Understanding user product queries in real time without using any rules

1. A computer-implementable method for performing a product query operation, comprising:

receiving user input from a user, the user input comprising chat customer interaction data, the chat interaction data being received via a chat agent;

identifying keywords contained within the user input;

comparing keywords to vector representations of words and document vector representation of product specifications from a backend product repository, each vector representation of a word being associated with a particular query type, the backend product repository comprising a plurality of entries, each of the plurality of entries comprising information regarding a specific information handling system product, the specific information handling system having a specific set of components;

identifying component details based upon the comparing;

providing a list of products corresponding to the component details to the user, the list of products comprising a list of information handling system products, each of the information handling system products having an associated set of components;

performing a document vectorization operation on columns of the backend product repository, the document vectorization operation generating the document vector representation of product specifications;

performing a scoring operation on identified component details, the scoring operation using a combination of a simple keyword match score and a semantic score, the semantic score being based upon the document vectorization operation;

ranking the list of products based upon the scoring operation;

performing a natural language analysis of the user input; and,

determining whether the user input comprises a valid user query; and wherein

the scoring operation determines suitability of a product to the valid user query; and,

the ranking sorts the list of products by suitability of a product to the valid user query;

and wherein the scoring operation and ranking are performed as follows:

A={set of product components mentioned in user query}

P={P 1 , P 2 , P 3 . . . P m } where P m is product/sku in the backend product repository (with total m products)

each P containing multiple components={C 1 , C 2 , C 3 . . . , C n } with total n components

for i=1 to m

begin for i=1 to n begin component match score+=count of words that are present in both A i and P i semantic score+=cosine similarity word2vec (A i ,P i ) (for non-matching words) end total score i =component match score+semantic score

end

sort(total score, descending).

2. The method of claim 1 , wherein:

the columns of the backend product repository include a column regarding a base system identification, a column providing details of a processor corresponding to the base system, a column providing details of an operating system corresponding to the base system, a column providing details of random access memory associated with the base system and a column providing details of storage capacity associated with the base system; and,

the document vector representation of product specifications is based upon content of all of the columns of the backend product repository.

3. The method of claim 1 , further comprising:

generating clusters of keywords within the vector representations of words; and

wherein

the identifying identifies a cluster of keywords from the clusters of keywords, the cluster of keywords corresponding to the particular query type.

4. The method of claim 3 , wherein:

the clusters of keywords are generated using a neural network model, the neural network model comprising a word2vec model.

5. A system comprising:

a processor;

a data bus coupled to the processor; and

a non-transitory, computer-readable storage medium embodying computer program code, the non-transitory, computer-readable storage medium being coupled to the data bus, the computer program code interacting with a plurality of computer operations and comprising instructions executable by the processor and configured for:

receiving user input from a user, the user input comprising chat customer interaction data, the chat interaction data being received via a chat agent;

identifying keywords contained within the user input;

comparing keywords to vector representations of words and document vector representation of product specifications from a backend product repository, each vector representation of a word being associated with a particular query type, the backend product repository comprising a plurality of entries, each of the plurality of entries comprising information regarding a specific information handling system product, the specific information handling system having a specific set of components;

identifying component details based upon the comparing;

providing a list of products corresponding to the component details to the user, the list of products comprising a list of information handling system products, each of the information handling system products having an associated set of components;

performing a document vectorization operation on columns of the backend product repository, the document vectorization operation generating the document vector representation of product specifications

performing a scoring operation on identified component details, the scoring operation using a combination of a simple keyword match score and a semantic score, the semantic score being based upon the document vectorization operation;

ranking the list of products based upon the scoring operation;

performing a natural language analysis of the user input; and,

determining whether the user input comprises a valid user query; and wherein

the scoring operation determines suitability of a product to the valid user query; and

the ranking sorts the list of products by suitability of a product to the valid user query; and wherein

the scoring operation and ranking are performed as follows:

A={set of product components mentioned in user query}

P={P 1 , P 2 , P 3 . . . P m } where P m is product/sku in the backend product repository (with total m products)

each P containing multiple components={C 1 , C 2 , C 3 . . . , C n } with total n components

for i=1 to m

begin for i=1 to n begin component match score+=count of words that are present in both A i and P i semantic score+=cosine similarity word2vec (A i ,P i ) (for non-matching words) end total score, =component match score+semantic score

end

sort(total score, descending).

6. The system of claim 5 , wherein:

the columns of the backend product repository include a column regarding a base system identification, a column providing details of a processor corresponding to the base system, a column providing details of an operating system corresponding to the base system a column providing details of random access memory associated with the base system and a column providing details of storage capacity associated with the base system; and,

the document vector representation of product specifications is based upon content of all of the columns of the backend product repository.

7. The system of claim 5 , wherein the instructions executable by the processor are further configured for:

generating clusters of keywords within the vector representations of words; and

wherein

the identifying identifies a cluster of keywords from the clusters of keywords, the cluster of keywords corresponding to the particular query type.

8. The system of claim 7 , wherein:

the clusters of keywords are generated using a neural network model, the neural network model comprising a word2vec model.

9. A non-transitory, computer-readable storage medium embodying computer program code, the computer program code comprising computer executable instructions configured for:

receiving user input from a user, the user input comprising chat customer interaction data, the chat interaction data being received via a chat agent;

identifying keywords contained within the user input;

comparing keywords to vector representations of words and document vector representation of product specifications from a backend product repository, each vector representation of a word being associated with a particular query type, the backend product repository comprising a plurality of entries, each of the plurality of entries comprising information regarding a specific information handling system product, the specific information handling system having a specific set of components;

identifying component details based upon the comparing;

providing a list of products corresponding to the component details to the user, the list of products comprising a list of information handling system products, each of the information handling system products having an associated set of components; and,

performing a document vectorization operation on columns of the backend product repository, the document vectorization operation generating the document vector representation of product specifications

performing a scoring operation on identified component details, the scoring operation using a combination of a simple keyword match score and a semantic score, the semantic score being based upon the document vectorization operation;

ranking the list of products based upon the scoring operation;

performing a natural language analysis of the user input; and,

determining whether the user input comprises a valid user query; and wherein

the scoring operation determines suitability of a product to the valid user query; and,

the ranking sorts the list of products by suitability of a product to the valid user query;

and wherein

the scoring operation and ranking are performed as follows:

A={set of product components mentioned in user query}

P={P 1 , P 2 , P 3 . . . P m } where P m is product/sku in the backend product repository (with total m products)

each P containing multiple components={C 1 , C 2 , C 3 . . . , C n } with total n components

for i=1 to m

begin for i=1 to n begin component match score+=count of words that are present in both A i and P i semantic score+=cosine similarity word2vec (A i ,P i ) (for non-matching words) end total score, =component match score+semantic score

end

sort(total score, descending).

10. The non-transitory, computer-readable storage medium of claim 9 , wherein:

the columns of the backend product repository include a column regarding a base system identification, a column providing details of a processor corresponding to the base system, a column providing details of an operating system corresponding to the base system a column providing details of random access memory associated with the base system and a column providing details of storage capacity associated with the base system; and,

the document vector representation of product specifications is based upon content of all of the columns of the backend product repository.

11. The non-transitory, computer-readable storage medium of claim 9 , wherein the computer executable instructions are further configured for:

generating clusters of keywords within the vector representations of words; and

wherein

the identifying identifies a cluster of keywords from the clusters of keywords, the cluster of keywords corresponding to the particular query type.

12. The non-transitory, computer-readable storage medium of claim 11 , wherein:

the clusters of keywords are generated using a neural network model, the neural network model comprising a word2vec model.

13. The non-transitory, computer-readable storage medium of claim 9 , wherein:

the computer executable instructions are deployable to a client system from a server system at a remote location.

14. The non-transitory, computer-readable storage medium of claim 9 , wherein:

the computer executable instructions are provided by a service provider to a user on an on-demand basis.