Method and system for matching customer queries across diverse digital channels with a response in a multichannel contact center

1 . A computerized-method for matching customer queries across diverse digital channels with a response in a multichannel contact center, said computerized-method comprising:

(i) receiving queries from customers via one or more types of digital communication channels;

during each preconfigured time-window:

(ii) converting each customer query in the received queries into a high-dimensional vector representation and scoring the customer query to yield a similarity score by operating a semantic embedder module;

(iii) retrieving a preconfigured similarity-threshold, and clustering each high-dimensional vector representation of the customer query to a cluster based on the preconfigured similarity-threshold and a related cluster centroid by operating an Artificial intelligence (AI) driven Query Aggregator (AIQA) module, creating a new cluster when no cluster has been found, and updating the related centroid of the cluster after the clustering, of the high-dimensional vector representation to yield one or more clusters of customer queries with semantic similarity;

(iv) matching each cluster of customer queries with the response based on a preconfigured threshold by operating a Cosine Vector Search Engine (CVSE),

wherein for each cluster of customer queries with semantic similarity, the CVE is configured to:

a. extract phrases from the customer queries by using keyword extractions,

b. identify an intent of the customer queries in the cluster by mapping the extracted phrases to predefined intents by operating one of: intent classification models and a lookup table;

c. fetch a corresponding response template based on the identified intent from a predefined library; and

d. perform a quality check from the corresponding response template by checking that the response template matches the identified intent and includes dynamic placeholder values;

(v) automatically generating a notification for each customer query based on the matched response by operating a notification module; and

(vi) automatically sending the notification to each customer that is associated with the customer query.

2 . The computerized-method of claim 1 , wherein said operating of the semantic embedder module comprising:

(i) receiving the customer query;

(ii) preprocessing the customer query;

(iii) classifying the customer query as one of: structured and unstructured;

(iv) converting text of the customer query that was classified as unstructured into semantic embedding by using an Artificial Intelligence (AI) model;

(v) mapping the customer query that was classified as structured to the semantic embedding;

(vi) creating the high-dimensional vector representation of the customer query based on the semantic embedding, and storing is in an embedding database, wherein said high-dimensional vector representation comprising a meaning and context of the customer query, and

(vii) comparing the created high-dimensional vector representation of the customer query to high-dimensional vector representations of customer queries stored in the embeddings database to calculate a similarity score for each customer query in the embedding database.

3 . The computerized-method of claim 2 , wherein said preprocessing comprising: tokenizing the customer query; (ii) removing stop words; and (iii) operating lemmatization to the customer query.

4 . The computerized-method of claim 2 , wherein said calculating of the similarity score for each customer query is performed according to formula (I)

similarity

⁢

score

⁢

(

v

i

,

v

j

)

=

v

i

*

v

j

/



⁢

v

i



⁢



v

j



(

I

)

whereby:

v i is the semantic embedding of the received customer query, and

v j is the semantic embedding of the customer query stored in the embedding database.

5 . The computerized-method of claim 1 , wherein said semantic embedder module uses sentence-level contextual embedding techniques.

6 . The computerized-method of claim 1 , wherein said computerized-method further comprising routing queries that were not matched with the response to a human agent for a manual intervention.

7 . The computerized-method of claim 1 , wherein the computerized-method further comprising operating a feedback loop that continuously learns from customer interactions and agent corrections and improve clustering and intent identification accuracy.

8 . A computerized-system for matching customer queries across diverse digital channels with a response in a multichannel contact center, said computerized-system comprising:

one or more processors, said one or more processors are configured to:

(i) receive queries from customers via one or more types of digital communication channels;

during each preconfigured time-window:

(ii) convert each customer query in the received queries into a high-dimensional vector representation and score the customer query to yield a similarity score by operating a semantic embedder module;

(iii) retrieve a preconfigured similarity-threshold, and cluster each high-dimension vector representation of the customer query to a cluster based on the preconfigured similarity-threshold and a related cluster centroid by operating an Artificial intelligence (AI) driven Query Aggregator (AIQA) module, create a new cluster when no cluster has been found, and update the related centroid of the cluster after the clustering, of the high-dimensional vector representation to yield one or more clusters of customer queries with semantic similarity;

(iv) match each cluster of customer queries with the response based on a preconfigured threshold by operating a Cosine Vector Search Engine (CVSE),

wherein for each cluster of customer queries with semantic similarity, said CVSE is configured to:

a. extract phrases from the customer queries by using keyword extractions,

b. identify an intent of the customer queries in the cluster by mapping the extracted phrases to predefined intents by operating one of: intent classification models and a lookup table;

c. fetch a corresponding response template based on the identified intent from a predefined library; and

d. perform a quality check from the corresponding response template by checking that the response template matches the identified intent and includes dynamic placeholder values;

(v) automatically generate a notification for each customer query based on the matched response by operating a notification module; and

(vi) automatically send the notification to each customer that is associated with the customer query.