Natural language text processing for automated product gap clustering and display

One or more computer-readable non-transitory storage media storing instructions operable when executed by one or more processors to cause performance of:

• digitally storing a plurality of digital objects comprising first type digital objects and second type digital objects, each digital object comprising an electronic digital representation of natural language text;
• generating a corpus of digital documents, each digital document of the corpus comprising:
• the electronic digital representation of natural language text of each particular second type digital object of the plurality of digital objects;
• and
• the electronic digital representation of natural language text of each first type digital object associated with that particular second type digital object;
• vectorizing each digital document of the corpus to generate a first set of vectors;
• embedding the electronic digital representation of natural language text of each first type digital object not associated with one of the second type digital objects in a vector space of the first set of vectors, thereby generating a second set of vectors, each vector of the second set of vectors being associated with the first type digital object comprising the electronic digital representation of natural language text from which it was embedded;
• determining, for each vector of the second set of vectors, a corresponding set of nearest neighbor vectors selected from the first set of vectors;
• generating recommendation data indicating one or more potential new associations between first type digital objects and second type digital objects by:
• generating, for each vector of the second set of vectors, a similarity metric between that vector and each vector of the corresponding set of nearest neighbor vectors;
• and
• comparing the generated similarity metrics to a threshold similarity metric stored in computer memory;
• generating a third set of vectors from the specific vectors of the second set of vectors that did not generate any similarity metrics above the threshold similarity metric by embedding each first type digital object associated with those specific vectors in a new vector space;
• clustering a plurality of vectors of the third set of vectors, thereby generating one or more clusters of vectors;
• generating, for each cluster of vectors, a corresponding new second type digital object;
• and
• associating, for each cluster of vectors, for each particular first type digital object from which a vector in that cluster of vectors was generated, that particular first type digital object with the corresponding new second type digital object;
• and
• displaying, in a graphical user interface on a device display, at least one of a representation of the recommendation data or a representation of one or more of the new second type digital objects.

A system comprising:

• one or more processors;
• and
• one or more computer-readable non-transitory storage media coupled to one or more of the processors and storing instructions operable when executed by one or more of the processors to cause the system to perform operations comprising:
• digitally storing a plurality of digital objects comprising first type digital objects and second type digital objects, each digital object comprising an electronic digital representation of natural language text;
• generating a corpus of digital documents, each digital document of the corpus comprising:
• the electronic digital representation of natural language text of each particular second type digital object of the plurality of digital objects;
• and
• the electronic digital representation of natural language text of each first type digital object associated with that particular second type digital object;
• vectorizing each digital document of the corpus to generate a first set of vectors;
• embedding the electronic digital representation of natural language text of each first type digital object not associated with one of the second type digital objects in a vector space of the first set of vectors, thereby generating a second set of vectors, each vector of the second set of vectors being associated with the first type digital object comprising the electronic digital representation of natural language text from which it was embedded;
• determining, for each vector of the second set of vectors, a corresponding set of nearest neighbor vectors selected from the first set of vectors;
• generating recommendation data indicating one or more potential new associations between first type digital objects and second type digital objects by:
• generating, for each vector of the second set of vectors, a similarity metric between that vector and each vector of the corresponding set of nearest neighbor vectors;
• and
• comparing the generated similarity metrics to a threshold similarity metric stored in computer memory;
• generating a third set of vectors from the specific vectors of the second set of vectors that did not generate any similarity metrics above the threshold similarity metric by embedding each first type digital object associated with those specific vectors in a new vector space;
• clustering a plurality of vectors of the third set of vectors, thereby generating one or more clusters of vectors;
• generating, for each cluster of vectors, a corresponding new second type digital object;
• and
• associating, for each cluster of vectors, for each particular first type digital object from which a vector in that cluster of vectors was generated, that particular first type digital object with the corresponding new second type digital object;
• and
• displaying, in a graphical user interface on a device display, at least one of a representation of the recommendation data or a representation of one or more of the new second type digital objects.