting of the Association for Computational Linguistics! 
 The conference will be held in Toronto, Canada, July 9-14, 2023. 
Following the succession of the recent conferences in our field, ACL 2023 will adopt a hybrid format.
While the impact of Covid has considerably diminished in terms of traveling, obtaining visas to Canada
entails a very long process. Moreover, the global economic conditions pose challenges for many individuals to travel to conferences. Recognizing these circumstances, we know many participants may not be
able to attend the conference in person. Therefore, we are committed to providing a great virtual platform
so everyone has the opportunity to interact with other participants and enjoy the conference. Based on the
current registered participants, approxiately 30% have chosen to attend the conference virtually. Whether
you join us in person or virtually, we sincerely hope everyone has a remarkable conference experience. 
This General Chair’s message is where I express my gratitude to the many individuals who have made
enormous contributions to the conference over the past year.

Read [**ACL 2023 General Chair's message**](https://docs.google.com/document/d/1WobYM7norbG4dI48s75HfJoD89qgX5a_F-6U8AteLSA/edit?usp=sharing/) in full.

##### **[Conference Handbook](https://2023.aclweb.org/downloads/acl2023-handbook.pdf)**

The Association for Computational Linguistics (ACL) is the premier international scientific and professional society for people working on computational problems involving human language, a field often referred to as either computational linguistics or natural language processing.

workshop paper

Pre-trained encoder-only and sequence-to-sequence (seq2seq) models each have advantages, however training both model types from scratch is computationally expensive. We explore recipes to improve pre-training efficiency by initializing one model from the other. (1) Extracting the encoder from a seq2seq model, we show it under-performs a Masked Language Modeling (MLM) encoder, particularly on sequence labeling tasks. Variations of masking during seq2seq training, reducing the decoder size, and continuing with a small amount of MLM training do not close the gap. (2) Conversely, using an encoder to warm-start seq2seq training, we show that by unfreezing the encoder partway through training, we can match task performance of a from-scratch seq2seq model. Overall, this two-stage approach is an efficient recipe to obtain both a multilingual encoder and a seq2seq model, matching the performance of training each model from scratch while reducing the total compute cost by 27%.

Recipes for Sequential Pre-training of Multilingual Encoder and Seq2Seq Models

Parameter-efficient tuning aims to mitigate the large memory requirements of adapting pretrained language models for downstream tasks. For example, one popular method, prefix-tuning, prepends trainable tokens to sequences while freezing the rest of the model's parameters. Although such models attain comparable performance with fine-tuning when applied to sequences with short to moderate lengths, we show their inferior performance when modelling long sequences. To bridge this gap, we propose prefix-propagation, a simple but effective approach that conditions prefixes on previous hidden states. We empirically demonstrate that prefix-propagation outperforms prefix-tuning across long-document tasks, while using 50% fewer parameters. To further investigate the proposed architecture, we also show its advantage in calibration, and perform additional study on its relationship with kernel attention. To the best of our knowledge, this work is the first to focus on parameter-efficient learning for long-sequence language tasks.

Prefix-Propagation: Parameter-Efficient Tuning for Long Sequences

Past research has identified a rich set of handcrafted linguistic features that can potentially assist various tasks. However, their extensive number makes it difficult to effectively select and utilize existing handcrafted features. Coupled with the problem of inconsistent implementation across research works, there has been no categorization scheme or generally-accepted feature names. This creates unwanted confusion. Also, most existing handcrafted feature extraction libraries are not open-source or not actively maintained. As a result, a researcher often has to build such an extraction system from the ground up. 

We collect and categorize more than 220 popular handcrafted features grounded on past literature. Then, we conduct a correlation analysis study on several task-specific datasets and report the potential use cases of each feature. Lastly, we devise a multilingual handcrafted linguistic feature extraction system in a systematically expandable manner. We open-source our system for public access to a rich set of pre-implemented handcrafted features. Our system is coined LFTK and is the largest of its kind. Find it at github.com/brucewlee/lftk.

Handcrafted Features in Computational Linguistics

Effective human learning depends on a wide selection of educational materials that align with the learner's current understanding of the topic. While the Internet has revolutionized human learning or education, a substantial resource accessibility barrier still exists. Namely, the excess of online information can make it challenging to navigate and discover high-quality learning materials in a given subject area. In this paper, we propose an automatic pipeline for building an educational resource discovery system for new domains. The pipeline consists of three main steps: resource searching, feature extraction, and resource classification. We first collect frequent queries from a set of seed documents, and search the web with these queries to obtain candidate resources such as lecture slides and introductory blog posts. Then, we process these resources for BERT-based features and meta-features. Next, we train a tree-based classifier to decide whether they are suitable learning materials. The pipeline achieves F1 scores of 0.94 and 0.82 when evaluated on two similar but novel domains. Finally, we demonstrate how this pipeline can benefit two applications: prerequisite chain learning and leading paragraph generation for surveys. We also release a corpus of 39,728 manually labeled web resources and 659 queries from NLP, Computer Vision (CV), and Statistics (STATS).

A Transfer Learning Pipeline for Educational Resource Discovery with Application in Survey Generation

Single Choice exercises constitute a central exercise type for language learning in a learner's progression from mere implicit exposure through input enhancement to productive language use in open exercises. Distractors that support learning in the individual zone of proximal development should not be derived from static analyses of learner corpora, but rely on dynamic learning analytics based on half-open exercises. We demonstrate how a system's error diagnosis module can be re-used for automatic and dynamic generation and adaptation of distractors, as well as to inform exercise generation in terms of relevant learning goals and reasonable chunking in Jumbled Sentences exercises.

Using Learning Analytics for Adaptive Exercise Generation

Large Language Models (LLMs) offer novel opportunities for educational applications that have the potential to transform traditional learning for students. Despite AI-enhanced applications having the potential to provide personalized learning experiences, more studies are needed on the design of generative AI systems and evidence for using them in real educational settings. In this paper, we design, implement and evaluate \texttt{Reviewriter}, a novel tool to provide students with AI-generated instructions for writing peer reviews in German. Our study identifies three key aspects: a) we provide insights into student needs when writing peer reviews with generative models which we then use to develop a novel system to provide adaptive instructions b) we fine-tune three German language models on a selected corpus of 11,925 student-written peer review texts in German and choose German-GPT2 based on quantitative measures and human evaluation, and c) we evaluate our tool with fourteen students, revealing positive technology acceptance based on quantitative measures. Additionally, the qualitative feedback presents the benefits and limitations of generative AI in peer review writing.

AI-Generated Instructions For Peer Review Writing

We introduce the Korean-Learner-Morpheme (KLM) corpus, a manually annotated dataset consisting of 129,784 morphemes from second language (L2) learners of Korean, featuring morpheme tokenization and part-of-speech (POS) tagging. We evaluate the performance of four Korean morphological analyzers in tokenization and POS tagging on the L2- Korean corpus. Results highlight the analyzers’ reduced performance on L2 data, indicating the limitation of advanced deep-learning models when dealing with L2-Korean corpora. We further show that fine-tuning one of the models with the KLM corpus improves its accuracy of tokenization and POS tagging on L2-Korean dataset.

Towards L2-friendly pipelines for learner corpora: A case of written production by L2-Korean learners

We address the problem of generating high-quality question-answer pairs for educational materials. Previous work on this problem showed that using summaries as input improves the quality of question generation (QG) over original textbook text and that human-written summaries result in higher quality QG than automatic summaries. In this paper, a) we show that advances in Large Language Models (LLMs) are not yet sufficient to generate quality summaries for QG and b) we introduce a new methodology for rewriting bullet point student notes into fully-fledged summaries and find that our methodology yields higher quality QG. We conducted a large-scale human annotation study of generated question-answer pairs for the evaluation of our methodology. In order to aid in future research, we release a novel dataset of 9.2K human annotations of generated questions.

Enhancing Human Summaries for Question-Answer Generation in Education

Question generation (QG) for reading comprehension, a technology for automatically generating questions related to given reading passages, has been used in various applications, including in education. Recently, QG methods based on deep neural networks have succeeded in generating fluent questions that are pertinent to given reading passages. One example of how QG can be applied in education is a reading tutor that automatically offers reading comprehension questions related to various reading materials. In such an application, QG methods should provide questions with difficulty levels appropriate for each learner’s reading ability in order to improve learning efficiency. Several difficulty-controllable QG methods have been proposed for doing so. However, conventional methods focus only on generating questions and cannot generate answers to them. Furthermore, they ignore the relation between question difficulty and learner ability, making it hard to determine an appropriate difficulty for each learner. To resolve these problems, we propose a new method for generating question–answer pairs that considers their difficulty, estimated using item response theory. The proposed difficulty-controllable generation is realized by extending two pre-trained transformer models: BERT and GPT-2.

Difficulty-Controllable Neural Question Generation for Reading Comprehension using Item Response Theory

This paper presents Card-it, a web-based application for learning Italian verb conjugation. Card-it integrates a large-scale finite-state morphological~(FSM) analyzer and a flashcard application as a user-friendly way for learners to utilize the analyzer. While Card-it can be used by individual learners, to support classroom adoption, we implemented simple classroom management functionalities such as sharing flashcards to a class and tracking students' progression. We evaluated Card-it with teachers of Italian. Card-it was reported as engaging and supportive, especially by featuring two different quiz types combined with a verb form look-up feature. Teachers were optimistic about the potential of Card-it as a classroom supplementary tool for learners of Italian as L2. Future work includes sample sentences and a complete learners evaluation.

Downloads

Next from ACL 2023

Recipes for Sequential Pre-training of Multilingual Encoder and Seq2Seq Models

Similar lecture

TADA -- Efficient Task-Agnostic Domain Adaptation for Transformers