Blog

New infrastructure

Wordbank has moved to a new infrastructure! The site you are reading is now fully static: the database lives on Redivis as a versioned, freely downloadable dataset, and all visualizations run directly in your browser, computed from the raw data. This datapages-style rebuild replaces the previous Django and Shiny stack, making the site faster and much cheaper to maintain, and making every data release citable by version. The migration was carried out with the help of Claude Code and the Claude Fable model, working alongside the Wordbank team. Old links and the wordbankr R package continue to work, and the underlying data are unchanged.

New languages!

 We're pleased to announce a new release of the Wordbank database, bringing us to more than 100,000 CDI administrations! This release includes data from Japanese, Estonian, Saudi Arabic, and Catalan, as well as new data from Korean, Finnish, French, and French-English bilinguals. 

A meta-analysis of outcomes for late talkers

The extreme variability in outcome of late talkers raises challenges both for theoretical work on the sources and development of individual differences in early language development, and for clinical/educational practice. Research on long-term outcomes is sparse but growing, though primarily focused on English. There is need for increasing the research database, and especially for integration of research across languages. We are undertaking a cross-linguistic meta-analysis of outcomes for late talkers. In addition to published and unpublished papers specifically focused on this issue, we believe there are projects which may have another primary focus, but which include an early (18-36 month) assessment of expressive vocabulary, and language and/or literacy outcomes at age 5 years or beyond. The target studies can include broadly representative samples, or late talkers specifically. If you are interested in the possibility of collaboration on this project, please contact either of us listed below. You may also find the PROSPERO registration of this project helpful in understanding our plans:

https://www.crd.york.ac.uk/PROSPERO, registration CRD42023394687.

Emma Hayiou-Thomas, emma.hayiou-thomas@york.ac.uk

Philip Dale, dalep@unm.edu

Educational materials

Mike just taught a course at the University of Amsterdam LOT winter school on "Language Learning: A Data-Driven Approach." Here's the course description:

In this course, we will examine early language learning through the lens of new data resources that facilitate quantitative studies. Our framework will be the "Standard Model" of Kachergis, Marchman, and Frank (2022) that links language input to processing and learning outcomes, and we will consider the strengths and weaknesses of this model for describing vocabulary learning as well as the learning of some morphology and syntax. Our hands-on approach will involve learning the use of CHILDES and childes-db for studying language input, Wordbank for studying language outcomes, and Peekbank for studying processing.

Materials from Days 1 and 2 focus on Wordbank and reproduce several patterns from the Wordbank Book! All of the code and slides are available here: https://github.com/mcfrank/lot-language-learning-2023. 

Wordbank updates!

 

We are very pleased to announce a major update to Wordbank, including some significant changes to the database structure. We are also adding more than 10 new languages and data from many thousands of children. 

We now include data from multilingual children and children with diagnosed developmental disorders (as well as functionality for identifying these children via the shiny apps and the wordbankr API). 

The prior version of the Wordbank database will remain up and available for queries via the wordbankr 0.3.0 API (or earlier) for a period of at least 6 months, but if you upgrade wordbankr you will begin accessing the new Wordbank data. 

Wordbank data will also be versioned going forward so that older snapshots of the database will be available via S3 snapshots (see Documentation page). 

Wordbank Book!

We are very pleased to announce publication of the Wordbank Book, "Variability and Consistency in Early Language Learning," now available from MIT Press (2021). The book brings together many different ways of looking at data in the Wordbank database, in service of characterizing how children vary as well as shared patterns of learning. The book is also available free online at http://wordbank-book.stanford.edu, and all of the code necessary to generate it from the Wordbank data can be found at http://github.com/langcog/wordbank-book.

More languages (and some naming changes)

We are pleased to announce several pieces of news:

First, the arrival of several more languages and datasets, including French (European), more Korean data, as well as more Hebrew data and Spanish (European) in the works.

Second, we have a new licensing standard such that some datasets can be licensed Creative Commons for Non-Commercial use. These datasets are marked on the contributors page.

Finally, because of the new data, we have some new naming conventions for languages. "English" is now "English (American)"; this convention will generally be followed as "Language (Country/Region)." These are breaking changes unfortunately, we apologize for the inconvenience and are working on past database images available for purposes of reproducibility.

New frontpage and languages

Since our last update, we've reorganized the frontpage and revamped the contributors and citation policy. We've also added Latvian, Slovak, and Korean data!

New languages

We're pleased to announce a number of new languages in Wordbank, bringing us up above 20 languages. The new data are from ASL, Cypriot Greek, Kiswahili and Kirigama. We also have added British English short-form data from the Twins Early Development Study (TEDS). Take a look at these new additions!

New changes

We now have a FAQ!

Also, check out the semantic networks report and the new scoring tool.

Stay tuned for some new languages in the next few months.

Paper in press

We are pleased to announce that our paper on the wordbank site is in press at Journal of Child Language:

Frank, M. C., Braginsky, M., Yurovsky, D., Marchman, V. A. (in press). Wordbank: An open repository for developmental vocabulary data. Journal of Child Language. 

wordbankr on CRAN

Our R package for accessing wordbank is now on CRAN! You can now install and use wordbankr even more easily.

Licensing

We're pleased to announce that all derivatives from Wordbank – including downloaded data, tables, and graphs – are licensed CC-BY 4.0. To attribute derivatives, please cite:

Frank, M. C., Braginsky, M., Yurovsky, D., Marchman, V. A. (under revision). Wordbank: An open repository for developmental vocabulary data. Journal of Child Language.

Going forward, we will update the preferred citation for the site on the Publications page.

Paper under review

We are happy to announce that we have a new paper on Wordbank and the wordbankr package now under review. For now, this is the appropriate citation for the wordbank site:

Frank, M. C., Braginsky, M., Yurovsky, D., Marchman, V. A. (under review). Wordbank: An open repository for developmental vocabulary data.

wordbankr package

We've been working on improving how you can access the Wordbank data from R, and the result is the wordbankr package. It's still a work in progress, so a lot of it may change, but using it you can now pull in data by-administration, by-item, and administration-by-item, with minimal hassle. Check out the package vignette for a tutorial!

Reorganized reports

We've just pushed an update to our analysis pages so that they include much more content. We not only have interactive visualizations (like the norms and items tools), we also have a number of non-interactive reports that we've put together on topics like gender differences, SES effects, etc. You can also see analyses from our CogSci paper this year on grammar-lexicon relationships. Let us know if you have comments (or analyses you'd like to add)!

CLEX data

We're excited to announce we've imported all of the CDI norming data from CLEX. Thank you to Rune Nørgaard Jørgensen! Wordbank now has almost 40,000 CDI administrations, across 10 different languages (Croatian, Danish, English, German, Italian, Norwegian, Russian, Spanish, Swedish, and Turkish).

Data downloading

Check out our new reports that let you explore our data, filter and sort it to suit your needs, and download it as a csv file. There's one report for by-administration data, with overall vocabulary sizes and demographic information, and one report for the full administration-by-item data. Also, all of the analysis reports now let you download the plot and the data being plotted!

Cross-linguistic data

We've spent the last few months overhauling Wordbank's data import to support adding in many datasets in different languages. Yesterday these updates went live, and Wordbank now has data in English, Spanish, Danish, and Norwegian, including both Words & Sentences and Words & Gestures. It's a total of almost 30,000 CDI administrations! Check out our reports to play around with all of these datasets, or look at this tutorial on accessing Wordbank from R if you're interested in running your own analyses.

Welcome to Wordbank's blog!

This is the place where we will be announcing new features and datasets for the wordbank site (wordbank.stanford.edu). Stay tuned!

Resources

DatapagesStatic data-portal infrastructure used by this site MetaLabDatabase of developmental meta-analyses CHILDESCorpora of child and child-directed language childes-dbFlexible and reproducible interface to CHILDES PeekbankData from experiments measuring children's word recognition ManyBabiesMulti-lab replications in developmental psychology