GXJoin: Generalized Cell Transformations for Explainable Joinability

Arash Dargahi Nobari; Davood Rafiei; Soroush Omidvartehrani

REVIEW 4 major objections 5 minor 31 references

GXJoin: Generalized Cell Transformations for Explainable Joinability

T0 review · 4 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read Four generalization principles make mismatched table columns joinable with fewer, simpler rules.

desk verdict A solid, honest extension of CST with four generalization techniques, but the recurrence gains may be inflated by set-valued coverage and need a tighter evaluation before the results can be trusted. read the letter →

arxiv 2505.21860 v1 pith:R64OF4OH submitted 2025-05-28 cs.DB

Soroush Omidvartehrani , Arash Dargahi Nobari , Davood Rafiei This is my paper

classification cs.DB

keywords explainablejoincelltransformationtableprogramsynthesisgeneralizationdataintegrationstringentityresolution

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

Two columns from different sources often describe the same entities in different formats, so an equality join fails even though the data are joinable after a syntactic rewrite. The paper tries to establish that such rewrites—cell transformations built from copy, substring, split, and literal units—can be made more general by four principles: relative indices that ignore input length, repetition and optional removal of units for recurring patterns, choosing the source and target direction that yields fewer rules, and preferring simpler rules when coverage ties. On two real-world benchmarks it reports that these generalizations raise best-transformation coverage from 0.577 to 0.679 on web tables and from 0.667 to 0.725 on spreadsheets, cut the number of transformations needed to cover all rows from 431 to 394 and from 970 to 876, and improve end-to-end join F1 from 0.713 to 0.777 on the web benchmark. If true, the practical payoff is that a data analyst can join and explain mismatched columns with fewer, more readable rules that also hold up on newly added rows.

What carries the argument

The machinery is a cell transformation: a sequence of string units—literal, substr, split, and splitSubstr—that maps a source cell to a target cell, with coverage defined as the fraction of source–target pairs the sequence maps correctly. Onto this the paper adds four generalization operators: relative indices anchored at the start or end of the input to make rules length-invariant; a repetition factor r that applies a split or splitSubstr unit to r consecutive tokens; an optional-removal marker ? that lets a unit produce no output; and a bidirectional search that chooses the mapping direction with fewer transformations. A simplicity score, defined as the number of units or parameters, serves as the final tie-breaker. These operators work on top of an existing search for covering transformations, generalizing each candidate rule before or during selection.

What would settle it

Take a held-out set of table pairs where the true formatting rule is deliberately the more complex one (for example, names with variable middle initials require an extra splitSubstr unit while a simpler first-initial rule coincidentally covers the training rows), run the method with simplicity-based tie-breaking, and compare coverage on unseen rows against a random tie-breaking variant; if the simpler rule does not systematically cover more held-out rows, the simplicity prior is falsified.

Watch

Extended reading notes

Core claim

The central claim is that generalization, not just coverage of the given examples, should drive the search for table-join transformations, and that four concrete generalizations achieve this. Relative indices replace absolute positions so a rule like "take the first letter of the first token and the whole last token" works whether or not a middle name sits between them. Unit repetition and unit removal let one pattern absorb variable-length sequences, such as any number of middle-name initials. Bidirectional search picks the mapping direction that yields fewer transformations on a sample, avoiding the failure mode where a longer but less informative column is wrongly treated as the source. Simplicity tie-breaking—preferring fewer units or parameters when coverage is equal—is offered as a way to avoid accidental patterns and to cover unseen rows. The reported experiments on web tables and spreadsheet tables show the generalized transformations matching or beating the state-of-the-art baseline on coverage, number of rules, held-out data, and end-to-end join quality.

Load-bearing premise

The load-bearing premise is that a simpler transformation—one with fewer units or parameters—is more likely to describe unseen rows when two transformations cover the same seen rows, and the paper offers no distributional argument or independent evidence for that link.

Editorial extensions

If this is right

A single generalized rule can replace several single-row rules, so joins over heterogeneous rows (for example, names with and without middle initials) need fewer rules to audit.
Rules learned from as little as 10% of rows transfer to the remaining rows, and the reported held-out coverage gap over the baseline widens as the sample shrinks.
End-to-end unequal joins gain in recall and F1 on noisy web tables—reported F1 rises from 0.713 to 0.777—without changing the explainable unit-based format.
Simplicity tie-breaking cuts the total number of units by at least 5% and parameters by about 17%, which should make re-running transformations on dynamic tables cheaper.
Choosing mapping direction by which side yields fewer rules matters when the longer column is not the more informative one, a situation that is common in web data.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

The simplicity prior is untested against distributions where the true rule is the complex one; one could build a benchmark of formatting rules of deliberately varying complexity and measure held-out coverage to see when tie-breaking by simpler rules hurts.
The paper limits repetition to a factor of 2 and to single-unit or consecutive-unit sequences; extending recursion to nested delimiters or to repetition over arbitrary subsequences could cover hierarchical formats such as address-like or JSON-like cells.
Relative indices are anchored at the start and end of the whole cell; anchoring at token boundaries or punctuation could generalize further, especially for cells with mixed alphabets.
The two benchmarks differ sharply in noise level, so the method's advantage is clearest on messy web tables; a cross-domain stress test with more than two datasets would indicate how portable the generalizations are.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

The paper is a useful, systematic extension of CST for joinable transformations. The four generalization techniques — relative indices, unit repetition/removal, direction generalization, and simplicity-based tie-breaking — are clearly defined and individually evaluated. The headline claim that coverage improves and transformation counts drop is backed by consistent numbers across two benchmarks, and the unseen-data experiment is a genuine held-out evaluation, which is more than many papers in this area do. Credit where due: the authors are open about what each technique contributes and do not hide the fact that gains are modest on the cleaner spreadsheet dataset.

The main soft spot is the one the stress-test flags. Coverage is defined as the fraction of rows where the output set includes the target (Section 3.2). Repetition and removal (Definitions 1 and 2) deliberately make transformations set-valued: repetition emits a range of indexed outputs, and removal adds an empty alternative. So a transformation can cover more rows simply by emitting more candidates, with no increase in semantic specificity. The paper does not quantify output-set sizes or report coverage under a single-output restriction, and the claim that repetition degree one ensures a single output (Section 5.3) does not address removal, which still produces a set including the empty string. This means Tables 1–3 and possibly the unseen-data experiment may overstate the benefit of recurrence generalization. The authors need to either restrict coverage to single-output transformations or show that the gains persist when output-set size is held fixed.

Other issues are more minor but worth fixing. The simplicity tie-breaker is a heuristic with no distributional argument; the unseen-data gains could be dataset-specific, though the comparison between random and simplicity tie-breaking is a good start. The 0.05 support threshold is reported as chosen based on observations; a sensitivity analysis would help. And the abstract says "outperforms" but Table 5a shows GXJoin's F1 on Spreadsheet (0.810) is slightly below CST (0.812) — that should be softened to "on par."

Overall, the ideas are clear, the evaluation is honest in most places, and the paper makes a real contribution to a niche but active subfield. The set-valued coverage issue is load-bearing for the recurrence claims but fixable with additional experiments. I would send this to a serious referee, with the request that they focus on the coverage definition and the recurrence-specific experiments.

Referee Report

4 major / 5 minor

Summary. The paper studies the problem of discovering syntactic cell transformations that make two columns equi-joinable, and proposes four generalization techniques: relative indices, recurrence (unit repetition and unit removal), source-target direction selection, and simplicity-based tie-breaking. The approach is evaluated on the Web and Spreadsheet benchmarks against CST and Auto-join, with metrics including best transformation coverage, number of transformations required, coverage on held-out rows, and end-to-end join F1. The paper reports consistent improvements from relative indices and bidirectional direction selection, smaller gains from recurrence, and improved unseen-data coverage from simplicity tie-breaking.

Significance. If the reported results hold, the paper provides a practical extension of CST-style transformation discovery and is the first systematic study of generality principles for cell transformations. Strengths include a clear formalization of relative indices and recurrence, an explicit held-out row experiment, and public code and datasets. The main caveat is that the coverage metric in the recurrence experiments counts a row as covered when any output of a set-valued transformation matches, so the central generalization claim is only as strong as the single-output robustness analysis.

major comments (4)

[Section 3.2, Definitions 1-2, Tables 1-3] The coverage metric in Section 3.2 is set-inclusion: a pair is covered if the output set contains the target. Definitions 1 and 2 deliberately create set-valued units (split^r and unit?), and the baseline CST is deterministic. Under this asymmetric metric, the coverage improvements in Table 1a/b and Table 3, and the rule-count reductions in Table 2, can be inflated simply because a transformation emits more candidates. The paper must report the distribution of output-set sizes and repeat the coverage and rule-count experiments under a single-output restriction (for example, requiring that exactly one produced candidate equals the target, or disabling recurrence units entirely) before attributing these gains to the semantic generality of the learned rules.
[Section 4.4 and Table 4] The unseen-data experiment relies on the load-bearing premise that simpler transformations, when coverage on seen rows ties, generalize better to future rows. The paper offers no distributional argument or independent evidence for this premise beyond the same two benchmarks. Moreover, Table 4 averages over at least five random splits but reports no confidence intervals or significance tests, so the 0.108 gap at Web 10% could be within sampling variability. The authors should provide a comparison that removes simplicity tie-breaking while keeping the other generalization operators, and should report variance or paired tests for the 10/20/40 sampling conditions.
[Footnote 5 and Section 5.3] The minimum support threshold of 0.05 is chosen after observing that it gives the best performance-recall trade-off in the end-to-end experiments. Since this threshold affects all Table 5 results, the paper should either specify a threshold-selection protocol that does not use the test data or show sensitivity of the F1 results across a range of support values. Otherwise the end-to-end improvements may partly reflect test-set-based tuning.
[Section 5.3] The text states that setting the repetition degree to one 'ensures a single output for each transformation applied on an input,' but Definition 2 still allows unit removal (unit? = unit | ∅), which can produce two outputs for an input. If removal units are disabled in the end-to-end experiments, the paper must state this explicitly; if they remain active, the F1 gains in Table 5 may also be influenced by multi-candidate lookups. A strictly single-output configuration should be reported.

minor comments (5)

[Section 4.2] Definition 1 introduces the repetition factor r, but the relationship between r and the hyperparameter 'repetition degree' is not defined; the paper should specify whether the repetition degree is the maximum value of r or a bound on the total number of repetitions in a transformation.
[Table 3] The denominator for 'Affected Trans.' is unclear: the percentages should state whether they are relative to all transformations, to transformations with coverage strictly between 0 and 1, or to the set of relative-indexed transformations only.
[Section 6] The word 'joinablity' in the first sentence of the conclusion should be corrected to 'joinability'.
[Section 3.1 / Figure 2] The pipeline in Figure 2 shows 'Generalization' as a separate stage after transformation generation, but the text in Section 4 describes generalization applied both during and after generation; the figure and the surrounding explanation should be harmonized.
[Section 4.2] The sentence 'splitSubstr becomes substr when the splitting character does not exist in the input, hence substr is not considered' is confusing because substr appears in Definition 2; the paper should clarify the exact base unit set and the intended scope of the recurrence operations.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: the paper's claims are empirical evaluations on external benchmarks, and its self-citations are used as comparators rather than as load-bearing evidence.

full rationale

The paper's central claims are empirical: coverage, transformation counts, unseen-data coverage, and join F1 are measured on two external benchmarks (Web and Spreadsheet). There is no claimed derivation from first principles whose conclusion equals its premises. The main self-citation is the CST baseline [18], co-authored by two of the current authors; however, CST is used as a state-of-the-art comparator and as the example-generation pipeline, not as authority for the paper's own generalization operators, so it is not load-bearing circularity. The set-valued coverage definition in Section 3.2 (a row is covered if the output set includes the target) is a potential confound for recurrence-generalization gains, since Definitions 1 and 2 enlarge output sets; but the paper discloses this definition, and in Section 5.3 it sets repetition degree to one, 'which ensures a single output' for the end-to-end join experiments, so the headline F1 improvements are not forced by this construction. The simplicity tie-breaking reduction in unit/parameter counts is tautological by design, yet the paper treats it only as a byproduct and separately tests the generalization claim on held-out rows (Table 4), which is an independent empirical check. No uniqueness theorem is imported from the authors' prior work, and no transformation ansatz is smuggled in via citation; remaining concerns, such as the 0.05 support threshold tuned on benchmarks, are correctness and overfitting risks rather than circularity.

Assumptions & free parameters 3 free parameters · 3 assumptions · 0 invented entities

The paper introduces no new entities such as particles, forces, dimensions, or external objects. Its generalization operators are reformulations of transformation syntax. The central claim rests on the transformation language, the availability of example row pairs, and the simplicity prior, plus three manually chosen hyperparameters.

free parameters (3)

maximum number of units = 3
Set in Section 5: 'we set the maximum number of units and repetition degree to 3 and 2 respectively.' A manually chosen search-space bound, not derived.
repetition degree = 2
Set in the same sentence in Section 5; controls how many times units can repeat or be optional, manually chosen.
minimum support threshold = 0.05
Footnote 5: 'the best performance-recall trade-off is achieved when the minimum support for transformations is set to 0.05'; chosen after observing the benchmarks, with no separate validation set or sensitivity analysis.

assumptions (3)

domain assumption Cell transformations are expressible as concatenations of literal, substr, split, and splitSubstr units, extended with relative indices, optionality, and repetition.
Section 3.2 defines the transformation language. If real formatting differences require functions outside this language, the method will miss them. This is inherited from CST and Auto-join.
domain assumption A set of joinable row pairs is provided in advance.
Section 5 states: 'We assume a manually annotated set of joinable row pairs is provided, as finding the matching pairs is beyond the scope of this study.' All coverage and generalization results are measured only on these pairs.
ad hoc to paper Simpler transformations generalize better to unseen data; when coverage ties, fewer units or parameters indicate a more general rule.
Section 4.4 introduces simplicity as a proxy for generality, motivated by accidental patterns in small samples, but gives no proof or distributional argument. The unseen-data experiments in Section 5.2 depend on this heuristic.

how reviews work

0 comments

Cite this review

Pith. "Pith review of GXJoin: Generalized Cell Transformations for Explainable Joinability." pith.science (2026). https://pith.science/paper/R64OF4OH

@misc{pith2026250521860,
  author       = {Pith},
  title        = {Pith review of: GXJoin: Generalized Cell Transformations for Explainable Joinability},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/R64OF4OH}},
  note         = {Machine review of arXiv:2505.21860}
}

read the original abstract

Describing real-world entities can vary across different sources, posing a challenge when integrating or exchanging data. We study the problem of joinability under syntactic transformations, where two columns are not equi-joinable but can become equi-joinable after some transformations. Discovering those transformations is a challenge because of the large space of possible candidates, which grows with the input length and the number of rows. Our focus is on the generality of transformations, aiming to make the relevant models applicable across various instances and domains. We explore a few generalization techniques, emphasizing those that yield transformations covering a larger number of rows and are often easier to explain. Through extensive evaluation on two real-world datasets and employing diverse metrics for measuring the coverage and simplicity of the transformations, our approach demonstrates superior performance over state-of-the-art approaches by generating fewer, simpler and hence more explainable transformations as well as improving the join performance.

Figures

Figures reproduced from arXiv: 2505.21860 by the authors.

**Figure 2.** The end-to-end process of an explainable join [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗

**Figure 1.** The transformation <splitSubstr(‘ ’, s, s, s+1), literal (‘. ’), split(‘ ’, e)> maps the first row of the source table to the first row in the target table, but it cannot map the second row because of the presence of a middle name. However, the first two units can be applied multiple times under repetition. In particular, with at most two repetition of those two units, both rows are covered. Unit Removal As a specia… view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

31 extracted references · 28 canonical work pages

[1]

SyGuS-Comp 2016: Results and Analysis

Alur, R., Fisman, D., Singh, R., Solar-Lezama, A.: Sygus-comp 2016: Results and analysis. arXiv:1611.07627 (2016)

work page Pith review arXiv 2016
[2]

ACM TODS44(3) (2019)

Bonifati, A., Comignani, U., Coquery, E., Thion, R.: Interactive mapping specifi- cation with exemplar tuples. ACM TODS44(3) (2019)

work page 2019
[3]

Cafarella, M.J., Halevy, A., Wang, D.Z., Wu, E., Zhang, Y.: Webtables: Exploring the power of tables on the web. Proc. VLDB Endow.1(1) (Aug 2008)

work page 2008
[4]

ACM TODS42(2) (2017)

Cate, B.T., Kolaitis, P.G., Qian, K., Tan, W.C.: Approximation algorithms for schema-mapping discovery from data examples. ACM TODS42(2) (2017)

work page 2017
[5]

In: 22nd ICDE (2006)

Chaudhuri, S., Ganti, V., Kaushik, R.: A primitive operator for similarity joins in data cleaning. In: 22nd ICDE (2006)

work page 2006
[6]

NeurIPS33 (2020)

Chen, X., Liang, C., Yu, A.W., Song, D., Zhou, D.: Compositional generalization via neural-symbolic stack machines. NeurIPS33 (2020)

work page 2020
[7]

arXiv:2106.04252 (2021)

Conklin, H., Wang, B., Smith, K., Titov, I.: Meta-learning to compositionally gen- eralize. arXiv:2106.04252 (2021)

arXiv 2021
[8]

Gulwani, S.: Automating string processing in spreadsheets using input-output ex- amples. ACM Sigplan Notices46(1) (2011) 5 Our observations indicate that the best performance-recall trade-off is achieved when the minimum support for transformations is set to 0.05, and all experiments in this section are conducted with that threshold. S. Omidvartehrani et al

work page 2011

Show all 31 references

[9]

Communications of the ACM55(8) (2012)

Gulwani, S., Harris, W.R., Singh, R.: Spreadsheet data manipulation using exam- ples. Communications of the ACM55(8) (2012)

2012
[10]

In: Proceedings of the 2016 International Conference on Management of Data (2016)

He, J., Veltri, E., Santoro, D., Li, G., Mecca, G., Papotti, P., Tang, N.: Inter- active and deterministic data cleaning. In: Proceedings of the 2016 International Conference on Management of Data (2016)

2016
[11]

In: CIDR (2015)

Heer, J., Hellerstein, J.M., Kandel, S.: Predictive interaction for data transforma- tion. In: CIDR (2015)

2015
[12]

arXiv:2009.06040 (2020)

Herzig, J., Berant, J.: Span-based semantic parsing for compositional generaliza- tion. arXiv:2009.06040 (2020)

2020 arXiv
[13]

NeurIPS 32 (2019)

Lake, B.M.: Compositional generalization through meta sequence-to-sequence learning. NeurIPS 32 (2019)

2019
[14]

Li, Y., Li, J., Suhara, Y., Doan, A., Tan, W.C.: Deep entity matching with pre- trained language models. Proc. VLDB Endow.14(1) (sep 2020)

2020
[15]

In: Twenty-Third International Joint Conference on Artificial Intelligence (2013)

Ling, X., Halevy, A.Y., Wu, F., Yu, C.: Synthesizing union tables from the web. In: Twenty-Third International Joint Conference on Artificial Intelligence (2013)

2013
[16]

NeurIPS33 (2020)

Liu, Q., An, S., Lou, J.G., Chen, B., Lin, Z., Gao, Y., Zhou, B., Zheng, N., Zhang, D.: Compositional generalization by learning analytical expressions. NeurIPS33 (2020)

2020
[17]

Nargesian, F., Zhu, E., Pu, K.Q., Miller, R.J.: Table union search on open data. Proc. VLDB Endow.11(7) (Mar 2018)

2018
[18]

In: 2022 IEEE 38th ICDE

Nobari, A.D., Rafiei, D.: Efficiently transforming tables for joinability. In: 2022 IEEE 38th ICDE. IEEE (2022)

2022
[19]

https://openrefine.org/ docs/manual/cellediting (2010), accessed Jan 14, 2024

OpenRefine: Openrefine documentation: Cell editing. https://openrefine.org/ docs/manual/cellediting (2010), accessed Jan 14, 2024

2010
[20]

arXiv:1904.09708 (2019)

Russin, J., Jo, J., O’Reilly, R.C., Bengio, Y.: Compositional generalization in a deep seq2seq model by separating syntax and semantics. arXiv:1904.09708 (2019)

2019 arXiv
[21]

arXiv:2204.03758 (2022)

Shi, K., Hong, J., Zaheer, M., Yin, P., Sutton, C.: Compositional generalization and decomposition in neural program synthesis. arXiv:2204.03758 (2022)

2022 arXiv
[22]

Proceedings of the VLDB Endowment9(10) (2016)

Singh, R.: Blinkfill: Semi-supervised programming by example for syntactic string transformations. Proceedings of the VLDB Endowment9(10) (2016)

2016
[23]

arXiv:2010.11988 (2020)

Wang, B., Lapata, M., Titov, I.: Meta-learning for domain generalization in se- mantic parsing. arXiv:2010.11988 (2020)

2020 arXiv
[24]

In: 2011 IEEE 27th ICDE (2011)

Wang, J., Li, G., Fe, J.: Fast-join: An efficient method for fuzzy token matching based string similarity join. In: 2011 IEEE 27th ICDE (2011)

2011
[25]

ACM Trans

Wang, J., Li, G., Feng, J.: Extending string similarity join to tolerant fuzzy token matching. ACM Trans. Database Syst.39(1) (Jan 2014)

2014
[26]

In: Proceedings of the 42nd International ACM SIGIR Conference on Research and Development in Information Retrieval

Zhang, L., Zhang, S., Balog, K.: Table2vec: Neural word and entity embeddings for table population and retrieval. In: Proceedings of the 42nd International ACM SIGIR Conference on Research and Development in Information Retrieval. ACM, New York, NY, USA (2019)

2019
[27]

ACM Trans

Zhang, S., Balog, K.: Semantic table retrieval using keyword and table queries. ACM Trans. Web15(3) (May 2021)

2021
[28]

In: The World Wide Web Conference

Zhao, C., He, Y.: Auto-em: End-to-end fuzzy entity-matching using pre-trained deep models and transfer learning. In: The World Wide Web Conference. WWW ’19, ACM (2019)

2019
[29]

In: Proceedings of the 2019 International Conference on Management of Data

Zhu, E., Deng, D., Nargesian, F., Miller, R.J.: Josie: Overlap set similarity search for finding joinable tables in data lakes. In: Proceedings of the 2019 International Conference on Management of Data. SIGMOD ’19, ACM (2019)

2019
[30]

Proceedings of the VLDB Endowment10(10) (2017)

Zhu, E., He, Y., Chaudhuri, S.: Auto-join: Joining tables by leveraging transfor- mations. Proceedings of the VLDB Endowment10(10) (2017)

2017
[31]

Proceedings of the VLDB Endowment9(12) (2016)

Zhu, E., Nargesian, F., Pu, K.Q., Miller, R.J.: Lsh ensemble: Internet-scale domain search. Proceedings of the VLDB Endowment9(12) (2016)

2016

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.

[1] [1]

SyGuS-Comp 2016: Results and Analysis

Alur, R., Fisman, D., Singh, R., Solar-Lezama, A.: Sygus-comp 2016: Results and analysis. arXiv:1611.07627 (2016)

work page Pith review arXiv 2016

[2] [2]

ACM TODS44(3) (2019)

Bonifati, A., Comignani, U., Coquery, E., Thion, R.: Interactive mapping specifi- cation with exemplar tuples. ACM TODS44(3) (2019)

work page 2019

[3] [3]

Cafarella, M.J., Halevy, A., Wang, D.Z., Wu, E., Zhang, Y.: Webtables: Exploring the power of tables on the web. Proc. VLDB Endow.1(1) (Aug 2008)

work page 2008

[4] [4]

ACM TODS42(2) (2017)

Cate, B.T., Kolaitis, P.G., Qian, K., Tan, W.C.: Approximation algorithms for schema-mapping discovery from data examples. ACM TODS42(2) (2017)

work page 2017

[5] [5]

In: 22nd ICDE (2006)

Chaudhuri, S., Ganti, V., Kaushik, R.: A primitive operator for similarity joins in data cleaning. In: 22nd ICDE (2006)

work page 2006

[6] [6]

NeurIPS33 (2020)

Chen, X., Liang, C., Yu, A.W., Song, D., Zhou, D.: Compositional generalization via neural-symbolic stack machines. NeurIPS33 (2020)

work page 2020

[7] [7]

arXiv:2106.04252 (2021)

Conklin, H., Wang, B., Smith, K., Titov, I.: Meta-learning to compositionally gen- eralize. arXiv:2106.04252 (2021)

arXiv 2021

[8] [8]

Gulwani, S.: Automating string processing in spreadsheets using input-output ex- amples. ACM Sigplan Notices46(1) (2011) 5 Our observations indicate that the best performance-recall trade-off is achieved when the minimum support for transformations is set to 0.05, and all experiments in this section are conducted with that threshold. S. Omidvartehrani et al

work page 2011