Per Microsoft’s website, the Text Analytics REST API is a suite of text analytics web services built with Azure Machine Learning that can be used to analyze unstructured text. The API supports four text analysis operations:
The API returns a numeric score between 0 and 1. Scores close to 1 indicate positive sentiment and scores close to 0 indicate negative sentiment. Sentiment score is generated using classification techniques. The input features of the classifier include n-grams, features generated from part-of-speech tags, and word embeddings. English, French, Spanish and Portuguese text are supported.
This API returns the detected topics for a list of submitted text records. A topic is identified with a key phrase, which can be one or more related words. This API requires a minimum of 100 text records to be submitted, but is designed to detect topics across hundreds to thousands of records. The API is designed to work well for short, human-written text such as reviews and user feedback.
This API returns the detected language and a numeric score between 0 and 1. Scores close to 1 indicate 100% certainty that the identified language is correct. A total of 120 languages are supported.
This API returns a list of strings denoting the key talking points in the input text. English, German, Spanish, and Japanese text are supported.
You can either install the latest stable version from CRAN:
if ("mscstexta4r" %in% installed.packages()[,"Package"] == FALSE) {
install.packages("mscstexta4r")
}Or, you can install the development version:
if ("mscstexta4r" %in% installed.packages()[,"Package"] == FALSE) {
if ("devtools" %in% installed.packages()[,"Package"] == FALSE) {
install.packages("devtools")
}
devtools::install_github("philferriere/mscstexta4r")
}After loading {mscstexta4r} with library(), you must call textaInit() before you can call any of the core {mscstexta4r} functions.
The textaInit() configuration function will first check to see if the variable MSCS_TEXTANALYTICS_CONFIG_FILE exists in the system environment. If it does, the package will use that as the path to the configuration file.
If MSCS_TEXTANALYTICS_CONFIG_FILE doesn’t exist, it will look for the file .mscskeys.json in the current user’s home directory (that’s ~/.mscskeys.json on Linux, and something like C:\Users\Phil\Documents\.mscskeys.json on Windows). If the file is found, the package will load the API key and URL from it.
If using a file, please make sure it has the following structure:
{
"textanalyticsurl": "https://westus.api.cognitive.microsoft.com/texta/analytics/v2.0/",
"textanalyticskey": "...MSCS Text Analytics API key goes here..."
}If no configuration file is found, textaInit() will attempt to pick up its configuration from two Sys env variables instead:
MSCS_TEXTANALYTICS_URL - the URL for the Text Analytics REST API.
MSCS_TEXTANALYTICS_KEY - your personal Text Analytics REST API key.
textaInit() needs to be called only once, after package load.
The MSCS Text Analytics API is a RESTful API. HTTP requests over a network and the Internet can fail. Because of congestion, because the web site is down for maintenance, because of firewall configuration issues, etc. There are many possible points of failure.
The API can also fail if you’ve exhausted your call volume quota or are exceeding the API calls rate limit. Unfortunately, MSCS does not expose an API you can query to check if you’re about to exceed your quota for instance. The only way you’ll know for sure is by looking at the error code returned after an API call has failed.
Therefore, you must write your R code with failure in mind. Our preferred way is to use tryCatch(). Its mechanism may appear a bit daunting at first, but it is well documented. We’ve also included many examples, as you’ll see below.
All but one core text analytics functions execute exclusively in synchronous mode. textaDetectTopics() is the only function that can be executed either synchronously or asynchronously. Why? Because topic detection is typically a “batch” operation meant to be performed on thousands of related documents (product reviews, research articles, etc.).
When textaDetectTopics() executes synchronously, you must wait for it to finish before you can move on to the next task. When textaDetectTopics() executes asynchronously, you can move on to something else before topic detection has completed. In the latter case, you will need to call textaDetectTopicsStatus() periodically yourself until the Microsoft Cognitive Services server complete topic detection and results become available.
If you’re performing topic detection in batch mode (from an R script), we recommend using the textaDetectTopics() in synchronous mode, in which case, again, it will return only after topic detection has completed.
If you’re calling textaDetectTopics() in synchronous mode within the R console REPL (interactive mode), it will appear as if the console has hanged. This is EXPECTED. The function hasn’t crashed. It is simply in “sleep mode”, activating itself periodically and then going back to sleep, until the results have become available. In sleep mode, even though it appears “stuck”, textaDetectTopics() doesn’t use any CPU resources. While the function is operating in sleep mode, you WILL NOT be able to use the console before the function completes. If you need to operate the console while topic detection is being performed by the Microsoft Cognitive services servers, you should call textaDetectTopics() in asynchronous mode and then call textaDetectTopicsStatus() yourself repeatedly afterwards, until results are available.
Here’s some sample code that illustrates how to use tryCatch():
library('mscstexta4r')
tryCatch({
textaInit()
}, error = function(err) {
geterrmessage()
})If {mscstexta4r} cannot locate .mscskeys.json nor any of the configuration environment variables, the code above will generate the following output:
[1] "mscstexta4r: could not load config info from Sys env nor from file"Similarly, textaInit() will fail if {mscstexta4r} cannot find the textanalyticskey key in .mscskeys.json, or fails to parse it correctly, etc. This is why it is so important to use tryCatch() with all {mscstexta4r} functions.
The core API calls exposed by {mscstexta4r} are the following:
# Perform sentiment analysis
textaSentiment(
documents, # Input sentences or documents
languages = rep("en", length(documents))
# "en"(English, default)|"es"(Spanish)|"fr"(French)|"pt"(Portuguese)
)# Detect top topics in group of documents
textaDetectTopics(
documents, # At least 100 documents
stopWords = NULL, # Stop word list (optional)
topicsToExclude = NULL, # Topics to exclude (optional)
minDocumentsPerWord = NULL, # Threshold to exclude rare topics (optional)
maxDocumentsPerWord = NULL, # Threshold to exclude ubiquitous topics (optional)
resultsPollInterval = 30L, # Poll interval (in s, default: 30s, use 0L for async)
resultsTimeout = 1200L, # Give up timeout (in s, default: 1200s = 20mn)
verbose = FALSE # If set to TRUE, print every poll status to stdout
)# Detect languages used in documents
textaDetectLanguages(
documents, # Input sentences or documents
numberOfLanguagesToDetect = 1L # Default: 1L
) # Get key talking points in documents
textaKeyPhrases(
documents, # Input sentences or documents
languages = rep("en", length(documents))
# "en"(English, default)|"de"(German)|"es"(Spanish)|"fr"(French)|"ja"(Japanese)
)The functions textaDetectTopics() returns a S3 class object of the class textatopics. The textatopics object exposes formatted results using several dataframes (documents and their IDs, topics and their IDs, which topics are assigned to which documents), the REST API JSON response (should you care), and the HTTP request (mostly for debugging purposes).
The other functions return S3 class objects of the class texta. The texta object exposes results collected in a single data.frame, the REST API JSON response, and the original HTTP request.
The following code snippets illustrate how to use {mscstexta4r} functions and show what results they return with toy examples. If after reviewing this code there is still confusion regarding how and when to use each function, please refer to the original documentation.
docsText <- c(
"Loved the food, service and atmosphere! We'll definitely be back.",
"Very good food, reasonable prices, excellent service.",
"It was a great restaurant.",
"If steak is what you want, this is the place.",
"The atmosphere is pretty bad but the food is quite good.",
"The food is quite good but the atmosphere is pretty bad.",
"The food wasn't very good.",
"I'm not sure I would come back to this restaurant.",
"While the food was good the service was a disappointment.",
"I was very disappointed with both the service and my entree."
)
docsLanguage <- rep("en", length(docsText))
tryCatch({
# Perform sentiment analysis
textaSentiment(
documents = docsText, # Input sentences or documents
languages = docsLanguage
# "en"(English, default)|"es"(Spanish)|"fr"(French)|"pt"(Portuguese)
)
}, error = function(err) {
# Print error
geterrmessage()
})# Load yelpChReviews100 text reviews
load("../tests/testthat/data/yelpChineseRestaurantReviews100.rda")
tryCatch({
# Detect top topics
textaDetectTopics(
documents = yelpChReviews100, # At least 100 docs/sentences
stopWords = NULL, # Stop word list (optional)
topicsToExclude = NULL, # Topics to exclude (optional)
minDocumentsPerWord = NULL, # Threshold to exclude rare topics (optional)
maxDocumentsPerWord = NULL, # Threshold to exclude ubiquitous topics (optional)
resultsPollInterval = 60L, # Poll interval (in s, default: 30s, use 0L for async)
resultsTimeout = 1200L, # Give up timeout (in s, default: 1200s = 20mn)
verbose = TRUE # If set to TRUE, print every poll status to stdout
)
}, error = function(err) {
# Print error
geterrmessage()
})# Load yelpChReviews100 text reviews
load("../tests/testthat/data/yelpChineseRestaurantReviews100.rda")
tryCatch({
# Detect top topics
operation <- textaDetectTopics(
documents = yelpChReviews100, # At least 100 docs/sentences
resultsPollInterval = 0L, # Poll interval (in s, default: 30s, use 0L for async)
verbose = TRUE # If set to TRUE, print every poll status to stdout
)
# Poll the servers ourselves, until the work completes or until we time out
resultsPollInterval <- 60L
resultsTimeout <- 1200L
startTime <- Sys.time()
endTime <- startTime + resultsTimeout
while (Sys.time() <= endTime) {
sleepTime <- startTime + resultsPollInterval - Sys.time()
if (sleepTime > 0)
Sys.sleep(sleepTime)
startTime <- Sys.time()
# Poll for results
topics <- textaDetectTopicsStatus(operation, verbose = TRUE)
if (topics$status != "NotStarted" && topics$status != "Running")
break;
}
topics
}, error = function(err) {
# Print error
geterrmessage()
})docsText = c(
"The Louvre or the Louvre Museum is the world's largest museum.",
"Le musée du Louvre est un musée d'art et d'antiquités situé au centre de Paris.",
"El Museo del Louvre es el museo nacional de Francia.",
"Il Museo del Louvre a Parigi, in Francia, è uno dei più celebri musei del mondo.",
"Der Louvre ist ein Museum in Paris."
)
tryCatch({
# Detect languages
textaDetectLanguages(
documents = docsText, # Input sentences or documents
numberOfLanguagesToDetect = 1L # Number of languages to detect
)
}, error = function(err) {
# Print error
geterrmessage()
})docsText <- c(
"Loved the food, service and atmosphere! We'll definitely be back.",
"Very good food, reasonable prices, excellent service.",
"It was a great restaurant.",
"If steak is what you want, this is the place.",
"The atmosphere is pretty bad but the food is quite good.",
"The food is quite good but the atmosphere is pretty bad.",
"I'm not sure I would come back to this restaurant.",
"The food wasn't very good.",
"While the food was good the service was a disappointment.",
"I was very disappointed with both the service and my entree."
)
docsLanguage <- rep("en", length(docsText))
tryCatch({
# Get key talking points in documents
textaKeyPhrases(
documents = docsText, # Input sentences or documents
languages = docsLanguage
# "en"(English, default)|"de"(German)|"es"(Spanish)|"fr"(French)|"ja"(Japanese)
)
}, error = function(err) {
# Print error
geterrmessage()
})A test/demo Shiny web application is available here