Sentiment analysis is a crucial task in natural language processing that involves identifying and extracting subjective sentiment from text. Self-training has recently emerged as an economical and efficient technique for developing sentiment analysis models by leveraging a small amount of labeled data and a large amount of unlabeled data. However, given a set of training data, how to utilize them to conduct self-training makes a significant difference in the final performance of the model. We refer to this methodology as the self-training strategy. In this paper, we present an empirical study of various self-training strategies for sentiment analysis. First, we investigate the influence of the self-training strategy and hyper-parameters on the performance of traditional small language models (SLMs) in various few-shot settings. Second, we also explore the feasibility of leveraging large language models (LLMs) to help self-training. We propose and empirically compare several self-training strategies with the intervention of LLMs. Extensive experiments are conducted on three real-world sentiment analysis datasets.

Many existing end-to-end systems for hybrid question answering tasks can often be boiled down to a “prompt-and-pray” paradigm, where the user has limited control and insight into the intermediate reasoning steps used to achieve the final result. Additionally, due to the context size limitation of many transformer-based LLMs, it is often not reasonable to expect that the full structured and unstructured context will fit into a given prompt in a zero-shot setting, let alone a few-shot setting. We introduce BlendSQL, a superset of SQLite to act as a unified dialect for orchestrating reasoning across both unstructured and structured data. For hybrid question answering tasks involving multi-hop reasoning, we encode the full decomposed reasoning roadmap into a single interpretable BlendSQL query. Notably, we show that BlendSQL can scale to massive datasets and improve the performance of end-to-end systems while using 35% fewer tokens. Our code is available and installable as a package at https://github.com/parkervg/blendsql.

In this paper, we describe the different approaches explored by the Jetsons team for the Multi-Lingual ESG Impact Duration Inference (ML-ESG-3) shared task. The shared task focuses on predicting the duration and type of the ESG impact of a news article. The shared task dataset consists of 2,059 news titles and articles in English, French, Korean, and Japanese languages. For the impact duration classification task, we fine-tuned XLM-RoBERTa with a custom fine-tuning strategy and using self-training and DeBERTa-v3 using only English translations. These models individually ranked first on the leaderboard for Korean and Japanese and in an ensemble for the English language, respectively. For the impact type classification task, our XLM-RoBERTa model fine-tuned using a custom fine-tuning strategy ranked first for the English language.

In addressing the task of converting natural language to SQL queries, there are several semantic and syntactic challenges. It becomes increasingly important to understand and remedy the points of failure as the performance of semantic parsing systems improve. We explore semantic parse correction with natural language feedback, proposing a new solution built on the success of autoregressive decoders in text-to-SQL tasks. By separating the semantic and syntactic difficulties of the task, we show that the accuracy of text-to-SQL parsers can be boosted by up to 26% with only one turn of correction with natural language. Additionally, we show that a T5-base model is capable of correcting the errors of a T5-large model in a zero-shot, cross-parser setting.

Clinical risk scores enable clinicians to tabulate a set of patient data into simple scores to stratify patients into risk categories. Although risk scores are widely used to inform decision-making at the point-of-care, collecting the information necessary to calculate such scores requires considerable time and effort. Previous studies have focused on specific risk scores and involved manual curation of relevant terms or codes and heuristics for each data element of a risk score. To support more generalizable methods for risk score calculation, we annotate 100 patients in MIMIC-III with elements of CHA2DS2-VASc and PERC scores, and explore using question answering (QA) and off-the-shelf tools. We show that QA models can achieve comparable or better performance for certain risk score elements as compared to heuristic-based methods, and demonstrate the potential for more scalable risk score automation without the need for expert-curated heuristics. Our annotated dataset will be released to the community to encourage efforts in generalizable methods for automating risk scores.

In this paper, we discuss the various approaches by the Jetsons team for the “Pairwise Comparison” sub-task of the ERAI shared task to compare financial opinions for profitability and loss. Our BERT-Chinese model considers a pair of opinions and predicts the one with a higher maximum potential profit (MPP) with 62.07% accuracy. We analyze the performance of our approaches on both the MPP and maximal loss (ML) problems and deeply dive into why BERT-Chinese outperforms other models.

In this paper, we present a system that addresses the taxonomy enrichment problem for Environment, Social and Governance issues in the financial domain, as well as classifying sentences as sustainable or unsustainable, for FinSim4-ESG, a shared task for the FinNLP workshop at IJCAI-2022. We first created a derived dataset for taxonomy enrichment by using a sentence-BERT-based paraphrase detector (Reimers and Gurevych, 2019) (on the train set) to create positive and negative term-concept pairs. We then model the problem by fine-tuning the sentence-BERT-based paraphrase detector on this derived dataset, and use it as the encoder, and use a Logistic Regression classifier as the decoder, resulting in test Accuracy: 0.6 and Avg. Rank: 1.97. In case of the sentence classification task, the best-performing classifier (Accuracy: 0.92) consists of a pre-trained RoBERTa model (Liu et al., 2019a) as the encoder and a Feed Forward Neural Network classifier as the decoder.

Existing question answering (QA) datasets derived from electronic health records (EHR) are artificially generated and consequently fail to capture realistic physician information needs. We present Discharge Summary Clinical Questions (DiSCQ), a newly curated question dataset composed of 2,000+ questions paired with the snippets of text (triggers) that prompted each question. The questions are generated by medical experts from 100+ MIMIC-III discharge summaries. We analyze this dataset to characterize the types of information sought by medical experts. We also train baseline models for trigger detection and question generation (QG), paired with unsupervised answer retrieval over EHRs. Our baseline model is able to generate high quality questions in over 62% of cases when prompted with human selected triggers. We release this dataset (and all code to reproduce baseline model results) to facilitate further research into realistic clinical QA and QG: https://github.com/elehman16/discq.

We present emrKBQA, a dataset for answering physician questions from a structured patient record. It consists of questions, logical forms and answers. The questions and logical forms are generated based on real-world physician questions and are slot-filled and answered from patients in the MIMIC-III KB through a semi-automated process. This community-shared release consists of over 940000 question, logical form and answer triplets with 389 types of questions and ~7.5 paraphrases per question type. We perform experiments to validate the quality of the dataset and set benchmarks for question to logical form learning that helps answer questions on this dataset.

We address the problem of model generalization for sequence to sequence (seq2seq) architectures. We propose going beyond data augmentation via paraphrase-optimized multi-task learning and observe that it is useful in correctly handling unseen sentential paraphrases as inputs. Our models greatly outperform SOTA seq2seq models for semantic parsing on diverse domains (Overnight - up to 3.2% and emrQA - 7%) and Nematus, the winning solution for WMT 2017, for Czech to English translation (CzENG 1.6 - 1.5 BLEU).

We explore state-of-the-art neural models for question answering on electronic medical records and improve their ability to generalize better on previously unseen (paraphrased) questions at test time. We enable this by learning to predict logical forms as an auxiliary task along with the main task of answer span detection. The predicted logical forms also serve as a rationale for the answer. Further, we also incorporate medical entity information in these models via the ERNIE architecture. We train our models on the large-scale emrQA dataset and observe that our multi-task entity-enriched models generalize to paraphrased questions ~5% better than the baseline BERT model.

We propose a novel methodology to generate domain-specific large-scale question answering (QA) datasets by re-purposing existing annotations for other NLP tasks. We demonstrate an instance of this methodology in generating a large-scale QA dataset for electronic medical records by leveraging existing expert annotations on clinical notes for various NLP tasks from the community shared i2b2 datasets. The resulting corpus (emrQA) has 1 million questions-logical form and 400,000+ question-answer evidence pairs. We characterize the dataset and explore its learning potential by training baseline models for question to logical form and question to answer mapping.