To update the NLP thresholds and configurations
To use this API, app needs the Bot Builder API scope of NLP Configurations.
OR the Admin API Scope of NLP Configurations under Test and Train.
POST https://{{host}}/api/public/bot/{{BotID}}/configurations?language={{languageCode}}&groupName={{groupName}}
Query Parameters
| Parameter | Required/Optional | Description |
|---|---|---|
| host | Required | Environment URL, for example, https://bots.kore.ai |
| BotID | Required | Bot ID or Stream ID. You can access it from the General Settings page of the bot. |
| languageCode | Required | Bot language to which these configurations need to be updated. |
| groupName | Required for Multiple ML Model |
GroupName can be used to update the ML params of a specific group. To update Bot level Intent Model Configurations groupName should be set to “Bot Level Intent Model”. |
Body Parameters
The parameters would depend upon the threshold configuration that needs to be updated. Following is the comprehensive list of parameters for the various threshold configurations.
Update Settings
Machine Learning Engine
This section refers to the configurations related to the Machine Learning engine. For details about the configurations refer here.
{
"configurations": [
{
"mode": "ml", // Machine Learning Engine
"exactMatchThreshold": 90, // ML Definitive Score - value in range [80-100]
"minThreshold": 0.4, // ML threshold - value in range [0-1]
}
],
"mlParams": {
"intentParams": {
"useSynonyms": true, // Bot Synonyms
"useStopwords": true, // Stopwords
"usePlaceholders": true, // Entity Placeholders
"features": "n_gram" // Feature Extraction - value in range [skip_gram, n_gram]
"skip_gram": { // features should be 'skip_gram'
"seqLength": 2, // Sequence Length - value in range [2-4]
"maxSkipDistance": 1 // Maximum Skip Distance - value in range [1-3]
},
"ngram": 3, // ngram Sequence Length - value in range [1-4]
// features should be ‘n_gram’
},
"nerParams": {
"type": "corenlp" // NER Model
// could be "corenlp" for Conditional Random Field
// or "spacy" - Deep Neural Network
}
}
}
Fundamental Meaning Engine
This section refers to the configurations related to the Fundamental Meaning engine. For details about the configurations refer here.
{
"configurations": [
{
"mode": "cs", // Fundamental Meaning Engine
"labelMatch": true // Intent Detection using Task Name Words
"isFMThreshold": true, //FM Threshold
"fmThreshold": 15 //FM Threshold value
}
]
}
Knowledge Graph Engine
This section refers to the configurations related to the Knowledge Graph engine. For details about the configurations refer here.
{
"configurations": [ THRESHOLDS & CONFIGURATIONS
{
"mode": "faq",
"autoSpellCorrectEnabled": true, // Auto Correction
"useBotSynonyms": true, // Bot Synonyms
"enPatternLemma": false, // Lemmatization using Parts of Speech
"pathCoverage": 50, // Path Coverage - value in range [0-100]
"exactMatchThreshold": 95, // Minimum and Definitive Level for Knowledge Tasks
// value in range [0-100]
// should be more than minThreshold & maxThreshold
"maxThreshold": 80, // Probable range for Knowledge Task
// value in range [0-100]
// should be between minThreshold & exactMatchThreshold
"minThreshold": 60, // Low confidence range for Knowledge Task
// value in range [0-100]
// should be less than maxThreshold & exactMatchThreshold
"suggestionsCount": 3, // KG suggestionsCount | [0-5]
"taskMatchTolerance": 35, // Proximity of Suggested Matches | [0-50]
"longResponses": {
"readMore": true, // Manage Long Responses
"useCustomReadMoreURL": true, // readMore should be true
"customReadMoreURL": "www.siteurl.com" // readMore link; readMore should be true
},
"searchInAnswer": {
"enabled": true, // Search in Answer
"notifyUser": true, // Inform end user that it is a probable answer
"responseType": "relevantWithReadMore", // can be "complete" - Show complete response,
// or "relevant" - Show only the relevant paragraph,
// or "relevantWithReadMore" - Show only the relevant paragraph with “read more” link ]
"customReadMoreURL": "www.siteurl.com", // “read more” link
"useCustomReadMoreURL": true // enabled should be true
}
"qualifyContextualPaths": false, // Qualify Contextual Paths
}
]
}
Ranking & Resolver Engine
This section refers to the configurations related to the Ranking & Resolver engine. For details about the configurations refer here.
{
"configurations":
{
"mode": "rr", // Ranking and Resolver Engine
"taskMatchTolerance": 2, // Proximity of Probable Matches - value in range [0-20]
"useDependencyParser": true, // Dependency Parsing Model
"minMatchVal": 0.4, // Minimum Match Score - value in range [0-1]
// useDependencyParser should be true
"rankingParameters": "" // Advanced Configurations
// useDependencyParser should be true
"intentRescoring": false, // Rescoring of Intents
"isPreferDefinitiveMatch": true // Prefer Definitive Matches
}
]
}
Advanced NLP Configurations
This section refers to the configurations as given in the Advanced NLP Configurations section of Natural Language -> Training. For details about the configurations refer here.
Following is the sample request for setting the above-mentioned Advanced NLP Configurations.
{
"advancedNLPSettings":[
{
"configurationKeyName": "NoneIntent",
"configurationValue":true,
"nlpEngine":"ML"
}
]
}
Following is the complete list of the permissible advanced NLP configurations settings and the possible values:
Split Compound Words
{
"configurationName": "Split Compound Words",
"configurationKeyName": "splitCompoundWords",
"desc": "The settings enables splitting of the compound words into multiple stems and then process the individual stem",
"nlpEngine": "ML",
"dataType": "Boolean",
"range": [true, false],
"defaultValue": false,
"requiresTraining": true,
"language": ["de", "nl"],
"isChild": false
}
None Intent
{
"configurationName": "None Intent",
"configurationKeyName": "NoneIntent",
"desc": "Once enabled, a None intent is created which reduces the chances of getting a false positive for an intent match using ML engine",
"nlpEngine": "ML",
"dataType": "Boolean",
"range": [true, false],
"defaultValue": true,
"requiresTraining": true,
"language": "ALL",
"isChild": false
}
Cosine similarity dampening
{
"configurationName": "Cosine similarity dampening",
"configurationKeyName": "cosineSimilarityDampening",
"desc": "Avoid penalty on short length questions using Cosine Similarity Dampening",
"nlpEngine": "KG",
"dataType": "Boolean",
"range": [true, false],
"defaultValue": true,
"requiresTraining": false,
"language": "ALL",
"isChild": false
}
Network Type
{
"configurationName" : "Network Type",
"configurationKeyName" : "network",
"desc" : "Networks available for intent training",
"nlpEngine" : "ML",
"dataType" : "enum",
"range" : ["Standard","MLP-BOW","MLP-WordEmbeddings","LSTM","CNN", "KoreBERT"],
"defaultValue": "Standard",
"requiresTraining" : true,
"language" : "ALL",
"isChild" : false
}
Epochs
{
"configurationName": "Epochs",
"configurationKeyName": "epochs",
"desc": "Number of iterations in training the model",
"nlpEngine": "ML",
"dataType": "Number",
"range": [20, 300],
"defaultValue": 20,
"requiresTraining" : true,
"language" : "ALL",
"isChild" : true,
"parentConfiguration" : "network",
"requiredParentInput" : ["MLP-BOW","MLP-WordEmbeddings","LSTM","CNN","KoreBERT"]
}
Batch Size
{
"configurationName" : "Batch Size",
"configurationKeyName" : "batch_size",
"desc" : "Number of training samples used for each batch while training",
"nlpEngine" : "ML",
"dataType" : "Number",
"range" : [10,30],
"defaultValue":10,
"requiresTraining" : true,
"language" : "ALL",
"isChild" : true,
"parentConfiguration" : "network",
"requiredParentInput" : ["MLP-BOW","MLP-WordEmbeddings","LSTM","CNN","KoreBERT"]
}
Learning rate
{
"configurationName": "Learning rate",
"configurationKeyName": "lr",
"desc": "Learning rate is a hyper-parameter that controls how much we are adjusting the weights of our network with respect to the loss gradient",
"nlpEngine": "ML",
"dataType": "enum",
"range": [1e-4, 1e-3, 1e-2],
"defaultValue": 1e-3,
"requiresTraining" : true,
"language" : "ALL",
"isChild" : true,
"parentConfiguration" : "network",
"requiredParentInput" : ["MLP-BOW","MLP-WordEmbeddings","LSTM","CNN","KoreBERT"]
}
Dropout
{
"configurationName": "Dropout",
"configurationKeyName": "dropout",
"desc": "Regularization parameter to avoid overfitting of the model",
"nlpEngine": "ML",
"dataType": "Number",
"range": [0, 0.8],
"defaultValue": 0,
"requiresTraining" : true,
"language" : "ALL",
"isChild" : true,
"parentConfiguration" : "network",
"requiredParentInput" : ["MLP-BOW","MLP-WordEmbeddings","LSTM","CNN", "KoreBERT"]
}
Vectorizer
{
"configurationName": "Vectorizer",
"configurationKeyName": "vector_type",
"desc": "Feature extraction technique on training data",
"nlpEngine": "ML",
"dataType": "enum",
"range": ["count", "tfidf"],
"defaultValue": "count",
"requiresTraining": true,
"language": "ALL",
"isChild": true,
"parentConfiguration": "network",
"requiredParentInput": ["MLP-BOW"]
}
Maximum sequence length
{
"configurationName": "Maximum sequence length",
"configurationKeyName": "max_seq_length",
"desc": "Length of the training sample or user input",
"nlpEngine": "ML",
"dataType": "Number",
"range": [10, 30],
"defaultValue": 20,
"requiresTraining" : true,
"language" : "ALL",
"isChild" : true,
"parentConfiguration" : "network",
"requiredParentInput" : ["MLP-WordEmbeddings","LSTM","CNN","KoreBERT"]
}
Embeddings Type
{
"configurationName": "Embeddings Type",
"configurationKeyName": "word_embedding_type",
"desc": "Feature extraction technique on training data",
"nlpEngine": "ML",
"dataType": "enum",
"range": ["generated", "pretrained", "random"],
"defaultValue": "random",
"requiresTraining" : true,
"language" : "ALL",
"isChild" : true,
"parentConfiguration" : "network",
"requiredParentInput" : ["MLP-WordEmbeddings","LSTM","CNN", "KoreBERT"]
}
Embeddings Dimensions
{
"configurationName": "Embeddings Dimensions",
"configurationKeyName": "embedding_dim",
"desc": "Embeddings Dimensions to be used in featurization",
"nlpEngine": "ML",
"dataType": "Number",
"range": [100, 400],
"defaultValue": 300,
"requiresTraining" : true,
"language" : "ALL",
"isChild" : true,
"parentConfiguration" : "network",
"requiredParentInput" : ["MLP-WordEmbeddings","LSTM","CNN", "KoreBERT"]
}
K Fold Cross-Validation
{
"configurationName" : "kfold",
"configurationKeyName" : "kfold",
"desc" : "kfold parameter for Crossvalidation",
"nlpEngine" : "ML",
"dataType" : "Number",
"range" : [2,10],
"defaultValue": 5,
"requiresTraining" : true,
"language" : "ALL",
"isChild" : false
}
FAQ Name as Intent Name
{
"configurationName": "FAQ Name as Intent Name",
"configurationKeyName": "FAQ_Name_Intent_Name",
"desc": "Enable to use FAQ Name as Intent Name. ",
"nlpEngine": "KG",
"dataType": "Boolean",
"range": [true, false],
"defaultValue": false,
"requiresTraining" : false,
"language" : "ALL",
"isChild" : false
}
Fuzzy Match
{
"configurationName" : "Fuzzy Match",
"configurationKeyName" : "fuzzyMatch",
"desc" : "This setting enables the use of the fuzzy matching algorithm for intent identification",
"nlpEngine" : "ML",
"dataType" : "Boolean",
"range" : [true,false],
"defaultValue": true,
"requiresTraining" : false,
"language" : "ALL",
"isChild" : false
}
Handle Negation
{
"configurationName" : "Handle Negation",
"configurationKeyName" : "f_negation",
"desc" : "This setting enables the handling of negated words in intent identification",
"nlpEngine" : "ML",
"dataType" : "Boolean",
"range" : [true,false],
"defaultValue": true,
"requiresTraining" : true,
"language" : ["en", "es"],
"isChild" : false
}
Ignore Multiple Occurences
{
"configurationName" : "Ignore Multiple occurences",
"configurationKeyName" : "binary",
"desc" : "This setting enables to just use presence or absence of a term instead of the raw counts",
"nlpEngine" : "ML",
"dataType" : "Boolean",
"range" : [true,false],
"defaultValue": false,
"requiresTraining" : true,
"language" : ["en"],
"isChild" : true,
"parentConfiguration" : "network",
"requiredParentInput" : ["MLP-BOW"]
}
Entity Placeholders in User Utterances
{
"configurationName": "Test PlaceHolders",
"configurationKeyName": "TestPlaceHolders",
"desc": "Enable to replace entity values present in the user input with the corresponding entity placeholders during prediction",
"nlpEngine": "ML",
"dataType": "Boolean",
"range": [true, false],
"defaultValue": true,
"requiresTraining": false,
"language": "ALL",
"isChild": false
}
Delete Settings
To delete any settings use the name of the settings mentioned in above sections in this way:
curl --location -g --request POST \
'https://{{host}}/api/public/bot/{{bot-id}}/configurations?language={{languageCode}}' \
--header 'auth: YOUR_JWT_ACCESS_TOKEN' \
--header 'content-type: application/json' \
--data-raw '{
"delete": {
"advancedNLPSettings": [
"kfold"
]
}
}'
Reset Settings
To reset any settings use the name of the settings mentioned in above sections in this way:
curl --location -g --request POST \
'https://{{host}}/api/public/bot/{{bot-id}}/configurations?language={{languageCode}}' \
--header 'auth: YOUR_JWT_ACCESS_TOKEN' \
--header 'content-type: application/json' \
--data-raw '{
"reset": {
"faq": [
"pathCoverage", // Path Coverage
"suggestionsCount", // KG suggestionsCount
"minimumAndDefinitiveLevels", // Minimum and Definitive Level for Knowledge Tasks (will reset all three values)
"taskMatchTolerance" // Proximity of Suggested Matches
],
"ml": [
"exactMatchThreshold", // ML Definitive Score
"minThreshold" // ML threshold
],
"rr": [
"taskMatchTolerance", // Proximity of Probable Matches
"minMatchVal" // Minimum Match Score
],
"mlParams": [
"ngram", // ngram Sequence Length
"seqLength", // Sequence Length
"maxSkipDistance" // Maximum Skip Distance
]
}
}'
Authorization
Invoke the API with JWT in the header as follows:
auth: {{JWT}}
Response Content Type
application/json
Sample Request
curl --location -g --request POST \
'https://{{host}}/api/public/bot/{{bot-id}}/configurations?language={{languageCode}}' \
--header 'auth: YOUR_JWT_ACCESS_TOKEN' \
--header 'content-type: application/json' \
--data-raw '{
"advancedNLPSettings": [
{
"configurationKeyName": "NoneIntent",
"configurationValue": true,
"nlpEngine": "ML"
}
],
"configurations": [
{
"mode": "ml",
"exactMatchThreshold": 85,
"useDependencyParser": true,
"minThreshold": 0.2
},
{
"mode": "faq",
"useBotSynonyms": true,
"searchInAnswer": {
"enabled": true,
"notifyUser": false,
"responseType": "relevantWithReadMore",
"customReadMoreURL": "aa",
"useCustomReadMoreURL": true
}
}
],
"mlParams": {
"intentParams": {
"features": "skip_gram",
"skip_gram": {
"seqLength": 3,
"maxSkipDistance": 2
}
}
},
"nlSettings": {
"enableAutoUtteranceAddition": false,
"enableNegativePatterns": true
}
}'
Sample Response
[
{
"message": "Training Queued.",
"Training_ID": "5d14b03edba48abcb44375a1"
}
]