A data augmentation technique based on text for Vietnamese sentiment analysis

Huong, Thien Ho; Hoang, Vinh Truong

doi:10.1145/3406601.3406618

Cited by 12 publications

(4 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…To address this, various methods have been explored. These methods range from word swapping [8], deletion [16,17], the introduction of spelling errors [15,18], and paraphrasing [19], to synonym replacement [20][21][22][23], utilizing close embeddings [24][25][26], and employing language models for wordlevel prediction [27][28][29][30]. More advanced techniques include altering dependency trees [31][32][33][34], performing round-trip translation [35][36][37][38], or inserting existing input instances [39][40][41][42].…”

Section: Related Workmentioning

confidence: 99%

Event Extraction Based on Self-Data Augmentation with Large Language Models

Yang,

Fan,

Wang

et al. 2024

Preprint

View full text Add to dashboard Cite

Event extraction plays a crucial role in natural language processing (NLP), facilitating the transformation of unstructured text into structured representations. This conversion significantly enhances the performance of various applications, such as automated question answering and information retrieval systems. However, traditional event extraction methodologies often encounter challenges stemming from limited datasets, imbalanced sample distributions, the necessity for extra resources to annotate large datasets, and the potential for data quality degradation during the augmentation process. To surmount these obstacles, this study introduces an innovative self-data augmentation strategy that leverages a single large language model (LLM) to concurrently perform data augmentation and event extraction. By dynamically assessing and refining the quality of generated samples, this approach mitigates the inclusion of noisy data, ultimately bolstering the model's performance. Demonstrable enhancements in precision, recall, and F1 scores across various model configurations underscore the efficacy of this strategy in managing small and imbalanced datasets. Furthermore, the incorporation of Logical Thoughts for Self-Data Augmentation (LoTSA) ensures the superior quality of augmented data, culminating in more accurate and reliable extraction outcomes.

show abstract

Section: Related Workmentioning

confidence: 99%

Event Extraction Based on Self-Data Augmentation with Large Language Models

Yang,

Fan,

Wang

et al. 2024

Preprint

View full text Add to dashboard Cite

show abstract

“…In the data space, several works operate at the character level [7,16] by swapping, removing, adding letters; injecting common spelling mistakes; or replacing words with abbreviations, e.g., "I'm". Approaches that operate at the word level, use word swap/deletion [3,5,21,25,41] or replacement with synonyms, hypernyms, and antonyms [16]. At the document level, a popular method is round trip translation [1,59].…”

Section: Related Workmentioning

confidence: 99%

Generation of Training Examples for Tabular Natural Language Inference

Bussotti,

Veltri,

Santoro

et al. 2023

Proc. ACM Manag. Data

View full text Add to dashboard Cite

Tabular data is becoming increasingly important in Natural Language Processing (NLP) tasks, such as Tabular Natural Language Inference (TNLI). Given a table and a hypothesis expressed in NL text, the goal is to assess if the former structured data supports or refutes the latter. In this work, we focus on the role played by the annotated data in training the inference model. We introduce a system, Tenet, for the automatic augmentation and generation of training examples for TNLI. Given the tables, existing approaches are either based on human annotators, and thus expensive, or on methods that produce simple examples that lack data variety and complex reasoning. Instead, our approach is built around the intuition that SQL queries are the right tool to achieve variety in the generated examples, both in terms of data variety and reasoning complexity. The first is achieved by evidence-queries that identify cell values over tables according to different data patterns. Once the data for the example is identified, semantic-queries describe the different ways such data can be identified with standard SQL clauses. These rich descriptions are then verbalized as text to create the annotated examples for the TNLI task. The same approach is also extended to create counterfactual examples, i.e., examples where the hypothesis is false, with a method based on injecting errors in the original (clean) table. For all steps, we introduce generic generation algorithms that take as input only the tables. For our experimental study, we use three datasets from the TNLI literature and two crafted by us on more complex tables. Tenet generates human-like examples, which lead to the effective training of several inference models with results comparable to those obtained by training the same models with manually-written examples.

show abstract

“…Thus many methods have been tried out in research so far. Among them are methods for swapping [43], deleting [13,28], inducing spelling mistakes [5,8], paraphrasing [21], and replacing of synonyms [19,47], close embeddings [2,42] and words predicted by a language model [9,14,18] on word-level. On a broader level, methods which change the dependency tree [1,44], perform round-trip-translation [20,33], or interpolate the input instances [7,46] are used.…”

Section: Foundations Of Data Augmentationmentioning

confidence: 99%

Data Augmentation in Natural Language Processing: A Novel Text Generation Approach for Long and Short Text Classifiers

Bayer,

Kaufhold,

Buchhold

et al. 2021

Preprint

View full text Add to dashboard Cite

In many cases of machine learning, research suggests that the development of training data might have a higher relevance than the choice and modelling of classifiers themselves. Thus, data augmentation methods have been developed to improve classifiers by artificially created training data. In NLP, there is the challenge of establishing universal rules for text transformations which provide new linguistic patterns. In this paper, we present and evaluate a text generation method suitable to increase the performance of classifiers for long and short texts. We achieved promising improvements when evaluating short as well as long text tasks with the enhancement by our text generation method. In a simulated low data regime additive accuracy gains of up to 15.53% are achieved. As the current track of these constructed regimes is not universally applicable, we also show major improvements in several real world low data tasks (up to +4.84 F1 score). Since we are evaluating the method from many perspectives, we also observe situations where the method might not be suitable. We discuss implications and patterns for the successful application of our approach on different types of datasets.

show abstract

A data augmentation technique based on text for Vietnamese sentiment analysis

Cited by 12 publications

References 16 publications

Event Extraction Based on Self-Data Augmentation with Large Language Models

Event Extraction Based on Self-Data Augmentation with Large Language Models

Generation of Training Examples for Tabular Natural Language Inference

Data Augmentation in Natural Language Processing: A Novel Text Generation Approach for Long and Short Text Classifiers

Contact Info

Product

Resources

About