Topic Identification via Supervised Sentence Grouping

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Conventional technologies for extracting topic words from text data often assign narrower topic names and meanings than actual topics, due to extracting words with high speech density per unit time, leading to inappropriate topic naming.

Innovation Solution

An information processing device that divides sentences into groups based on a supervised topic list and identifies topic names using a combination of classification, probability calculation, and similarity analysis to align with actual topic meanings.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If words with high speech density per unit time are extracted as topic words, then topic words can be extracted efficiently from text data, but the topic names become narrower and less accurate compared to actual topics

Engineering Contradiction:

Improvetopic extraction efficiencyVSAvoidtopic naming accuracy

Core Design Contradiction:

ProductivityVSMeasurement precision

Solution Approach 1:

The patent segments the text data into multiple units of time, and for each unit, extracts candidate topic words based on speech density. It then aggregates these candidates across multiple time units and selects final topic words from the aggregated results, rather than selecting from each unit independently. This segmentation and aggregation approach maintains extraction efficiency while improving topic naming accuracy.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent performs preliminary extraction of candidate topic words from multiple time units before making the final topic selection. By pre-aggregating candidates from several time units and filtering them in advance, the system prepares a refined set of candidates that better represent actual topics, thus improving accuracy without sacrificing the efficiency of the overall extraction process.

Inventive Principle:

Principle #10Preliminary action

2Device complexity

If topic words are extracted for each unit time independently, then the extraction process is simple and fast, but the topic names do not align with actual topic meanings

Engineering Contradiction:

Improveextraction process complexityVSAvoidtopic meaning alignment

Core Design Contradiction:

Device complexityVSReliability

Solution Approach 1:

The patent merges the extraction results from multiple time units by aggregating candidate topic words across units. Instead of treating each unit independently, it combines the candidates and selects final topics from the merged set, which better captures the actual topic meanings while maintaining a relatively simple extraction framework.

Inventive Principle:

Principle #5Merging (Combining)

Solution Approach 2:

The patent maintains continuous extraction across multiple time units, where the extraction process operates continuously on successive units and accumulates results. This continuous approach ensures that topic words reflect the sustained meaning of topics across time, improving alignment with actual topic meanings while keeping the process straightforward.

Inventive Principle:

Principle #20Continuity of useful action

Data Source

PatentUS11334715B2Topic-identifying information processing device, topic-identifying information processing method, and topic-identifying computer program product

Publication Date: 2022.05.17 KK TOSHIBA

AI summary

An information processing device includes a divider and an identifier. The divider divides a plurality of sentences included in a processing target sentence set into a plurality of groups based on a supervised topic list representing a list of supervised topic name that is set in advance. The identifier identifies a topic name for each of the groups based on the supervised topic list.