Systems and methods to identify products from verbal utterances

A method of enabling retail product ordering comprising:

• receiving an audible utterance of a request comprising a series of request words;
• obtaining a plurality of search words corresponding to the series of request words by applying a tokenizer model to the series of request words, the tokenizer model trained based on training utterance words;
• generating features of the plurality of search words by applying a series of featurizer models to the set of individual search words, the series of featurizer models trained based on set of predefined product identifiers corresponding to a respective single product;
• obtaining a plurality of requested product entities based on the features of the plurality of search words by applying a classifier and extractor model, the classifier and extractor model trained based on the sets of predefined product identifiers comprising product terms, each product term having position labels, wherein each requested product entity of the plurality of requested product entities comprises a respective sub-set of position labeled product terms ordered in accordance with corresponding position labels;
• and
• generating a recommendation listing comprising recommended products, the recommended products corresponding to the plurality of requested product entities.

A non-transitory computer-readable media storing a plurality of instructions, which, when executed by a processor, cause the processor to perform operations comprising:

• receiving an audible utterance of a request comprising a series of request words;
• obtaining a plurality of search words corresponding to the series of request words by applying a tokenizer model to the series of request words, the tokenizer model trained based on training utterance words;
• generating features of the plurality of search words by applying a series of featurizer models to the set of individual search words, the series of featurizer models trained based on sets of predefined product identifiers corresponding to a respective single product;
• obtaining a plurality of requested product entities based on the features of the plurality of search words by applying a classifier and extractor model, the classifier and extractor model trained based on the sets of predefined product identifiers comprising product terms, each product term having position labels, wherein each requested product entity of the plurality of requested product entities comprises a respective sub-set of position labeled product terms ordered in accordance with corresponding position labels;
• and
• generating a recommendation listing comprising recommended products, the recommended products corresponding to the plurality of requested product entities.