You can fine-tune intent detection for each language enabled for your Virtual Assistant (VA). To perform this action, follow the below steps:

1. On the left pane, click Natural Language > Training > Thresholds & Configurations.
2. Under the Thresholds & Configurations section, you can perform by customizing
   • The Fundamental Meaning model – see here for details
   • Machine Learning model – see here for details
   • Knowledge Graph – see here for details
   • Ranking & Resolver engine – see here for details

Apart from these, under the Advanced NLP Configurations section, there are advanced settings that you can use for specific use cases and requirements.

Warning: The default settings for these configurations are ideal for most use cases. Do not change these settings unless you are fully acquainted with the functionality you are setting, as they might have a detrimental effect on the VA’s performance if not done properly.

If your bot is trained in one or more languages that use NLP Version 2, you will see a banner notifying you about the upcoming auto-upgrade to NLP Version 3.

The following table gives the details of the various configurations that can be set from this section. Apart from these you can add Custom configurations, reach out to our support team to know how.

Configuration	Description	Affected NLP Engine	Valid Inputs	Notes
Split Compound Words	The setting enables the splitting of the compound words into multiple stems and then processing the individual stem.	ML	Enable,¶ Disable (default)	Supported only for assistants in the German language
Split Combined Words	This setting enables the splitting of incorrectly combined words into valid words and then processing the individual words.	ML	Enable,¶ Disable (default)	Supported only for assistants in the Spanish language
None Intent	Once enabled, a dummy, placeholder intent is created which reduces the chances of getting a false positive for an intent match using the ML engine.	ML	Enable (default), Disable
Epochs	Number iterations for training the neural network.	ML	Between 20 and 300, increments of 10 (default setting 20)	Valid only when Network Type is set to MLP-BOW, MLP-WordEmbeddings, LSTM, CNN
Batch Size	Number of training samples used for each batch while training	ML	Between 10 and 30, increments of 5 (default setting 10)	Valid only when Network Type is set to MLP-BOW, MLP-WordEmbeddings, LSTM, CNN
Learning rate	A hyper-parameter to control how much the weights of the network are adjusted with respect to the loss gradient	ML	Between 1e-4 and 1e-3, increments of 1e-2 (default setting 1.00E-03)	Valid only when Network Type is set to MLP-BOW, MLP-WordEmbeddings, LSTM, CNN
Dropout	Regularization parameter to avoid overfitting of the model	ML	Between 0 and 0.8, increments of 0.1 (default setting 0)	Valid only when Network Type is set to MLP-BOW, MLP-WordEmbeddings, LSTM, CNN
Vectorizer	Feature extraction technique on training data	ML	count (default), tfidf	Valid only when Network Type is set to MLP-BOW
Maximum sequence length	Length of the training sample or user input	ML	Between 10 and 30, increments of 5 (default setting 20)	Valid only when Network Type is set to MLP-WordEmbeddings, LSTM, CNN
Embeddings Type	Feature extraction technique on training data	ML	generated, random (default)	Valid only when Network Type is set to MLP-WordEmbeddings, LSTM, CNN
Embeddings Dimensions	Embeddings Dimensions to be used in featurization	ML	Between 100 and 400, increments of 50 (default setting 300)	Valid only when Network Type is set to MLP-WordEmbeddings, LSTM, CNN
K Fold	kfold parameter for Cross-validation	ML	Between 2 and 10, increments of 1 (default setting 2)
Fuzzy Match	This setting enables the use of the fuzzy matching algorithm for intent identification	ML	Enable (default), Disable
Handle Negation	This setting enables the handling of negated words in intent identification	ML	Enable (default), Disable
Ignore Multiple Occurences	Once enabled, the frequency of the words are disregarded for vectorization	ML	Enable (default), Disable	Valid only when Network Type is set to MLP-BOW
Entity Placeholders in User Utterances	Enable to replace entities present in user utterances with corresponding placeholders	ML	Enable (default), Disable	Valid only when Network Type is set to MLP-BOW
Sentence Split	Split the sentences in user utterance and perform intent detection using the complete user input	ML	Enable (default), Disable
Multiple Intent Models	Enable separate ML models for each of the primary intents constituting all its sub-intents	ML	Enable, Disable (default)
Lemmatize KG Synonyms	Enable to use the lemmatized versions of KG Synonyms to detect intents.	KG	Enable (default), Disable.
Use only Tagged Utterances for NER Training	Uses only the Tagged Utterances for Named Entity Recognition (NER) Training. When enabled, the ML Engine uses the utterances with one or more NER tags, to train the NER model for entity detection.	ML	Enable (default), Disable
Neurons in Hidden Layer	Used to configure the number of Neurons used in the Hidden Layer	ML	Range – 0 to 1000	Applicable only for Standard Network Type
Softmax Temperature	Use to define how confidently the ML Engine should identify the winning intent from the ML Model. Temperature is a hyperparameter that is applied to logits(Model outputs) to affect the final probabilities from the softmax.	ML	Range 0 to 100	Any Network Type, except Standard Network
Spell Correction for ML	Enable to support spell correction on the ML bot dictionary while predicting.	Custom (ML)	Enable, Disable (default)	Applicable only for English, French and Dutch language VAs.
Spell Correction Version	Use this configuration to select the version of Spell Correction for the VA.	NA	Version 1, Version 2	Applicable only for English, French and Dutch language VAs. The spell-corrected input is used by all the engines for further processing
Spell Correction Status	Use this configuration to set the status of Spell Correction for the VA.	NA	Enable, Disable	Applicable only for English, French and Dutch language VAs. The spell-corrected input is used by all the engines for further processing
Intent Elimination Rules	Enable to apply prebuilt rules to eliminate intent matches.	RR	Enable (default), Disable	Applicable only for English, Spanish, French, and German language VAs.
Cosine similarity dampening	Avoid penalty on short length questions using Cosine Similarity Dampening	KG	Enable (default), Disable
FAQ Name as Intent Name	To use the Primary Question of the FAQ as the intent name even when the FAQ is linked to a Dialog	KG	Enable, Disable (default)
FAQs Order for Disambiguation	Configure the order in which the FAQs are to be presented for resolving the ambiguity	KG	Order by Hierarchy, Default Order (default)
Auto qualify FAQs from fully matched Paths	Automatically qualify all FAQs from the path if the path is fully matched even if no question from that path has matched the user query.	KG	Enable, Disable (default)
Taxonomy based KG	Enable this option if only a full match of all the terms in the path should be considered as a path qualification.	Custom (KG)	Enable, Disable (default)
Default Max. Wildcards for Intent Patterns	Use this option to define the maximum number of wildcards to be allowed by default between words for intent patterns. This does not limit you from explicitly writing patterns containing more wildcards.	FM	any number from 0-9; set to 3 by default
Default Max. Wildcards for Entity Patterns	Use this option to define the maximum number of wildcards to be allowed by default between words for entity patterns. This does not limit you from explicitly writing patterns containing more wildcards.	FM	any number from 0-5; set to 2 by default
Matching Order of Intent Patterns	Choose whether to pick the first pattern match of the intent (as per the order in which patterns are defined) or to go through all the patterns defined for the intent and find the best one.	FM	First (default), Best
Grading of Pattern Matches	Choose whether the Pattern Matches should be classified as Probable matches, based on the number of wildcards present in the user input when compared to the pattern definition.	FM	any number from 0-9; set to 3 by default
Precedence for Intents with Ambiguous Entities	Enable to Use Precedence, Intent Over Entity, or Entity Over Intent options for intents with ambiguous entities.	FM	Enable Disable (default) /td>
Prefer Only the First Pattern Match in a Sentence	Choose whether to prefer only the first pattern match from a sentence or to pick all the pattern matches in a sentence, when multiple patterns are matched in a single sentence	FM	Enable (default), Disable
Exact Task Name Match	Choose whether the system should auto-generate the strict pattern when ‘Intent Detection using Task Name Words’ is Disabled.	FM	Enable (default), Disable

[/vc_column_text]

ML Engine related settings

Split Compound Words

Compound words are formed when two or more words are joined together to create a new word that has an entirely new meaning. This is particularly the case with the German language, where two (or more) words are combined to form a compound, leading to an infinite amount of new compounds. For example, the components are connected with a transitional element, as the -er in Bilder | buch (picture book); or parts of the modifier can be deleted. For example, Kirch | turm (church tower), where the final -e of the lemma Kirche is deleted. Often the compound words mean something entirely different from the stem words. For example, Grunder (founder) with stem words grun | der (green|the). From an NLP perspective, it is important to understand when the NLP engines should split the words and process and when the entire word should be processed.

This setting is used to choose how the compound words should be processed. Once enabled, compound words present in the user utterance splits into their stem words and then considered for Intent Detection.

Split Combined Words

In Spanish, users often combine two words while writing. For example, ‘los’, ‘hombres’, and ‘como’ are combined into one word – ‘loshombrescomo’. Although this is not the correct representation, words are frequently combined this way. The combined words negatively impact the NLP performance and accuracy in identifying the right intent and entities, especially for Standard and Few-shot network types. 

When the setting is enabled, combined words in user utterances are split into valid words (maximum of three words) and then considered for Intent Detection.

None Intent

The Machine Learning (ML) engine detects and evaluates utterances from customer inputs to build and train a model. The ML model tries to classify the user input into one of the intents it has been trained for. However, when there is an out of context utterance, the ML still tries to map it to the wrong intent.

This can be avoided by enabling the None intent under Training > Thresholds & Configurations for different languages. The None intent is a placeholder that prevents the ML from identifying the wrong intent for untrained or ambiguous utterances. This flag is enabled by default for a new virtual assistant.

For e.g. there is a Airlines VA which has an Intent “Book Flight Ticket”. If a customer requests to “Book a Movie Ticket” the ML tries to associate the closest intent “Book Flight Ticket” (wrong intent) as the successful intent and triggers it. This error can be eliminated by enabling the None intent.

Adding an extra None Intent ensures classifying random input to these intents in the VA. Once enabled, the ML Model is tuned to identify these none intents when a user utterance contains the words that are not used in the VA’s training. i.e., bot vocabulary. You can define the None Intent for the Linked Bots in a Universal Bot. For more information, see Creating a Universal Bot article.

Externalization of the ML Engine

In machine learning, a hyperparameter is a parameter whose value is used to control the learning process. The hyperparameters provide you with additional customization options for your VAs. The following are the ML configurations that can be customized.

Network Type

You can choose the Neural Network that you want to use. This setting is moved to the Machine Learning section post v8.1. Refer here for details.

Epochs

In terms of artificial neural networks, an epoch refers to one cycle through the full training dataset. To get a good performance on non-training data, it usually (but not always) takes more than one pass over the training data. The number of epochs is a hyperparameter that controls the number of complete passes through the training dataset.

Batch Size

Batch size is a term used in machine learning and refers to the number of training examples utilized in one iteration. It controls the accuracy of the estimate of the error gradient when training neural networks. The batch size is a hyperparameter that controls the number of training samples to work through before the model’s internal parameters are updated.

Learning Rate

In machine learning and statistics, the learning rate is a tuning parameter in an optimization algorithm that determines the step size at each iteration while moving toward a minimum of a loss function. It can be thought of as a parameter for controlling the weight update in the neural network based on the loss.

Dropout

The term dropout refers to dropping out units (both hidden and visible) in a neural network. Simply put, dropout refers to ignoring units (i.e. neurons) during the training phase of a certain set of neurons which is chosen at random. It is a regularization technique to prevent overfitting of data.

Vectorizer

Vectorization is a way to optimize algorithms by using vector operations for computations instead of element-by-element operations. It is used to determine the feature extraction technique on training data. It can be set to one of the following:

• Count Vectorizer is used to convert the given text documents to a vector of term/token counts based on the frequency (count) of each word occurrence in the text. This is helpful when there are multiple texts, and each word in the text needs to be converted into vectors for use in further text analysis. It enables the ​pre-processing of text data prior to generating the vector representation.
• TFIDF Vectorizer is a statistical measure that evaluates how relevant a word is to a document in a collection of documents. This is done by multiplying two metrics: how many times a word appears in a document (Term Frequency TF), and the Inverse Document Frequency (IDF) of the word across a set of documents.

Maximum Sequence Length

When processing a sentence (for training or prediction), the length of the sequence is the number of words in the sentence. The maximum sequence length parameter is the maximum number of words to be considered for training. If the user input or training phrase sentence sequence length is more than the maximum sentence length it is trimmed to this length and if it is less than then the sentence is padded with special tokens.

Embeddings Type

A (word) embedding is a vector representation of a word or phrase in an input/training text. Words with similar meaning will have similar vector representations in n-dimensional space and the vector values are learned in a way that resembles a neural network.

Embeddings Type can be set to one of the following:

• Random (default setting): At first, all the words are assigned random embeddings, then the embeddings are optimized for the given training data while training.
• Generated: Word Embeddings are generated just before the training starts. Word2Vec model is used for generating word embeddings. These generated embeddings are used while training. These generated word embeddings are optimized for the given training data while training.

Embeddings Dimensions

The embedding dimension defines the size of the embedding vector. If the word embeddings are random or generated, any number can be used as an embedding dimension.

K Fold Cross-Validation

Cross-validation is a resampling procedure used to evaluate machine learning models on a limited data sample. The procedure has a single parameter called k that refers to the number of groups that a given data sample is to be split into. This setting allows you to configure the K parameter. Refer here for more on cross-validation.

Fuzzy Match

Fuzzy matching is an approximate string matching technique that helps the system identify non-exact matches. The ML Engine uses fuzzy matching logic to identify definitive matches. The fuzzy match algorithm assigns a Fuzzy Search score to the intents based on their similarity with the user utterance. Any intent with a fuzzy match score of 95 or higher (on a scale of 0-100) is identified as a definitive match.

However, fuzzy matching can produce false positives when there are words with similar spellings but different meanings. For example, possible vs. impossible or available vs. unavailable. This behavior is problematic in some cases. You can disable this option and discourage the ML engine from using this matching algorithm.

Negation Handling

This setting is configured to choose the ML engine’s behavior when negated words are present in the user utterance. When the Negation Handling configuration is enabled, the intent’s ML score would be penalized if there are any negated predilection words present in the user utterance.

Ignore Multiple Occurences

Sometimes the intent identification gets skewed if multiple occurrences of the same word are present in the user utterance. When the Ignore Multiple Occurrences configuration is enabled, then multiple occurrences of the same word present in the user utterance are discarded. The repeated word is considered only once for the vectorization and the subsequent intent matching.

Entity Placeholders in User Utterances

Sometimes you want the system to replace the entity values present in the user utterance with entity placeholders so that the intent detection can be improved. Note that the entities that are not resolved by the NER model would not be used for replacement, so if you enable this option we strongly urge that you annotate all the training utterances. These entities are replaced in user utterance in End-user interactions, Batch testing, Utterance testing, Conversation testing.

Sentence Split

If the user input has multiple sentences, multiple intent calls are made, one for each sentence. This might not be an ideal situation in some cases. For example user utterance, I want to book tickets. Redirect me to Book My Show. will result in a 0.6 ML score for I want to book tickets and Redirect me to Book My Show and the total ML score of 0.6.

Disabling this configuration sends the original user input to ML for intent identification and results in a definite score like 0.99 for the above example.

Multiple Intent Model

Enabling this feature creates multiple ML intent models for your VA. All the Primary Dialog Intents will be part of the Bot Level Intent Model. Separate Dialog Level ML Models are created for each of the other Dialog Tasks and Sub Dialog Tasks, consisting of all the sub-intents used in the respective task definition. Refer here for details.

Lemmatization of KG Synonyms

Lemmatization in linguistics is the process of grouping together the inflected forms of a word so they can be analyzed as a single item, identified by the word’s lemma, or dictionary form. Using Parts of Speech information from the user utterance in the process of lemmatization can improve identifying a more accurate FAQ.

Use only Tagged Utterances for NER Training

Enabling this flag allows only Tagged Utterances for NER Training and avoids training of utterances without NER tags. This is an accurate and time-saving approach. By default, this flag is enabled for all new VAs. If the flag is not set for a VA, then it is in the disabled state under the Advanced NLP Configurations.

Neurons in Hidden Layer

Neurons in Hidden Layer determine the intensity/rigor to be adopted while performing intent identification by the ML Model. A higher number of neurons increases the accuracy but would require a longer duration for completing the training. A lower number of neurons decreases the accuracy but would speed up the training time. By default, it is fixed as 1000. Ideally, it should be 1x times the number of intents in a VA and can go up to 2x for better accuracy. This is a general recommendation and would vary depending on the quality of the training

Softmax Temperature

Softmax temperature allows you to define how confidently the ML Engine should identify the winning intent from the ML Model. Temperature is a hyperparameter that is applied to logits (Model outputs) to affect the final probabilities from the softmax. Any value between 0 to 1 indicates that the ML Engine should identify the winning intent with lower confidence. 0 being very low confidence and 1 being regular confidence. Any value between 1 to 100 indicates that the ML Engine should associate a high amount of confidence for the winning intent. 1 being regular confidence and 100 being the higher confidence possible.

Spell Correction in ML

For VAs in the English language, spell correction does not happen on the ML bot dictionary. This might cause an issue for VAs that are heavily dependent on ML training. The issue can be rectified by enabling spell correction on the ML bot dictionary while predicting. This is achieved by adding custom config in NLP Advanced Settings.

This is a Custom configuration, to enable follow these steps:

1. Add Custom
2. Enter name as ML_spell_correction
3. Enter the value as enabled or disabled

Please note: This configuration is not available in Spell Correction Version 2 is enabled for the VA.

[/vc_column_text] [/vc_column][/vc_row]