This webpage brings functional analysis to some of the most useful open-source lexica available, providing sparse functional linear regression with group lasso through the push of a button. Simply upload your language data to have your texts and paralinguistics analyzed. The results will be generated as a .zip file available for download.

This resource is provided by Grant Packard, Yang Li, and Jonah Berger. Please send comments, suggestions, and bug reports to Please cite this tool as follows:

Grant Packard, Yang Li, and Jonah Berger (2024), “When Language Matters,” Journal of Consumer Research,

Select data file

A CSV data file you upload must adhere to the following format requirement. See the sample data file link at the end of this paragraph for reference. The first four columns are indices for conversation, turn, speaker (A and C, representing Agent and Customer in our context), and time. The 5th and 6th columns are conversation outcomes – a continuous measure and a count measure (DV_continuous, DV_count). Predictor/control variables start from the 7th column. Any focal functional variables that are to be left out of Group Lasso need to have a “__focal_func” suffix in their column names (starting with double underscores). Other functional variables that are to be considered for penalization by Group Lasso are indicated by a “__func” suffix (starting with double underscores). You can add a custom name to these before the suffix to replace the generic "v1", "v3" etc. shown in the sample data file. Any static predictor/control variables can use your own variable names (replacing the generic "v10", "v11", etc. in the sample data file). The algorithm recognizes the three types of variables using the suffix information. (Sample CSV file, R code)

Upload file(.csv or .xlsx)