TEAR is the first framework that supports automated text-driven attribute recommendation, enabling exploratory schema design for table extraction and attribute recommendation on naturally occurring texts and achieves state-of-the-art performance on both tasks.
Abstract
Table extraction from texts is an important task for information systems, and recent approaches that prompt large language models (LLMs) with instructions have drawn great attention for their strong performance. Existing works have assumed the input texts to be table descriptions or specialized documents. However, these efforts have largely overlooked another prevalent category of texts, commonly found in news reports and social media: naturally occurring texts. Extracting tabular information from such texts poses two distinct challenges. First, high variability and the absence of explicit structural cues make fixed heuristic LLM prompts limited in precisely delineating extraction boundaries. Second, manually predefined schemas cannot capture open-ended, unseen attributes in naturally occurring text. In this paper, we propose a framework, TEAR, to address these challenges. It comprises two synergistic workflows: a Table Extraction Workflow that dynamically adapts instructions to overcome the limitation of heuristic instructions, and an Attribute Recommendation Workflow that discovers new attributes from texts to complement the heuristic schema. To our knowledge, TEAR is the first framework that supports automated text-driven attribute recommendation, enabling exploratory schema design for table extraction. To evaluate TEAR, we establish the benchmark for table extraction and attribute recommendation on naturally occurring texts, including two real-world datasets, manual annotations, appropriate metrics, and baseline comparisons. Experiments show that TEAR achieves state-of-the-art performance on both tasks, and the recommended attributes effectively enhance extraction performance in exploratory scenarios.
Results show that attributes comparable to those identified by humans can be discovered automatically, enabling the creation of high-quality structured datasets economically and at scale.
This study presents a systematic study evaluating URL extraction across six input formats and indicates that TEXTWAL achieves the best performance among single-format inputs, while TEXTWAL+LaTeX achieves the best overall URL extraction performance.
Rochana R. Obadage, Lamia Salsabil, Sawood Alam et al.· Proceedings of the 26th ACM/...· 1 citation
Large language models can extract useful signals from heterogeneous enterprise data, but high-recall extraction often produces outputs that are duplicated, uneven in granularity, semantically overlapping, or too numerous for downstream systems and human reviewers to use effectively. We present a dataset-adaptive post-p...
Mahesh Viswanathan, Joan Rossello, L. Fernandes et al.· 0 citations
This survey conducts a comprehensive review of the literature on table mining with Large Language Models, exploring the challenges unique to this domain, such as heterogeneous table structures, contextual ambiguity, and domain-specific knowledge requirements.
Mingyue Cheng, Qingyang Mao, Qi Liu et al.· ACM Computing Surveys· 13 citations
In the data extraction process for job-person recommendation systems, the data structures present in various forms including structured, semi-structured, and unstructured formats, which leads to significant difficulty and low efficiency in manual knowledge graph construction. Although deep learning techniques have adva...
Yuan-Han Xie, Hao-Chen Gong, Dan-Ni Hu et al.· 2026 6th International Confe...· 0 citations
EviMap is an interactive system providing researchers and practitioners with an auditable thematic overview of heterogeneous corpora spanning 2,108 to 101,699 documents, with a comparison against flat and hierarchical LLM baselines.