294 Hits in 3.4 sec

Structure-Grounded Pretraining for Text-to-SQL [article]

Xiang Deng, Ahmed Hassan Awadallah, Christopher Meek, Oleksandr Polozov, Huan Sun, Matthew Richardson
2020 arXiv   pre-print
In this paper, we present a novel weakly supervised Structure-Grounded pretraining framework (StruG) for text-to-SQL that can effectively learn to capture text-table alignment based on a parallel text-table  ...  Learning to capture text-table alignment is essential for table related tasks like text-to-SQL.  ...  Conclusion In this paper, we propose a novel while effective structure-grounded pretraining technique for textto-SQL.  ... 
arXiv:2010.12773v1 fatcat:5nqxshdayvayrp3yi7odhlvmme

Structure-Grounded Pretraining for Text-to-SQL

Xiang Deng, Ahmed Hassan Awadallah, Christopher Meek, Oleksandr Polozov, Huan Sun, Matthew Richardson
2021 Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies   unpublished
In this paper, we present a novel weakly supervised Structure-Grounded pretraining framework (STRUG) for text-to-SQL that can effectively learn to capture text-table alignment based on a parallel text-table  ...  Learning to capture text-table alignment is essential for tasks like text-to-SQL.  ...  Acknowledgements We thank Bo Pang, Tao Yu for their help with the official Spider evaluation. We also thank anonymous reviewers for their constructive feedback.  ... 
doi:10.18653/v1/2021.naacl-main.105 fatcat:j23kjdrtgndxroks4hs5glwrfi

SPARQLing Database Queries from Intermediate Question Decompositions [article]

Irina Saparina, Anton Osokin
2022 arXiv   pre-print
We chose SPARQL because its queries are structurally closer to our intermediate representations (compared to SQL).  ...  We observe that the execution accuracy of queries constructed by our model on the challenging Spider dataset is comparable with the state-of-the-art text-to-SQL methods trained with annotated SQL queries  ...  Related Work Text-to-SQL.  ... 
arXiv:2109.06162v2 fatcat:44ngmotje5fs5a7jevlchsaocy

Awakening Latent Grounding from Pretrained Language Models for Semantic Parsing [article]

Qian Liu, Dejian Yang, Jiahui Zhang, Jiaqi Guo, Bin Zhou, Jian-Guang Lou
2021 arXiv   pre-print
Taking text-to-SQL as a case study, we successfully couple our approach with two off-the-shelf parsers, obtaining an absolute improvement of up to 9.8%.  ...  To better understand and leverage what PLMs have learned, several techniques have emerged to explore syntactic structures entailed by PLMs.  ...  Acknowledgement We would like to thank all the anonymous reviewers for their constructive feedback and useful comments.  ... 
arXiv:2109.10540v1 fatcat:j7y2jbponjfkpolzniyu5y5xxq

Self-supervised Text-to-SQL Learning with Header Alignment Training [article]

Donggyu Kim, Seanie Lee
2021 arXiv   pre-print
We utilize the task-specific properties of Text-to-SQL task and the underlying structures of table contents to train the models to learn useful knowledge of the header-column alignment task from unlabeled  ...  In order to tackle such discrepancy in Text-to-SQL task, we propose a novel self-supervised learning framework.  ...  for Text-to-SQL.  ... 
arXiv:2103.06402v1 fatcat:ymstvrwqjvffrp6syqklssm4e4

RASAT: Integrating Relational Structures into Pretrained Seq2Seq Model for Text-to-SQL [article]

Jiexing Qi, Jingyao Tang, Ziwei He, Xiangpeng Wan, Chenghu Zhou, Xinbing Wang, Quanshi Zhang, Zhouhan Lin
2022 arXiv   pre-print
However, introducing these structural relations comes with prices: they often result in a specialized model structure, which largely prohibits the use of large pretrained models in text-to-SQL.  ...  Relational structures such as schema linking and schema encoding have been validated as a key component to qualitatively translating natural language into SQL queries.  ...  Faced with the multi-table and complex SQL setting, using graph structures to encode a variety of complex relationships is a major trend in the text-to-SQL task.  ... 
arXiv:2205.06983v1 fatcat:tau5dylefja6fdzbtil2jgyu2y

Topic Transferable Table Question Answering [article]

Saneem Ahmed Chemmengath, Vishwajeet Kumar, Samarth Bharadwaj, Jaydeep Sen, Mustafa Canim, Soumen Chakrabarti, Alfio Gliozzo, Karthik Sankaranarayanan
2021 arXiv   pre-print
In response, we propose T3QA (Topic Transferable Table Question Answering) a pragmatic adaptation framework for TableQA comprising of: (1) topic-specific vocabulary injection into BERT, (2) a novel text-to-text  ...  Weakly-supervised table question-answering(TableQA) models have achieved state-of-art performance by using pre-trained BERT transformer to jointly encoding a question and a table to produce structured  ...  Correspondingly, there is also less topic specific text to pretrain the TaBERT encoder.  ... 
arXiv:2109.07377v1 fatcat:4y4y5gj2tfdz7iysutdl2coo4m

CQR-SQL: Conversational Question Reformulation Enhanced Context-Dependent Text-to-SQL Parsers [article]

Dongling Xiao, Linzheng Chai, Qian-Wen Zhang, Zhao Yan, Zhoujun Li, Yunbo Cao
2022 arXiv   pre-print
Secondly, we train CQR-SQL models to map the semantics of multi-turn questions and auxiliary self-contained questions into the same latent space through schema grounding consistency task and tree-structured  ...  Context-dependent text-to-SQL is the task of translating multi-turn questions into database-related SQL queries.  ...  We conduct a self-training approach with a pretrained single-turn text-to-SQL model θ SQL to collect full self-contained question data D cqr for textto-SQL datasets D, as show in Algorithm 1.  ... 
arXiv:2205.07686v2 fatcat:nh2utaetdfes5noa4esf64ijyq

End-to-End Cross-Domain Text-to-SQL Semantic Parsing with Auxiliary Task [article]

Peng Shi, Tao Yu, Patrick Ng, Zhiguo Wang
2021 arXiv   pre-print
In this work, we focus on two crucial components in the cross-domain text-to-SQL semantic parsing task: schema linking and value filling.  ...  filling in the synthesized SQL.  ...  The decoder firstly generates the overall structure of the target SQL such as SELECT __ WHERE __ ORDER BY __, and then conducts schema grounding in the fine-grained decoding process, including the column  ... 
arXiv:2106.09588v1 fatcat:zalz33l5rndj7ifsyyy7cxpydu

DuoRAT: Towards Simpler Text-to-SQL Models [article]

Torsten Scholak, Raymond Li, Dzmitry Bahdanau, Harm de Vries, Chris Pal
2020 arXiv   pre-print
Contrary to this trend, in this paper we identify the aspects in which text-to-SQL models can be simplified.  ...  Recent research has shown that neural text-to-SQL models can effectively translate natural language questions into corresponding SQL queries on unseen databases.  ...  Acknowledgments We thank Christopher Manning from Stanford and Eric Laufer from the Knowledge Scout team at Element AI for fruitful discussions on the subject of text-to-SQL translation.  ... 
arXiv:2010.11119v1 fatcat:i54pxirvgfbhvbvqs2zyqwzbli

Natural Language to Code Translation with Execution [article]

Freda Shi, Daniel Fried, Marjan Ghazvininejad, Luke Zettlemoyer, Sida I. Wang
2022 arXiv   pre-print
In this work, we introduce execution result--based minimum Bayes risk decoding (MBR-EXEC) for program selection and show that it improves the few-shot performance of pretrained code models on natural-language-to-code  ...  Generative models of code, pretrained on large corpora of programs, have shown great success in translating natural language to code (Chen et al., 2021; Austin et al., 2021; Li et al., 2022, inter alia  ...  The Spider dataset (Yu et al., 2018) 5 is a text-to-SQL dataset, which requires a model to translate text descriptions into SQL commands.  ... 
arXiv:2204.11454v1 fatcat:tufiesp5vvcgxho45obcon67rq

Dual Reader-Parser on Hybrid Textual and Tabular Evidence for Open Domain Question Answering [article]

Alexander Hanbo Li, Patrick Ng, Peng Xu, Henghui Zhu, Zhiguo Wang, Bing Xiang
2021 arXiv   pre-print
In a detailed analysis, we demonstrate that the being able to generate structural SQL queries can always bring gains, especially for those questions that requires complex reasoning.  ...  However, a large amount of world's knowledge is stored in structured databases, and need to be accessed using query languages such as SQL.  ...  For tabular evidence, the models either predict direct answer text or generate structure SQL queries.  ... 
arXiv:2108.02866v2 fatcat:t4zs7qglnrahxedkxugwbxu5xa

Linking-Enhanced Pre-Training for Table Semantic Parsing [article]

Bowen Qin, Lihan Wang, Binyuan Hui, Ruiying Geng, Zheng Cao, Min Yang, Jian Sun, Yongbin Li
2022 arXiv   pre-print
objectives to impose the desired inductive bias into the learned representations for table pre-training.  ...  their semantic and structural correspondence.  ...  It represents the maximum potential benefit of schema linking for the text-to-SQL task (Liu et al. 2021) .  ... 
arXiv:2111.09486v3 fatcat:h4kkm5arcrad3ef2tl4nwpoob4

SeqZero: Few-shot Compositional Semantic Parsing with Sequential Prompts and Zero-shot Models [article]

Jingfeng Yang, Haoming Jiang, Qingyu Yin, Danqing Zhang, Bing Yin, Diyi Yang
2022 arXiv   pre-print
Recent research showed promising results on combining pretrained language models (LMs) with canonical utterance for few-shot semantic parsing.  ...  The canonical utterance is often lengthy and complex due to the compositional structure of formal languages.  ...  The LMs are usually pretrained on large unlabeled open-domain natural language data and achieve impressive performance on fewshot text-to-text generation problems via proper prompt designing (Brown et  ... 
arXiv:2205.07381v1 fatcat:kjewe6sphjc2jgbtsnyo6e5wme

FORTAP: Using Formulas for Numerical-Reasoning-Aware Table Pretraining [article]

Zhoujun Cheng, Haoyu Dong, Ran Jia, Pengfei Wu, Shi Han, Fan Cheng, Dongmei Zhang
2022 arXiv   pre-print
We design two formula pretraining tasks to explicitly guide FORTAP to learn numerical reference and calculation in semi-structured tables.  ...  FORTAP is the first method for numerical-reasoning-aware table pretraining by leveraging large corpus of spreadsheet formulae.  ...  Some works mine large-scale table-text pairs as pretraining corpus adopts BART(Lewis et al.)as a neural executor for synthesized SQLs to improve table reasoning.  ... 
arXiv:2109.07323v2 fatcat:yfcm7yjppffk7mvwjhdupelpjm
« Previous Showing results 1 — 15 out of 294 results