Thailand

Interpretability is an important aspect of the trustworthiness of a model&#39;s predictions. Transformer&#39;s predictions are widely explained by the attention weights, i.e., a probability distribution generated at its self-attention unit (head). Current empirical studies provide shreds of evidence that attention weights are not explanations by proving that they are not unique. A recent study showed theoretical justifications to this observation by proving the non-identifiability of attention weights. For a given input to a head and its output, if the attention weights generated in it are unique, we call the weights identifiable. In this work, we provide deeper theoretical analysis and empirical observations on the identifiability of attention weights. Ignored in the previous works, we find the attention weights are more identifiable than we currently perceive by uncovering the hidden role of the key vector. However, the weights are still prone to be non-unique attentions that make them unfit for interpretation. To tackle this issue, we provide a variant of the encoder layer that decouples the relationship between key and value vector and provides identifiable weights up to the desired length of the input. We prove the applicability of such variations by providing empirical justifications on varied text classification tasks. The implementations are available at https://github.com/declare-lab/identifiable-transformers.

ACL-IJCNLP 2021

More Identifiable yet Equally Performant Transformers for Text Classification

text-classification.

identifiability

interpretability

transformers

**Welcome to ACL-IJCNLP 2021!**

The great event is jointly organized by the Association for Computational Linguistics (ACL) and Asian Federation of Natural Language Processing (AFNLP). 

As in previous years, the program of the conference includes a poster session, tutorials, workshops and demonstrations in addition to the main conference.


We were able to keep the registration fees similar to those charged for the virtual ACL 2020. The one fee allows attendance at the main conference and any/all tutorials and workshops. These fees would be $125 Regular Early and $175 Regular Late; $50 Student Early and $75 Student Late. Early registration closes at midnight July 11, 2021 (Eastern Daylight Time).

**Reminder:** It is ACL’s policy that at least one author of each accepted paper (including ACL Finding papers) must register for the conference.

**Reminder2:** Underline site will open closer to the event. If you already registered you will receive access detail

Registration is now open

The great event is jointly organized by the Association for Computational Linguistics (ACL) and Asian Federation of Natural Language Processing (AFNLP).

Interpretability is an important aspect of the trustworthiness of a model's predictions. Transformer's predictions are widely explained by the attention weights, i.e., a probability distribution generated at its self-attention unit (head). Current empirical studies provide shreds of evidence that attention weights are not explanations by proving that they are not unique. A recent study showed theoretical justifications to this observation by proving the non-identifiability of attention weights. For a given input to a head and its output, if the attention weights generated in it are unique, we call the weights identifiable. In this work, we provide deeper theoretical analysis and empirical observations on the identifiability of attention weights. Ignored in the previous works, we find the attention weights are more identifiable than we currently perceive by uncovering the hidden role of the key vector. However, the weights are still prone to be non-unique attentions that make them unfit for interpretation. To tackle this issue, we provide a variant of the encoder layer that decouples the relationship between key and value vector and provides identifiable weights up to the desired length of the input. We prove the applicability of such variations by providing empirical justifications on varied text classification tasks. The implementations are available at https://github.com/declare-lab/identifiable-transformers.

poster

Text simplification reduces the language complexity of professional content for accessibility purposes. End-to-end neural network models have been widely adopted to directly generate the simplified version of input text, usually functioning as a blackbox. We show that
text simplification can be decomposed into a compact pipeline of tasks to ensure the transparency and explainability of the process. The first two steps in this pipeline are often neglected: 1) to predict whether a given piece of text needs to be simplified, and 2) if yes, to identify complex parts of the text. The two tasks can be solved separately using either lexical or deep learning methods, or solved jointly. Simply applying explainable complexity prediction as a preliminary step, the out-of-sample text simplification performance of the state-of-the-art, black-box simplification models can be improved by a large margin.

Explainable Prediction of Text Complexity: The Missing Preliminaries for Text Simplification

NLP is currently dominated by language models like RoBERTa which are pretrained on billions of words. But what exact knowledge or skills do Transformer LMs learn from large-scale pretraining that they cannot learn from less data? To explore this question, we adopt five styles of evaluation: classifier probing, information-theoretic probing, unsupervised relative acceptability judgments, unsupervised language model knowledge probing, and fine-tuning on NLU tasks. We then draw learning curves that track the growth of these different measures of model ability with respect to pretraining data volume using the MiniBERTas, a group of RoBERTa models pretrained on 1M, 10M, 100M and 1B words. We find that these LMs require only about 10M to 100M words to learn to reliably encode most syntactic and semantic features we test. They need a much larger quantity of data in order to acquire enough commonsense knowledge and other skills required to master typical downstream NLU tasks. The results suggest that, while the ability to encode linguistic features is almost certainly necessary for language understanding, it is likely that other, unidentified, forms of knowledge are the major drivers of recent improvements in language understanding among large pretrained models.

When Do You Need Billions of Words of Pretraining Data?

In Neural Machine Translation (and, more generally, conditional language modeling), the generation of a target token is influenced by two types of context: the source and the prefix of the target sequence. While many attempts to understand the internal workings of NMT models have been made, none of them explicitly evaluates relative source and target contributions to a generation decision. We argue that this relative contribution can be evaluated by adopting a variant of Layerwise Relevance Propagation (LRP). Its underlying 'conservation principle' makes relevance propagation unique: differently from other methods, it evaluates not an abstract quantity reflecting token importance, but the proportion of each token's influence. We extend LRP to the Transformer and conduct an analysis of NMT models which explicitly evaluates the source and target relative contributions to the generation process. We analyze changes in these contributions when conditioning on different types of prefixes, when varying the training objective or the amount of training data, and during the training process. We find that models trained with more data tend to rely on source information more and to have more sharp token contributions; the training process is non-monotonic with several stages of different nature.

Analyzing the Source and Target Contributions to Predictions in Neural Machine Translation

Recent years have seen numerous NLP datasets introduced to evaluate the performance of fine-tuned models on natural language understanding tasks. Recent results from large pretrained models, though, show that many of these datasets are largely saturated and unlikely to be able to detect further progress. What kind of datasets are still effective at discriminating among strong models, and what kind of datasets should we expect to be able to detect future improvements? To measure this uniformly across datasets, we draw on Item Response Theory and evaluate 29 datasets using predictions from 18 pretrained Transformer models on individual test examples. We find that Quoref, HellaSwag, and MC-TACO are best suited for distinguishing among state-of-the-art models, while SNLI, MNLI, and CommitmentBank seem to be saturated for current strong models. We also observe span selection task format, which is used for QA datasets like QAMR or SQuAD2.0, is effective in differentiating between strong and weak models.

Comparing Test Sets with Item Response Theory

A growing body of literature has focused on detailing the linguistic knowledge embedded in large, pretrained language models. Existing work has shown that non-linguistic biases in models can drive model behavior away from linguistic generalizations. We hypothesized that competing linguistic processes within a language, rather than just non-linguistic model biases, could obscure underlying linguistic knowledge. We tested this claim by exploring a single phenomenon in four languages: English, Chinese, Spanish, and Italian. While human behavior has been found to be similar across languages, we find cross-linguistic variation in model behavior. We show that competing processes in a language act as constraints on model behavior and demonstrate that targeted fine-tuning can re-weight the learned constraints, uncovering otherwise dormant linguistic knowledge in models. Our results suggest that models need to learn both the linguistic constraints in a language and their relative ranking, with mismatches in either producing non-human-like behavior.

Uncovering Constraint-Based Behavior in Neural Models via Targeted Fine-Tuning

The application of transformer-based contextual representations has became a de facto solution for solving complex NLP tasks. Despite their successes, such representations are arguably opaque as their latent dimensions are not directly interpretable. To alleviate this limitation of contextual representations, we devise such an algorithm where the output representation expresses human-interpretable information of each dimension. 
We achieve this by constructing a transformation matrix based on the semantic content of the embedding space and predefined semantic categories using Hellinger distance.
We evaluate our inferred representations on supersense prediction task. Our experiments reveal that the interpretable nature of transformed contextual representations makes it possible to accurately predict the supersense category of a word by simply looking for its transformed coordinate with the largest coefficient. We quantify the effects of our proposed transformation when applied over traditional dense contextual embeddings. We additionally investigate and report consistent improvements for the integration of sparse contextual word representations into our proposed algorithm.

Changing the Basis of Contextual Representations with Explicit Semantics

Despite the development of pre-trained language models (PLMs) significantly raise the performances of various Chinese natural language processing (NLP) tasks, the vocabulary (vocab) for these Chinese PLMs remains to be the one provided by Google Chinese BERT~\citep{Devlin2019BERTPO}, which is based on Chinese characters (chars). Second, the masked language model pre-training is based on a single vocab, limiting its downstream task performances. In this work, we first experimentally demonstrate that building a vocab via Chinese word segmentation (CWS) guided sub-word tokenization (SGT) can improve the performances of Chinese PLMs. Then we propose two versions of multi-vocab pre-training (MVP), Hi-MVP and AL-MVP, to improve the models' expressiveness. Experiments show that: (a) MVP training strategies improve PLMs' downstream performances, especially it can improve the PLM's performances on span-level tasks; (b) our AL-MVP outperforms the recent AMBERT \citep{Zhang2020AMBERTAP} after large-scale pre-training, and it is more robust against adversarial attacks.

MVP-BERT: Multi-Vocab Pre-training for Chinese BERT

The internet has actually come to be an essential resource of health knowledge for individuals around the world in the present situation of the coronavirus condition pandemic(COVID- 19). During pandemic situations, myths, sensationalism, rumours and misinformation, generated intentionally or unintentionally, spread rapidly through social networks. Twitter is one of these popular social networks people use to share COVID-19 related news, information, and thoughts that reflect their perception and opinion about the pandemic. Evaluation of tweets for recognizing misinformation can create beneficial understanding to review the top quality and also the readability of online information concerning the COVID-19. This paper presents a multilingual COVID-19 related tweet analysis method, CMTA, that uses BERT, a deep learning model for multilingual tweet misinformation detection and classification. CMTA extracts features from multilingual textual data, which is then categorized into specific information classes. Classification is done by a Dense-CNN model trained on tweets manually annotated into information classes (i.e., ’false’, ’partly false’, ’misleading’). The paper presents an analysis of multilingual tweets from February to June, showing the distribution type of information spread across different languages. To access the performance of the CMTA multilingual model, we performed a comparative analysis of 8 monolingual model and CMTA for the misinformation detection task. The results show that our proposed CMTA model has surpassed various monolingual models which consolidated the fact that through transfer learning a multilingual framework could be developed

CMTA: COVID-19 Misinformation Multilingual Analysis on Twitter

Individuals with autism spectrum disorder (ASD) experience difficulties in social aspects of communication, but the linguistic characteristics associated with deficits in discourse and pragmatic expression are often difficult to precisely identify and quantify. We are currently collecting a corpus of transcribed natural conversations produced in an experimental setting in which participants with and without ASD complete a number of collaborative tasks with their neurotypical peers. Using this dyadic conversational data, we investigate three pragmatic features – politeness, uncertainty, and informativeness – and present a dataset of utterances annotated for each of these features on a three-point scale. We then introduce ongoing work in developing and training neural models to automatically predict these features, with the goal of identifying the same between-groups differences that are observed using manual annotations. We find the best performing model for all three features is a feedforward neural network trained with BERT embeddings. Our models yield higher acc racy than ones used in previous approaches for deriving these features, with F1 exceeding 0.82 for all three pragmatic features.

Predicting pragmatic discourse features in the language of adults with autism spectrum disorder

We present the first Chinese adversarial natural language inference (NLI) evaluation set (CANLI) to probe the limits of pretrained language models (PLMs). CANLI is based on the linguistic phenomenon of causative- passive homonymy (CPH). The disambiguation of CPH is easy for humans, but difficult for machines, as the causative and the passive are not distinguished by the sentences’ syntactic structure. We created the CANLI in two related settings: (i) CA: a syntactic heuristic setting and (ii) CB: a semantic role labeling setting. We find that state-of-the-art PLMs that are fine-tuned on existing large-scale Chinese NLI benchmark datasets perform poorly on CANLI. Further, we notice that the PLM’s performance on CA and CB is independent of each other, suggesting that the PLM has not learned the shared underlying linguistic knowledge CPH, but instead the respective spurious patterns. We conclude that there is considerable room for improvement in the NLI system.

Downloads

Next from ACL-IJCNLP 2021

Explainable Prediction of Text Complexity: The Missing Preliminaries for Text Simplification

Similar lecture

Causal Analysis of Syntactic Agreement Mechanisms in Neural Language Models