Categorizing a use scenario of a product

1. A computer-implemented method for categorizing a use scenario of a product, the computer-implemented comprising:

selecting, by a processor of a computer, text from purchase reviews for the product on shopping websites and contents on social media databases related to the product posted, wherein the computer is communicatively coupled to the shopping websites and the social media databases, wherein the purchase reviews and the contents describe the use scenario of the product, by:

detecting scenario information from the contents on social media,

searching the product scenario category repository for at least one product scenario category matching the detected scenario information, and

provisioning the product corresponding to the at least one product scenario category;

extracting, from the text by the processor of the computer, at least one feature tuple configured to characterize the use scenario by:

selecting via a syntax and semantic analysis on the text a target word representing the product or an event involved in the use scenario, a context word representing context of the use scenario, and a sentiment word expressing an opinion, wherein a plurality of feature words includes the target word, the context word, and the sentiment word, and

constructing the at least one feature tuple comprising the plurality of feature words;

performing, by the processor of the computer, a latent semantic association analysis on the at least one feature tuple into respective semantic categories to obtain a latent topic sequence for the use scenario by:

generalizing the plurality of feature words in the at least one feature tuple to produce generalized feature words,

constructing a contextual clue for the at least one feature tuple, wherein the contextual clue at least comprises adjacent feature tuples of the at least one feature tuple,

applying a pre-built latent semantic association model to the generalized feature words of the at least one feature tuple and a corresponding contextual clue to generate latent topics for the at least one feature tuple, and

generating the latent topic sequence for the use scenario utilizing at least the generalized feature words, the latent topic sequence comprising all the latent topics of the at least one feature tuple;

determining, by the processor of the computer, a product scenario category to which the use scenario belongs according to the latent topic sequence for the use scenario by:

calculating similarities between the latent topic sequence and pre-set tag sequences of product scenario categories,

determining the product scenario category with maximum similarity as the product scenario category to which the use scenario belongs, if the calculated maximum similarity is greater than or equal to a similarity threshold, and

generating a new product scenario category if the calculated maximum similarity is less than the similarity threshold, wherein the tag sequence of the new product scenario category is the latent topic sequence; and

recommending, by the processor of the computer, one or more subsequent products based on the new product scenario category or the product scenario category, the one or more subsequent products being filtered according to the at least one feature tuple so as to be relevant to the product.

2. A computer apparatus for categorizing a use scenario of a product, the computer apparatus comprising:

a text selection module configured to select a text from purchase reviews for the product on shopping websites and contents on social media databases related to the product posted, wherein the computer apparatus is communicatively coupled to the shopping websites and the social media databases, wherein the purchase reviews and the contents describe the use scenario of the product, by:

detecting scenario information from the contents on social media,

searching the product scenario category repository for at least one product scenario category matching the detected scenario information, and

provisioning the product corresponding to the at least one product scenario category;

an extraction module configured to extract, from the text, at least one feature tuple configured to characterize the use scenario by utilizing:

a word selection unit configured to select via a syntax and semantic analysis on the text a target word representing the product or an event involved in the use scenario, a context word representing context of the use scenario, and a sentiment word expressing an opinion, wherein a plurality of feature words includes the target word, the context word, and the sentiment word, and

construction unit configured to construct the at least one feature tuple comprising the plurality of feature words;

an analysis module configured to perform a latent semantic association analysis on the at least one feature tuple to obtain a latent topic sequence for the use scenario by utilizing:

a generalization unit configured to generalize the plurality of feature words in the at least one feature tuple into respective semantic categories to produce generalized feature words, wherein the generalized feature words comprise the respective semantic categories,

a contextual clue construction unit configured to construct a contextual clue for the at least one feature tuple, wherein the contextual clue at least comprises adjacent feature tuples of the at least one feature tuple,

a latent topic generation unit configured to apply a pre-built latent semantic association model to the at least one feature tuple and a corresponding contextual clue to generate latent topics for the at least one feature tuple, and

a latent topic sequence generation unit configured to generate the latent topic sequence utilizing at least the generalized feature words, the latent topic sequence comprising all the latent topics of the at least one feature tuple;

a product scenario category determination module configured to determine a product scenario category to which the use scenario belongs according to the latent topic sequence for the use scenario by utilizing:

a calculation unit configured to calculate similarities between the latent topic sequence and pre-set tag sequences of product scenario categories,

a determination unit configured to determine the product scenario category with maximum similarity as the product scenario category to which the use scenario belongs, if the calculated maximum similarity is greater than or equal to a similarity threshold, and

a product scenario category generation unit configured to generate a new product scenario category if the calculated maximum similarity is less than the similarity threshold, wherein the tag sequence of the new product scenario category is the latent topic sequence; and

recommending one or more subsequent products based on the new product scenario category or the product scenario category, the one or more subsequent products being filtered according to the at least one feature tuple so as to be relevant to the product.