Pith. sign in

REVIEW 3 major objections 5 minor 88 references

Texture: Structured Exploration of Text Datasets

T0 review · 3 major / 5 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read One tool, ten datasets: configurable attributes speed text exploration

desk verdict Solid systems contribution with a credibly expressive schema, but the effectiveness claims outrun a non-counterbalanced, self-reported study. read the letter →

arxiv 2504.16898 v1 pith:YGSLISLP submitted 2025-04-23 cs.HC

Will Epperson , Arpit Mathur , Adam Perer , Dominik Moritz This is my paper · ORCID
classification cs.HC
keywords textvisualizationexploratorydataanalysisconfigurableschemacross-filteringdocumentembeddingssimilaritysearchdatasetsuserstudy
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

This paper tries to establish that exploratory analysis of text corpora can be served by one general-purpose tool rather than a patchwork of task- and domain-specific systems. The proposed system, Texture, rests on a configurable schema that describes any descriptive attribute of a document (single values, lists, text spans, or vector embeddings) and on linked interactions that let users overview attributes, cross-filter them, and read the matching documents. In a two-session study, ten analysts each brought their own dataset; the paper reports that every attribute from their existing workflows could be represented, that participants iterated more quickly, and that several uncovered insights their prior analyses missed. The significance, if the claim holds, is that flexible attribute-based exploration can be built once and reused across domains, with attribute derivation left to whatever code or model the analyst prefers.

What carries the argument

The central object is the five-type attribute schema (text, single-value, list, span list, and embedding) that lets any text dataset be described at arbitrary granularity. The load-bearing mechanism is normalization: multi-valued list attributes are split into separate tables linked back to documents, and span lists additionally store the character positions of each occurrence. That layout lets every overview chart, filter, and search be executed as relational SQL queries over joined tables, which is what makes cross-filtering between word-level and document-level attributes fast and makes span highlighting unambiguous. Embeddings are treated as a first-class attribute type so that projection overviews and similarity search plug into the same filtering pipeline as any other attribute.

What would settle it

A counterbalanced study where each participant analyzes the same dataset twice, once with Texture and once with their usual workflow, would settle the central claim: if the usual-workflow second session produces comparable new insights at comparable speed, the reported gains are repeated-exposure effects, not tool effects.

Watch

Extended reading notes

Core claim

The paper's central discovery is that a text exploration tool can be made general-purpose by separating what an analyst derives from how it is explored. It defines five attribute types (text, single-value, list, span list, and embedding) and prescribes storing multi-valued attributes in normalized tables linked to documents by foreign keys, with span lists keeping character offsets. Around this schema the interface combines automatic per-attribute overview charts, cross-filtering across any combination of attributes, an embedding projection with similarity search, and a document view that highlights matching spans. The user study with ten participants and ten distinct datasets is offered as evidence that this design is expressive enough to absorb all the attributes analysts already use, fast enough to shorten the iteration loop, and capable of surfacing new findings such as duplicate-heavy clusters in a corpus.

Load-bearing premise

The evaluation assumes that the faster iteration and new insights seen in the second session came from the tool rather than from participants having already spent a session thinking about the same familiar dataset, because the baseline always came first and the sessions were not counterbalanced.

Editorial extensions

If this is right

  • Analysts can move between datasets from different domains (research abstracts, song lyrics, Reddit posts, chatbot logs) without switching tools, as long as they can express their attributes in the five-type schema.
  • Because attribute derivation is decoupled from exploration, any attribute computable in code can be explored immediately, including LLM-derived topics or custom metadata.
  • Embedding overviews and similarity search become routine rather than expert-only, letting users find duplicate clusters and outliers even when they would not have computed embeddings themselves.
  • Cross-filtering across granularities lets a user start from a word-level filter and immediately see document-level distributions (for example, the years or authors associated with that word), supporting top-down and bottom-up analysis paths.

Reading between the lines

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

  • The schema's normalized table layout implies a direct path to scaling, because all filters compile to joins over indexed tables; the approach could likely extend well beyond the 16,000-document maximum in the study.
  • The two data-quality discoveries (duplicate posts and near-duplicate prompts) suggest that embedding-based exploration doubles as a corpus audit tool, a use the paper documents but does not frame as a primary contribution.
  • A natural, untested extension is bringing attribute derivation into the interface (for example, LLM-generated attributes verified immediately), which would close the loop the paper identifies as future work and could make the reported iteration gains larger.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. Texture proposes a configurable data schema for text datasets (text, single-value, list, span-list, and embedding attribute types) and an interactive tool that combines automatic attribute overview visualizations, cross-filtering, embedding-based overview and similarity search, and document-level contextualization. The paper reports a two-part user study with 10 participants, each of whom first demonstrated their baseline analysis workflow on their own dataset and later explored the same dataset with Texture. The central claims are that the schema is expressive enough to represent all participant-derived attributes, that Texture enables users to iterate more quickly, and that it leads to new insights about their data.

Significance. If the effectiveness claims held, this would be a solid contribution to text visualization: the attribute schema is a simple and sensible formalization that covers common granularities, and the system integrates familiar interaction techniques in a coherent, open-source implementation. The use of ten real datasets contributed by participants is a real strength, and the direct demonstration that all participant attributes fit the schema is convincing. However, the comparative evaluation is not strong enough to support the speed and insight-discovery claims, because the design conflates the interface with added attributes and with mere repeated exposure. The core conceptual and engineering contribution is defensible, but the empirical claims as stated in the abstract require either stronger evidence or careful tempering.

major comments (3)
  1. [Section 4.1 and Section 6.4] The evaluation does not control for learning or order effects. All participants completed the baseline session before the Texture session on the same dataset, with no counterbalancing and no control condition. The baseline session was an active think-aloud interview that required participants to re-articulate their analysis questions, demonstrate their workflow, and reflect on the data; that process alone can generate additional insights on a second encounter. Section 6.4 acknowledges this possibility ('additional insights could come simply from additional exposure') but dismisses it because participants already knew their datasets, without any test or evidence. This assumption is load-bearing for the abstract claims that Texture 'enabled participants to more quickly iterate' and 'discover new insights.' As written, the observed differences could be explained by repeated exposure or by the added baseline reflection, so the comparative claims are not supported.
  2. [Section 4.1 and Section 6.3] The Texture session introduced new analytical attributes not present in the baseline condition. Section 4.1 states that the research team 'derived words from the text attributes with the spans if words were discussed in the baseline interview, and added document embeddings for the text using the OpenAI text-embedding-3-small model if participants did not already provide them.' The headline discoveries in Section 6.3 (P9's duplicate Reddit posts and P10's near-duplicate training prompts) are both based on the embedding projection, which was not part of the baseline workflow. The comparison therefore conflates the interface design with the addition of embedding-based overview and search capabilities. While Texture provides the interaction scaffolding, the claim that 'TEXTURE enabled participants to discover new insights' cannot be cleanly attributed to the system's interaction design rather than to the new embedding attributes that were precomputed and inserted into the data before the session.
  3. [Section 6.2 and Figure 7] The 'more quickly iterate' claim is not directly measured. The survey items in Appendix A ask participants whether tasks were 'easier' on a 5-point Likert scale, not whether they were faster; no objective timing data, interaction logs, or behavioral measures are reported. The paper's abstract states that 'TEXTURE enabled participants to more quickly iterate,' but the only quantitative evidence is self-reported comparative ease (Figure 7), and the supporting quotes such as 'This just gets me there so much faster' are anecdotal. Even if the order and attribute confounds were resolved, the current data would not support a quantitative speed claim. The claims should be tempered to self-reported ease and perceived speed, with the absence of objective measures noted in the abstract and conclusion.
minor comments (5)
  1. [Throughout] There are many missing spaces after the product name (e.g., 'TEXTUREhelps', 'TEXTUREis', 'TEXTUREdata') in the figure captions, abstract, and body text; these should be fixed in proofreading.
  2. [Section 2.2.3] The heading 'Corpus Overview Methods and Embedddings' contains a typo: 'Embedddings' should be 'Embeddings.'
  3. [Section 3.4] The substring-highlighting example is confusing because both sentences in the sentence 'This way “we won the wonderful match” is highlighted and not “we won the wonderful match”' are identical. The second example presumably should contain a different word (e.g., 'wonderful') to illustrate that the substring 'won' inside 'wonderful' is not highlighted.
  4. [Reference [14]] The journal name and volume in the Evans and Aceves reference are typeset with an unusual space ('V olume 42'), which should be corrected.
  5. [Figure 7] The note for 'Calculate & Verify New*' says the item reflects the seven participants who provided ratings; it would be helpful to state briefly why three participants did not provide ratings so readers can judge the coverage.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the central claims are empirical system and usability findings, and the only self-cited dependency (Mosaic) is implemented as prior infrastructure rather than as a definitional input.

full rationale

The paper does not present a mathematical derivation or a fitted prediction; it presents a configurable schema, an implemented tool, and a two-session user study. The central effectiveness claims rest on observed participant behavior, self-reported ratings, and qualitative insights, not on equations whose outputs duplicate their inputs. The one overlapping-author dependency, Mosaic [22], is used as an implementation substrate for coordinated queries and is cited as prior published work; no uniqueness claim or fitted parameter is imported from that citation. The expressiveness claim that all participant attributes fit the TEXTURE schema is partly a consequence of the study procedure, since the research team formatted each dataset into the schema before the Texture session (Section 4.1), but this is a methodological limitation of a case study rather than a circular derivation: the schema was defined before the datasets were encoded, and the study demonstrates the encoding on ten external datasets. Section 6.4 explicitly concedes that repeated exposure to the same dataset could explain additional insights; that is a validity threat to the comparative effectiveness claim, not a circularity in the argument. Mosaic and other prior systems are cited contextually and do not carry the load of the effectiveness conclusion. Overall, no step reduces by construction to its input, and no load-bearing self-citation chain is present.

Assumptions & free parameters 0 free parameters · 3 assumptions · 1 invented entities

The system itself introduces no fitted parameters. Its evaluation rests on the expressiveness of the schema, which is shown directly with 10 datasets, and on a user study whose interpretation depends on the stated domain assumptions.

assumptions (3)
  • domain assumption First normal form normalization into relational tables enables scalable interactive filtering.
    Section 3.1.1; relies on standard database practice and Mosaic's query engine, not empirically validated in this paper.
  • domain assumption Document embeddings and cosine similarity meaningfully capture document similarity for overview and search.
    Sections 3.2 and 3.3; adopted from prior work without independent validation in this study.
  • ad hoc to paper Self-reported ratings and thematic analysis of 10 participants provide valid evidence for the system's effectiveness.
    Section 4 and Section 6; no objective measures or control condition are used.
invented entities (1)
  • Texture five-type attribute schema (text, single-value, list, span list, embedding) independent evidence
    purpose: Represents descriptive text attributes at arbitrary granularity and cardinality for exploration.
    The schema is implemented in the open-source Texture tool and shown to represent all attributes in 10 participant datasets, giving it direct evidence outside itself.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Texture: Structured Exploration of Text Datasets." pith.science (2026). https://pith.science/paper/YGSLISLP

@misc{pith2026250416898,
  author       = {Pith},
  title        = {Pith review of: Texture: Structured Exploration of Text Datasets},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/YGSLISLP}},
  note         = {Machine review of arXiv:2504.16898}
}
read the original abstract

Exploratory analysis of a text corpus is essential for assessing data quality and developing meaningful hypotheses. Text analysis relies on understanding documents through structured attributes spanning various granularities of the documents such as words, phrases, sentences, topics, or clusters. However, current text visualization tools typically adopt a fixed representation tailored to specific tasks or domains, requiring users to switch tools as their analytical goals change. To address this limitation, we present Texture, a general-purpose interactive text exploration tool. Texture introduces a configurable data schema for representing text documents enriched with descriptive attributes. These attributes can appear at arbitrary levels of granularity in the text and possibly have multiple values, including document-level attributes, multi-valued attributes (e.g., topics), fine-grained span-level attributes (e.g., words), and vector embeddings. The system then combines existing interactive methods for text exploration into a single interface that provides attribute overview visualizations, supports cross-filtering attribute charts to explore subsets, uses embeddings for a dataset overview and similar instance search, and contextualizes filters in the actual documents. We evaluated Texture through a two-part user study with 10 participants from varied domains who each analyzed their own dataset in a baseline session and then with Texture. Texture was able to represent all of the previously derived dataset attributes, enabled participants to more quickly iterate during their exploratory analysis, and discover new insights about their data. Our findings contribute to the design of scalable, interactive, and flexible exploration systems that improve users' ability to make sense of text data.

Figures

Figures reproduced from arXiv: 2504.16898 by the authors.

Figure 1
Figure 1. TEXTURE helps users explore text datasets through structured descriptive attributes. Its configurable data schema supports attributes at any level of granularity in the text, such as document-level attributes like the conference and embedding or word-level counts shown in this example analysis of an abstract corpus. The system organizes list attributes like words with multiple values per document into new tables and… view at source ↗
Figure 3
Figure 3. All attributes are automatically visualized according to their data [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 2
Figure 2. Following the TEXTURE data schema requires placing list at￾tributes into new tables that map back to the documents. 3.2 Overview: Automatic Attribute Visualizations Providing a dataset and schema to TEXTURE enables immediate explo￾ration in the UI [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Document embeddings enable a projection overview and similarity [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 5
Figure 5. Figure 5: TEXTURE helps users contextualize attribute filters in the actual documents by showing documents that match current filters and highlighting the spans of text for filtered span list attributes. 1. How well can users working across different tasks represent their text d…
Figure 6
Figure 6. Figure 6: Participants used TEXTURE to explore a wide variety of datasets including LLM outputs, song lyrics, and Reddit posts. (also used by 7/10 participants in TEXTURE), along with filters over multiple attribute charts. The automatic attribute visualizations in TEXTURE sped …
Figure 7
Figure 7. Figure 7: Participants rated if Texture made it easier to perform certain [PITH_FULL_IMAGE:figures/full_fig_p008_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

88 extracted references · 57 canonical work pages

  1. [1]

    Alexander, J

    E. Alexander, J. Kohlmann, R. Valenza, M. Witmore, and M. Gleicher. Serendip: Topic model-driven visual exploration of text corpora. In2014 IEEE Conference on Visual Analytics Science and Technology (VAST), pp. 173–182. IEEE, 2014. 3, 5

    2014

  2. [2]

    Alsakran, Y

    J. Alsakran, Y . Chen, Y . Zhao, J. Yang, and D. Luo. Streamit: Dy- namic visualization and interactive exploration of text streams. In2011 IEEE Pacific Visualization Symposium, pp. 131–138, 2011. doi: 10.1109/ PACIFICVIS.2011.5742382 2

    arXiv 2011

  3. [3]

    H. An, A. Narechania, E. Wall, and K. Xu. vitality 2: Reviewing academic literature using large language models. Presented at the NLVIZ Workshop, IEEE VIS 2024, 2024. 2

    2024

  4. [4]

    D. M. Blei, A. Y . Ng, and M. I. Jordan. Latent dirichlet allocation.Journal of machine Learning research, 3(Jan):993–1022, 2003. 3

    2003

  5. [5]

    Bosch, D

    H. Bosch, D. Thom, F. Heimerl, E. Püttmann, S. Koch, R. Krüger, M. Wörner, and T. Ertl. Scatterblogs2: Real-time monitoring of mi- croblog messages through user-guided filtering.IEEE Transactions on Visualization and Computer Graphics, 19(12):2022–2031, 2013. 2

    2022

  6. [6]

    A. A. Cabrera, E. Fu, D. Bertucci, K. Holstein, A. Talwalkar, J. I. Hong, and A. Perer. Zeno: An interactive framework for behavioral evaluation of machine learning. InProceedings of the 2023 CHI Conference on Human Factors in Computing Systems, CHI ’23, article no. 419, 14 pages. Association for Computing Machinery, New York, NY , USA, 2023. doi: 10.1145...

    arXiv 2023

  7. [7]

    E. F. Codd. A relational model of data for large shared data banks.Com- mun. ACM, 13(6):377–387, 11 pages, June 1970. doi: 10.1145/362384. 362685 4

    work page doi:10.1145/362384 1970

  8. [8]

    Collins, F

    C. Collins, F. B. Viegas, and M. Wattenberg. Parallel tag clouds to explore and analyze faceted text corpora. In2009 IEEE Symposium on Visual Analytics Science and Technology, pp. 91–98, 2009. doi: 10.1109/V AST. 2009.5333443 2

    arXiv 2009

Show all 88 references
  1. [9]

    Correll, M

    M. Correll, M. Witmore, and M. Gleicher. Exploring collections of tagged text for literary scholarship.Computer Graphics Forum, 30(3):731–740,

  2. [10]

    Devlin, M

    J. Devlin, M. Chang, K. Lee, and K. Toutanova. BERT: pre-training of deep bidirectional transformers for language understanding. In J. Burstein, C. Doran, and T. Solorio, eds.,Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational...

    2019 doi

  3. [11]

    A. Don, E. Zheleva, M. Gregory, S. Tarkan, L. Auvil, T. Clement, B. Shnei- derman, and C. Plaisant. Discovering interesting usage patterns in text collections: integrating text mining with visualization. InProceedings of the sixteenth ACM conference on Conference on informatio...

    2007

  4. [12]

    Elazar, A

    Y . Elazar, A. Bhagia, I. Magnusson, A. Ravichander, D. Schwenk, A. Suhr, P. Walsh, D. Groeneveld, L. Soldaini, S. Singh, et al. What’s in my big data?arXiv preprint arXiv:2310.20707, 2023. 2

    2023 arXiv

  5. [13]

    Epperson, V

    W. Epperson, V . Gorantla, D. Moritz, and A. Perer. Dead or alive: Con- tinuous data profiling for interactive data science.IEEE Transactions on Visualization and Computer Graphics, 30(1):197–207, 2024. doi: 10. 1109/TVCG.2023.3327367 2, 4

    2024

  6. [14]

    J. A. Evans and P. Aceves. Machine translation: Mining text for social theory.Annual Review of Sociology, 42(V olume 42, 2016):21–50, 2016. doi: 10.1146/annurev-soc-081715-074206 1

    2016 doi

  7. [15]

    E. Fast, B. Chen, and M. S. Bernstein. Empath: Understanding topic signals in large-scale text. InProceedings of the 2016 CHI conference on human factors in computing systems, pp. 4647–4657, 2016. 2, 3

    2016

  8. [16]

    Felix, S

    C. Felix, S. Franconeri, and E. Bertini. Taking word clouds apart: An empirical investigation of the design space for keyword summaries.IEEE Transactions on Visualization and Computer Graphics, 24(1):657–666,

  9. [17]

    Felix, A

    C. Felix, A. V . Pandey, and E. Bertini. Texttile: An interactive visualization tool for seamless exploratory analysis of structured data and unstructured text.IEEE Transactions on Visualization and Computer Graphics, 23:161– 170, 2017. 3

    2017

  10. [18]

    Gentzkow, B

    M. Gentzkow, B. Kelly, and M. Taddy. Text as data.Journal of Economic Literature, 57(3):535–574, 2019. 1

    2019

  11. [19]

    Github copilot - your ai pair programmer

    Github. Github copilot - your ai pair programmer. https://github. com/features/copilot. Accessed 03-2025. 9

    2025

  12. [20]

    Griggs, C

    P. Griggs, C. Demiralp, and S. Rahman. Towards integrated, interactive, and extensible text data analytics with leam. In E. Dragut, Y . Li, L. Popa, and S. Vucetic, eds.,Proceedings of the Second Workshop on Data Science with Human in the Loop: Language Advances, pp. 52–58. As...

    2021 doi

  13. [21]

    Gururaja, N

    S. Gururaja, N. Gandhi, J. Milbauer, and E. Strubell. Beyond text: Expert needs in document research.ACL, 2025. 2

    2025

  14. [22]

    Heer and D

    J. Heer and D. Moritz. Mosaic: An architecture for scalable & interop- erable data views.IEEE Transactions on Visualization and Computer Graphics, 30(1):436–446, 2024. doi: 10.1109/TVCG.2023.3327189 4, 5, 12

    2024

  15. [23]

    Heimerl, M

    F. Heimerl, M. John, Q. Han, S. Koch, and T. Ertl. Docucompass: Effective exploration of document landscapes. In2016 IEEE conference on visual analytics science and technology (VAST), pp. 11–20. IEEE, 2016. 3, 5

    2016

  16. [24]

    E. Horvitz. Principles of mixed-initiative user interfaces. InProceedings of the SIGCHI conference on Human Factors in Computing Systems, pp. 159–166, 1999. 9

    1999

  17. [25]

    Huggingface datasets

    Huggingface. Huggingface datasets. https://huggingface.co/ datasets, 2024. Accessed 03-2025. 6

    2024

  18. [26]

    Ittoo, A

    A. Ittoo, A. van den Bosch, et al. Text analytics in industry: Challenges, desiderata and trends.Computers in Industry, 78:96–107, 2016. 2

    2016

  19. [27]

    Kahng, I

    M. Kahng, I. Tenney, M. Pushkarna, M. X. Liu, J. Wexler, E. Reif, K. Kallarackal, M. Chang, M. Terry, and L. Dixon. Llm comparator: Interactive analysis of side-by-side evaluation of large language models. IEEE Transactions on Visualization and Computer Graphics, 2024. 1, 2, 3, 4

    2024

  20. [28]

    Kandel, R

    S. Kandel, R. Parikh, A. Paepcke, J. M. Hellerstein, and J. Heer. Profiler: Integrated statistical analysis and visualization for data quality assessment. InProceedings of the International Working Conference on Advanced Vi- sual Interfaces, A VI ’12, 8 pages, p. 547–554. Asso...

    2012

  21. [29]

    M. B. Kery, D. Ren, F. Hohman, D. Moritz, K. Wongsuphasawat, and K. Patel. Mage: Fluid moves between code and graphical work in compu- tational notebooks. InProceedings of the 33rd Annual ACM Symposium on User Interface Software and Technology, UIST ’20, 12 pages, p. 140–151. ...

    2020

  22. [30]

    M. S. Lam, J. Teoh, J. A. Landay, J. Heer, and M. S. Bernstein. Concept induction: Analyzing unstructured text with high-level concepts using lloom. InProceedings of the CHI Conference on Human Factors in Computing Systems, CHI ’24, article no. 766, 28 pages. Association for C...

    2024 doi

  23. [31]

    Lance db

    Lance DB. Lance db. https://lancedb.com/, 2024. Accessed 03-2025. 5

    2024

  24. [32]

    D. Lange. Vispubs: A visualization publications repository, 2024. doi: 10. 31219/osf.io/dg3p2 4

    2024

  25. [33]

    B. Lee, M. Czerwinski, G. Robertson, and B. B. Bederson. Understanding research trends in conferences using paperlens. InCHI’05 extended abstracts on Human factors in computing systems, pp. 1969–1972, 2005. 3

    1969

  26. [34]

    S. Liu, X. Wang, C. Collins, W. Dou, F. Ouyang, M. El-Assady, L. Jiang, and D. A. Keim. Bridging text visualization and mining: A task-driven survey.IEEE transactions on visualization and computer graphics, 25(7):2482–2504, 2018. 2

    2018

  27. [35]

    S. Liu, M. X. Zhou, S. Pan, Y . Song, W. Qian, W. Cai, and X. Lian. Tiara: Interactive, topic-based visual text summarization and analysis.ACM Transactions on Intelligent Systems and Technology (TIST), 3(2):1–28,

  28. [36]

    S. Loria. textblob documentation.Release 0.15, 2, 2018. 7

    2018

  29. [37]

    McInnes, J

    L. McInnes, J. Healy, N. Saul, and L. Grossberger. Umap: Uniform manifold approximation and projection.The Journal of Open Source Software, 3(29):861, 2018. 3, 4

    2018

  30. [38]

    Muralidharan and M

    A. Muralidharan and M. A. Hearst. Supporting exploratory text analysis in literature study.Literary and linguistic computing, 28(2):283–295, 2012. 2

    2012

  31. [39]

    Narechania, A

    A. Narechania, A. Karduni, R. Wesslen, and E. Wall. Vitality: Promoting serendipitous discovery of academic literature with transformers & visual analytics.IEEE Transactions on Visualization and Computer Graphics, 28(1):486–496, 2022. doi: 10.1109/TVCG.2021.3114820 2, 3, 4

    2022

  32. [40]

    Nomic atlas

    Nomic AI. Nomic atlas. https://github.com/nomic-ai/nomic,

  33. [41]

    Oelke, D

    D. Oelke, D. Spretke, A. Stoffel, and D. A. Keim. Visual readability analysis: How to make your writings easier to read.IEEE Transactions on Visualization and Computer Graphics, 18(5):662–674, 2011. 2

    2011

  34. [42]

    Openai vector embeddings, 2025

    OpenAI. Openai vector embeddings, 2025. Accessed 03-25. 6

    2025

  35. [43]

    Pandas: Python data analysis library

    Pandas. Pandas: Python data analysis library. https://pandas.pydata. org, 2024. Accessed 03-2025. 4

    2024

  36. [44]

    Pennington, R

    J. Pennington, R. Socher, and C. D. Manning. Glove: Global vectors for word representation. InEmpirical Methods in Natural Language Processing (EMNLP), pp. 1532–1543, 2014. 3

    2014

  37. [45]

    E. Reif, M. Kahng, and S. Petridis. Visualizing linguistic diversity of text datasets synthesized by large language models. In2023 IEEE Visualization and Visual Analytics (VIS), Melbourne, Australia, October 21-27, 2023, pp. 236–240. IEEE, New York, NY , USA, 2023. doi: 10.110...

    2023

  38. [46]

    E. Reif, C. Qian, J. Wexler, and M. Kahng. Automatic histograms: Lever- aging language models for text dataset exploration. In F. F. Mueller, P. Kyburz, J. R. Williamson, and C. Sas, eds.,Extended Abstracts of the CHI Conference on Human Factors in Computing Systems, CHI EA 20...

    2024

  39. [47]

    Reimers and I

    N. Reimers and I. Gurevych. Sentence-bert: Sentence embeddings us- ing siamese bert-networks. In K. Inui, J. Jiang, V . Ng, and X. Wan, eds., Proceedings of the 2019 Conference on Empirical Methods in Natural Lan- guage Processing and the 9th International Joint Conference on ...

    2019 doi

  40. [48]

    Robertson, Z

    S. Robertson, Z. J. Wang, D. Moritz, M. B. Kery, and F. Hohman. Angler: Helping machine translation practitioners prioritize model improvements. InProceedings of the 2023 CHI Conference on Human Factors in Comput- ing Systems, CHI ’23, article no. 832, 20 pages. Association fo...

    2023

  41. [49]

    everyone wants to do the model work, not the data work

    N. Sambasivan, S. Kapania, H. Highfill, D. Akrong, P. Paritosh, and L. M. Aroyo. “everyone wants to do the model work, not the data work”: Data cascades in high-stakes ai. InProceedings of the 2021 CHI Conference on Human Factors in Computing Systems, CHI ’21, article no. 39, ...

    2021

  42. [50]

    Sievert and K

    C. Sievert and K. Shirley. Ldavis: A method for visualizing and inter- preting topics. InProceedings of the workshop on interactive language learning, visualization, and interfaces, pp. 63–70, 2014. 2

    2014

  43. [51]

    G. Staff. Octoverse: Ai leads python to top language as the number of global developers surges, 2024. Published 10-29-24. 5

    2024

  44. [52]

    Stasko, C

    J. Stasko, C. Gorg, Z. Liu, and K. Singhal. Jigsaw: supporting investigative analysis through interactive visualization. In2007 IEEE Symposium on Visual Analytics Science and Technology, pp. 131–138. IEEE, 2007. 3

    2007

  45. [53]

    Strobelt, D

    H. Strobelt, D. Oelke, B. C. Kwon, T. Schreck, and H. Pfister. Guidelines for effective usage of text highlighting techniques.IEEE transactions on visualization and computer graphics, 22(1):489–498, 2015. 2

    2015

  46. [54]

    Swayamdipta, R

    S. Swayamdipta, R. Schwartz, N. Lourie, Y . Wang, H. Hajishirzi, N. A. Smith, and Y . Choi. Dataset cartography: Mapping and diagnosing datasets with training dynamics. InProceedings of the 2020 Conference on Empir- ical Methods in Natural Language Processing (EMNLP), pp. 9275...

    2020

  47. [55]

    J. W. Tukey.Exploratory data analysis / John W. Tukey.Addison-Wesley series in behavioral science. Addison-Wesley Pub. Co., Reading, Mass,

  48. [56]

    J. W. Tukey. We need both exploratory and confirmatory.The American Statistician, 34:23–25, 1980. doi: 10.2307/2682991 2, 9

    1980 doi

  49. [57]

    F. B. Viégas and M. Wattenberg. Timelines tag clouds and the case for vernacular visualization.interactions, 15(4):49–52, 2008. 2

    2008

  50. [58]

    Z. J. Wang, F. Hohman, and D. H. Chau. WizMap: Scalable Interactive Visualization for Exploring Large Machine Learning Embeddings.arXiv 2306.09328, 2023. 3, 5

    2023 arXiv

  51. [59]

    Wanner, A

    F. Wanner, A. Stoffel, D. Jäckle, B. C. Kwon, A. Weiler, D. A. Keim, K. E. Isaacs, A. Giménez, I. Jusufi, T. Gamblin, et al. State-of-the-art report of visual analysis for event detection in text data streams. InEuroVis (STARs),

  52. [60]

    Wongsuphasawat, Y

    K. Wongsuphasawat, Y . Liu, and J. Heer. Goals, process, and challenges of exploratory data analysis: An interview study, 2019. 2

    2019

  53. [61]

    Wongsuphasawat, D

    K. Wongsuphasawat, D. Moritz, A. Anand, J. Mackinlay, B. Howe, and J. Heer. V oyager: Exploratory analysis via faceted browsing of visual- ization recommendations.IEEE Trans. Visualization & Comp. Graphics (Proc. InfoVis), 2016. doi: 10.1109/TVCG.2015.2467191 2

    2016

  54. [62]

    T. Wu, K. Wongsuphasawat, D. Ren, K. Patel, and C. DuBois. Tempura: Query analysis with structural templates. InProceedings of the 2020 CHI Conference on Human Factors in Computing Systems, pp. 1–12, 2020. 3

    2020

  55. [63]

    Ziegler, E

    A. Ziegler, E. Kalliamvakou, X. A. Li, A. Rice, D. Rifkin, S. Simister, G. Sittampalam, and E. Aftandilian. Productivity assessment of neural code completion. InProceedings of the 6th ACM SIGPLAN International Symposium on Machine Programming, MAPS 2022, 9 pages, p. 21–29. Ass...

    2022

  56. [66]

    How would your rate your experience with Python?

  57. [67]

    How many years have you been using Python?

  58. [68]

    How often do you work with text data?

  59. [69]

    When was the last time you worked with this dataset in particular?

  60. [70]

    What are the analysis questions you were trying to answer with this data? A.2 Session 1: baseline exploration questions Discussion questions:

  61. [71]

    How did you understand what is in the dataset? Can you show us which code or visualizations you used?

  62. [72]

    Were all these attributes already present in the data? If not, how did you derive them?

  63. [73]

    How did you verify or explore the results after adding a new attribute?

  64. [74]

    What kind of filtering or subset analysis did you do of this dataset?

  65. [75]

    Do you look at similar instances or use embeddings as part of your exploration / how?

  66. [76]

    What was the hardest part about this task and working with this text data for you?

  67. [77]

    If you had a tool to help you understand the text data, what do you wish it could do? A.3 Session 2: TEXTUREusage questions We asked the following Likert-scale questions in a survey immediately after the task:

  68. [78]

    I found it easier to understand the text attributes in my dataset with Texture than the last session (5 point Likert scale)

  69. [79]

    I found it easier to filter to subsets of my dataset with Texture than the last session (5 point Likert scale)

  70. [80]

    I found it easier to calculate a new attribute and verify its correct- ness with Texture than the last session (5 point Likert scale)

  71. [81]

    I found it easier to find similar instances in my dataset with Texture than the last session (5 point Likert scale)

  72. [82]

    I learned new things about my dataset while exploring it with Texture (5 point Likert scale) Discussion questions:

  73. [83]

    Talk me through your rating on how TEXTUREimpacted your understanding of your text attributes

  74. [84]

    Talk me through your rating on how TEXTUREimpacted your ability to filter to subsets

  75. [85]

    Talk me through your rating on how TEXTUREimpacted your ability to add a new attribute and verify

  76. [86]

    Talk me through your rating on how TEXTUREimpacted your ability to make sense of similar instances

  77. [87]

    Talk me through your rating on how you learned new things about your dataset with TEXTURE

  78. [88]

    Therefore we must be able to filter across joined tables to enable interactive cross-filtering

    What other features would better support your workflow? B TECHNICALDETAILS: FILTERSACROSSMULTIPLETABLES In TEXTURE, attributes may come from different tables. Therefore we must be able to filter across joined tables to enable interactive cross-filtering. In Mosaic, filters are...

  79. [2011]

    doi: 10.1111/j.1467-8659.2011.01922.x 2

    2011

  80. [2018]

    doi: 10.1109/TVCG.2017.2746018 2

    2017

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.