Method and system for text classification
AssigneeAlibaba Group Holding Limited
InventorXiang Sun
Life of this patent
Estimating the term…
Abstract
The present disclosure discloses simplified text classification with improved classification accuracy. The content of receive text is segmented to obtain multiple words. A sum of word vectors of the multiple words at a first level vector space of a vector space group is determined. A vector space in the vector space group is composed of one or more direct child classes that are non-leaf child classes in a tree class structure. The text is classified into a class among various class of the first level vector space that has a shortest distance to the sum of the word vectors.
Claims18
A computer-implemented method for text classification comprising:
- segmenting content of a text into multiple words;
- determining a sum of a plurality of word vectors of an individual word of the multiple words at a first level vector space of a vector space group that includes a plurality of vector spaces, at least one of the plurality of vector spaces comprising one or more direct child classes that are non-leaf child classes in a tree class structure, the first level vector space having a plurality of classes, an individual word vector of the plurality word vectors including word frequency vectors of a corresponding word of the multiple words at multiple classes of the vector space;
- and
- classifying the text into a class of the plurality of classes that has a shortest distance to the sum of the plurality of word vectors of the word vectors.
An apparatus for text classification, comprising:
- a segmentation module that segments content of text into multiple words;
- a calculation module that determines a sum of word vectors of an individual word of the multiple words at a first level vector space of a vector space group that includes a plurality of vector spaces, at least one of the plurality of vector spaces comprising one or more direct child classes that are non-leaf child classes in a tree class structure, the first level vector space having a plurality of classes, the first level vector space including multiple classes that are in a top level of the tree class structure;
- and
- a classification module that classifies the text into a class of the plurality of classes that has a shortest distance to the sum of the word vectors.
A computer-implemented method for text classification, the method comprising:
- receiving a text for classification into a plurality of classes, at least one of the plurality of classes including one or more direct child classes that are non-leaf child classes in a tree class structure;
- segmenting content of the received text into multiple words, each of the multiple word having corresponding word frequency vectors;
- determining a value for each of the multiple words at individual classes of the plurality of classes using the corresponding word frequency vectors of each of the multiple words;
- determining a sum of the corresponding word frequency vectors for each of the multiple words that has the value higher than a pre-defined filtering threshold;
- and
- classifying the text into a class of the plurality of classes that is associated with a sum of the corresponding word frequency vectors that has a highest value.
Dependencies are inferred from the claim text.