US 2023/0067435 A1 · METHOD AND DEVICE OF SEGMENTING TOPICS OF CONTENT

Abstract

A method of segmenting topics of content according to an embodiment is configured to preprocess text data configured of content, and divide a plurality of utterances into two topic segmented bodies based on the preprocessed data. The preprocessing may be performed by processing the text data in a continuous form of the plurality of utterances, and calculating a conceptual similarity between utterances based on the processed data. The topic segmented bodies may be divided into two by calculating similarity cohesion for the two topic segmented bodies based on the conceptual similarity and a consistency metric while changing a segmentation point which distinguishes the two topic segmented bodies, and determining the segmentation point based on the similarity cohesion.

Claims20

Dependent claims

A method of segmenting topics of content, the method including:

preprocessing text data configuring content;
and
dividing a plurality of utterances into two topic segmented bodies based on the preprocessed data,
wherein the preprocessing includes:
processing the text data in a continuous form of the plurality of utterances;
and
calculating a conceptual similarity between utterances based on the processed data, and
wherein the dividing includes:
calculating similarity cohesion for the two topic segmented bodies based on the conceptual similarity and a consistency metric while changing a segmentation point which distinguishes the two topic segmented bodies;
and
determining the segmentation point based on the similarity cohesion.

8 more elements

A device of segmenting topics of content, the device including:

an input unit for receiving text data configuring content;
a processor for preprocessing the text data and dividing a plurality of utterances into two topic segmented bodies based on the preprocessed data;
and
a visualization unit for visualizing and expressing the divided topic segmented bodies,
wherein the processor is configured to:
process the text data in a continuous form of the plurality of utterances and calculate a conceptual similarity between the utterances based on the processed data;
and
calculate similarity cohesion for the two topic segmented bodies based on the conceptual similarity and a consistency metric while changing a segmentation point which distinguishes the two topic segmented bodies, and determine the segmentation point based on the similarity cohesion.

5 more elements

Dependencies are inferred from the claim text.

Method and device of segmenting topics of content

Life of this patent

Method and device of segmenting topics of content

Life of this patent

Abstract

Claims20