Wrex: A Unified Programming-by-Example Interaction for Synthesizing Readable Code for Data Scientists

Drosos, Ian; Barik, Titus; Guo, Philip J.; DeLine, Robert; Gulwani, Sumit

doi:10.1145/3313831.3376442

Cited by 95 publications

(70 citation statements)

References 27 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Instead of eliciting additional examples from users, several interactive approaches have been proposed to help users directly inspect and refine synthesized programs [42,10,30,64]. FlashProg [42] paraphrases synthesized programs in English and clusters programs with common structures for ease of navigation.…”

Section: Direct Program Manipulation To Guide Synthesismentioning

confidence: 99%

“…FlashProg [42] paraphrases synthesized programs in English and clusters programs with common structures for ease of navigation. Wrex [10] transforms synthesized programs to readable code and allows users to directly edit them.…”

Section: Direct Program Manipulation To Guide Synthesismentioning

confidence: 99%

“…Even when they do, inexperienced users are not good at providing representative examples that cover the hypothetical input space [34,44]. Several interactive approaches have been proposed to enable users to directly refine synthesized programs rather than crafting additional examples [42,10,30,46]. Unfortunately, these approaches assume users are capable of and willing to inspect and edit programs.…”

Section: Introductionmentioning

confidence: 99%

“…This fundamental assumption may not hold in many domains, especially for novice programmers and end-users without enough expertise in programming. Some approaches have explored the middle ground by rendering synthesized code in a humanfriendly format [42,10], but users still need to spend substantial time comprehending the program. This is cognitively demanding, especially when multiple candidate programs are synthesized and provided to users.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Interactive Program Synthesis by Augmented Examples

Zhang

Lowmanstone

Wang

et al. 2020

Proceedings of the 33rd Annual ACM Symposium on User Interface Software and Technology

View full text Add to dashboard Cite

Mark parts of an input as literal or general 2 Annotate parts of synthesized programs to be desired or undesired 3 View other similar examples and corner cases 4 Figure 1: A user interactively synthesizes a regular expression (regex) that accepts phone numbers starting with an optional '+'. After 1) adding some positive and negative examples, the user 2) marks the + sign as literal and the following numbers as a general class of digits to specify how individual characters should be treated by the synthesizer. To refine the synthesis result, the user can either 3) directly mark subexpressions as desired or undesired in the final regex, or 4) ask for additional examples to validate and enhance her understanding of a regex candidate as well as identify and label counterexamples, therefore adding them to the set of user-provided input-output examples.

show abstract

Section: Direct Program Manipulation To Guide Synthesismentioning

confidence: 99%

Section: Direct Program Manipulation To Guide Synthesismentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Interactive Program Synthesis by Augmented Examples

Zhang

Lowmanstone

Wang

et al. 2020

Proceedings of the 33rd Annual ACM Symposium on User Interface Software and Technology

View full text Add to dashboard Cite

show abstract

“…Meeting data scientists where they work is a motivation that B2 shares with Wrex [20], a recent system that embeds a Session 1B: Bridging Programming and Visualization UIST '20, October 20-23, 2020, Virtual Event, USA visual data wrangling interface within the Jupyter notebook. Building on the previous theme, a key insight of Wrex is that it is not enough to simply embed PBE systems in context; rather, to respect literate computing principles, the code these systems synthesize must be human-readable.…”

Section: Interactions Generating Codementioning

confidence: 99%

B2

Hellerstein

Satyanarayan

2020

Proceedings of the 33rd Annual ACM Symposium on User Interface Software and Technology

View full text Add to dashboard Cite

Data scientists have embraced computational notebooks to author analysis code and accompanying visualizations within a single document. Currently, although these media may be interleaved, they remain siloed: interactive visualizations must be manually specified as they are divorced from the analysis provenance expressed via dataframes, while code cells have no access to users' interactions with visualizations, and hence no way to operate on the results of interaction. To bridge this divide, we present B2, a set of techniques grounded in treating data queries as a shared representation between the code and interactive visualizations. B2 instruments data frames to track the queries expressed in code and synthesize corresponding visualizations. These visualizations are displayed in a dashboard to facilitate interactive analysis. When an interaction occurs, B2 reifies it as a data query and generates a history log in a new code cell. Subsequent cells can use this log to further analyze interaction results and, when marked as reactive, to ensure that code is automatically recomputed when new interaction occurs. In an evaluative study with data scientists, we find that B2 promotes a tighter feedback loop between coding and interacting with visualizations. All participants frequently moved from code to visualization and vice-versa, which facilitated their exploratory data analysis in the notebook.

show abstract

avatar—Automated Feature Wrangling for Machine Learning

Verbruggen

Wolputte

Dumančić

et al. 2021

Advances in Intelligent Data Analysis XIX

View full text Add to dashboard Cite

A large part of the time invested in data science is spent on manual preparation of data. Transforming wrongly formatted columns into useful features takes up a significant part of this time. We present the avatar algorithm for automatically learning programs that perform this type of feature wrangling. Instead of relying on users to guide the wrangling process, avatar directly uses the predictive performance of machine learning models to measure its progress during wrangling. We use datasets from Kaggle to show that avatar improves raw data for prediction, and square it off against human data scientists.

show abstract

Wrex: A Unified Programming-by-Example Interaction for Synthesizing Readable Code for Data Scientists

Cited by 95 publications

References 27 publications

Interactive Program Synthesis by Augmented Examples

Interactive Program Synthesis by Augmented Examples

B2

avatar—Automated Feature Wrangling for Machine Learning

Contact Info

Product

Resources

About