| Type: | Package |
| Title: | Process Text and Compute Linguistic Alignment in Conversation Transcripts |
| Version: | 0.3.2 |
| Maintainer: | Jamie Reilly <jamie_reilly@temple.edu> |
| Description: | Imports conversation transcripts into R, concatenates them into a single dataframe appending event identifiers, cleans and formats the text, then yokes user-specified psycholinguistic database values to each word. 'ConversationAlign' then computes alignment indices between two interlocutors across each transcript for >40 possible semantic, lexical, and affective dimensions. In addition to alignment, 'ConversationAlign' also produces a table of analytics (e.g., token count, type-token-ratio) in a summary table describing your particular text corpus. |
| License: | LGPL (≥ 3) |
| Encoding: | UTF-8 |
| Depends: | R (≥ 3.5) |
| Imports: | DescTools, dplyr (≥ 0.4.3), httr, magrittr, purrr, rlang, stringi, stringr, textstem, tibble, tidyr, tidyselect, stats, utils, YRmisc, zoo |
| Suggests: | devtools, knitr, rmarkdown, testthat (≥ 3.0.0) |
| URL: | https://github.com/Reilly-ConceptsCognitionLab/ConversationAlign |
| RoxygenNote: | 7.3.2 |
| LazyData: | true |
| VignetteBuilder: | knitr |
| Collate: | 'ConversationAlign-package.R' 'compute_auc.R' 'compute_lagcorr.R' 'corpus_analytics.R' 'data.R' 'globals.R' 'prep_dyads.R' 'read_1file.R' 'read_dyads.R' 'replacements_25.R' 'summarize_dyads.R' 'utils.R' 'zzz.R' |
| Config/testthat/edition: | 3 |
| NeedsCompilation: | no |
| Packaged: | 2025-07-30 23:44:56 UTC; Jamie |
| Author: | Jamie Reilly 
[aut, cre],
Virginia Ulichney [aut],
Ben Sacks [aut],
Sarah Weinstein [ctb],
Chelsea Helion [ctb],
Gus Cooney [ctb] |
| Repository: | CRAN |
| Date/Publication: | 2025-07-31 00:00:02 UTC |
ConversationAlign: Process Text and Compute Linguistic Alignment in Conversation Transcripts
Description
Imports conversation transcripts into R, concatenates them into a single dataframe appending event identifiers, cleans and formats the text, then yokes user-specified psycholinguistic database values to each word. 'ConversationAlign' then computes alignment indices between two interlocutors across each transcript for >40 possible semantic, lexical, and affective dimensions. In addition to alignment, 'ConversationAlign' also produces a table of analytics (e.g., token count, type-token-ratio) in a summary table describing your particular text corpus.
Author(s)
Maintainer: Jamie Reilly jamie_reilly@temple.edu (ORCID)
Authors:
Virginia Ulichney
Ben Sacks
Other contributors:
Sarah Weinstein [contributor]
Chelsea Helion [contributor]
Gus Cooney [contributor]
See Also
Useful links:
Sample Dyadic Interview Transcript: Marc Maron and Terry Gross Radio Interview 2013
Description
Text and talker information delineated, raw transcript, multiple lines per talker
Usage
MaronGross_2013
Format
## "MaronGross_2013" A data.frame with 546 obs, 2 vars:
- text
text from interview
- speaker
speaker identity
...
Text and talker information delineated, 3 separate nursery rhymes, good for computing analytics and word counts
Description
Text and talker information delineated, 3 separate nursery rhymes, good for computing analytics and word counts
Usage
NurseryRhymes
Format
## "NurseryRhymes" A data.frame with 100 observations, 2 vars:
- Event_ID
factor 3 different simulated conversations
- Participant_ID
fictional speaker names, 2 each conversation
- Text_Raw
simulated language production, actually looped phrases from nursery rhymes
...
Text and talker information delineated, 3 separate nursery rhymes, good for computing analytics and word counts
Description
Text and talker information delineated, 3 separate nursery rhymes, good for computing analytics and word counts
Usage
NurseryRhymes_Prepped
Format
## "NurseryRhymes_Prepped" A data.frame with 1507 x 7 observations, 5 vars:
- Event_ID
factor 3 different simulated conversations
- Participant_ID
fictional speaker names, 2 each conversation
- Exchange_Count
sequential numbering of exchanges by conversation, 1 exchange = 2 turns
- Turn_Count
sequential numbering of turns by conversation
- Text_Clean
content words
- emo_anger
raw value of anger salience yoked to each word
...
corpus_analytics
Description
Produces a table of corpus analytics including numbers of complete observations at each step, word counts, lexical diversity (e.g., TTR), stopword ratios, etc. Granularity of the summary statistics are guided by the user (e.g., by conversation, by conversation and speaker, collapsed all)
Usage
corpus_analytics(dat_prep)
Arguments
dat_prep |
takes dataframe produced from the df_prep() function |
Value
dataframe with summary statistics (mean, SD, range) for numerous corpus analytics (e.g., token count, type-token-ratio, word-count-per-turn) for the target conversation corpus. Summary data structured in table format for easy export to a journal method section.
Load all .rda files from a GitHub data folder into the package environment
Description
Load all .rda files from a GitHub data folder into the package environment
Usage
load_github_data(
repo = "Reilly-ConceptsCognitionLab/ConversationAlign_Data",
branch = "main",
data_folder = "data",
envir = parent.frame()
)
Arguments
repo |
GitHub repository (e.g., "username/repo") |
branch |
Branch name (default: "main") |
data_folder |
Remote folder containing .rda files (default: "data/") |
envir |
Environment to load into (default: package namespace) |
Value
nothing, loads data (as rda files) from github repository needed for other package functions
prep_dyads
Description
Cleans, vectorizes and appends lexical norms to all content words in a language corpus. User guides options for stopword removal and lemmatization. User selects up to three psycholinguistic dimensions to yoke norms on each content word in the original conversation transcript.
Usage
prep_dyads(
dat_read,
lemmatize = TRUE,
omit_stops = TRUE,
which_stoplist = "Temple_stops25",
verbose = TRUE
)
Arguments
dat_read |
dataframe produced from read_dyads() function |
lemmatize |
logical, should words be lemmatized (switched to base morphological form), default is TRUE |
omit_stops |
option to remove stopwords, default TRUE |
which_stoplist |
user-specified stopword removal method with options including "none", "SMART", "MIT_stops", "CA_OriginalStops", or "Temple_Stopwords25". "Temple_Stopwords25 is the default list |
verbose |
display detailed output such as error messages and progress (default is TRUE) |
Value
dataframe with text cleaned and vectorized to a one word per-row format. Lexical norms and metadata are appended to each content word. Cleaned text appears under a new column called 'Text_Clean'. Any selected dimensions (e.g., word length) and metadata are also appended to each word along with speaker identity, turn, and Event_ID (conversation identifier).
read_1file
Description
Reads pre-formatted dyadic (2 interlocutor) conversation transcript already imported into your R environment.
Usage
read_1file(my_dat)
Arguments
my_dat |
one conversation transcript already in the R environment |
Value
a dataframe formatted with 'Event_ID', "Participant_ID", "Text_Raw" fields – ready for clean_dyads()
read_dyads
Description
Reads pre-formatted dyadic (2 interlocutor) conversation transcripts from your machine. Transcripts must be either csv or txt format. IF you are supplying a txt file, your transcript must be formatted as an otter.ai txt file export. Your options for using csv files are more flexible. ConversationAlign minimally requires a csv file with two columns, denoting interlocutor and text. Each separate conversation transcript should be saved as a separate file. ConversationAlign will use the file names as a document ID. Within the read dyads function, set the my_path argument as the directory path to the local folder containing your transcripts on your machine (e.g., "my_transcripts"). Please see our github page for examples of properly formatted transcripts: https://github.com/Reilly-ConceptsCognitionLab/ConversationAlign
Usage
read_dyads(my_path = "my_transcripts")
Arguments
my_path |
folder of conversation transcripts in csv or txt format |
Value
a dataframe where each individual conversation transcript in a user's directory has been concatenated. read_dyads appends a unique document identifier to each conversation transcript appending its unique filename as a factor level to 'Event_ID'.
summarize_dyads
Description
Calculates and appends 3 measures for quantifying alignment. Appends the averaged value for each selected dimension by turn and speaker. Calculates and Spearman's rank correlation between interlocutor time series and appends by transcript. Calculates the area under the curve of the absolute difference time series between interlocutor time series. The length of the difference time series can be standardized the shortest number of exchanges present in the group using an internally defined resampling function, called with resample = TRUE. Spearman's rank correlation and area under the curve become less reliable for dyads under 30 exchanges.
Usage
summarize_dyads(
df_prep,
custom_lags = NULL,
sumdat_only = TRUE,
corr_type = "Pearson"
)
Arguments
df_prep |
produced in the align_dyads function |
custom_lags |
integer vector, should any lags be added in addition to -2, 0, 2 |
sumdat_only |
default=TRUE, group and summarize data, two rows per conversation, one row for each participant, false will fill down summary statistics across all exchanges |
corr_type |
option for computing lagged correlations turn-by-turn covariance (default='Pearson') |
Value
either: - a grouped dataframe with summary data aggregated by converation (Event_ID) and participant if sumdat_only=T. - the origoinal dataframe 'filled down' with summary data (e.g., AUC, turn-by-turn correlations) for each conversation is sumdat_only=F.