Cause-effect sentence analysis device, cause-effect sentence analysis system, program, and cause-effect sentence analysis method

A non-transitory computer-readable medium configured to store instruction which, when executed by at least one processor, cause the at least one processor to execute:

• a cause-effect sentence extraction step of obtaining a text from a database and extracting a plurality of cause-effect sentences from the text, wherein each cause-effect sentence includes a cause expression and an effect expression;
• an importance degree calculation step of calculating a plurality of importance degrees corresponding to the plurality of cause-effect sentences, wherein the plurality of importance degrees indicate an importance of the plurality of cause-effect sentences with respect to the text;
• a selection step of selecting, from among the plurality of cause-effect sentences, a plurality of target cause-effect sentences, wherein an importance degree corresponding to each target cause-effect sentence from among the plurality of target cause-effect sentences is above an importance threshold;
• an acquisition step of acquiring information indicating a reference expression serving as a reference for analyzing a degree of similarity;
• a similarity degree analysis step of calculating, for the target each cause-effect sentence, a cause similarity degree that is the degree of similarity between the cause expression included in the target cause-effect sentence and the reference expression and an effect similarity degree that is the degree of similarity between the effect expression included in the target cause-effect sentence and the reference expression, wherein at least one of the cause similarity degree and the effect similarity degree is calculated using a similarity degree model based on a vector representation;
• and
• an output step of selecting, from among the plurality of target cause-effect sentences, a target cause-effect sentence in which one of the cause similarity degree and the effect similarity degree is above a similarity threshold, and the other of the cause similarity degree and the effect similarity degree is below the similarity threshold,
• wherein the importance degree comprises a cause importance degree corresponding to the cause expression and an effect importance degree corresponding to the effect expression, and
• wherein the importance degree is determined to be above the importance threshold based on at least one of the cause importance degree being above a cause importance threshold, and the effect importance degree being above an effect importance threshold.

A cause-effect sentence analysis method performed in a computer, comprising:

• obtaining a text from a database and extracting a plurality of cause-effect sentences from the text, wherein each cause-effect sentence includes a cause expression and an effect expression;
• calculating a plurality of importance degrees corresponding to the plurality of cause-effect sentences, wherein the plurality of importance degrees indicate an importance of the plurality of cause-effect sentences with respect to the text;
• selecting, from among the plurality of cause-effect sentences, a plurality of target cause-effect sentences, wherein an importance degree corresponding to each target cause-effect sentence from among the plurality of target cause-effect sentences is above an importance threshold;
• acquiring information indicating a reference expression serving as a reference for analyzing a degree of similarity;
• performing a similarity degree analysis for calculating, for the each target cause-effect sentence, a cause similarity degree that is the degree of similarity between the cause expression included in the target cause-effect sentence and the reference expression and an effect similarity degree that is the degree of similarity between the effect expression included in the target cause-effect sentence and the reference expression, wherein at least one of the cause similarity degree and the effect similarity degree is calculated using a similarity degree model based on a vector representation;
• and
• selecting and outputting, from among the plurality of target cause-effect sentences, a search result comprising a target cause-effect sentence in which one of the cause similarity degree and the effect similarity degree is above a similarity threshold, and the other of the cause similarity degree and the effect similarity degree is below the similarity threshold,
• wherein the importance degree comprises a cause importance degree corresponding to the cause expression and an effect importance degree corresponding to the effect expression, and
• wherein the importance degree is determined to be above the importance threshold based on at least one of the cause importance degree being above a cause importance threshold, and the effect importance degree being above an effect importance threshold.

A cause-effect sentence analysis device comprising:

• at least one processor configured to:
• obtain a text from a database and extract a plurality of cause-effect sentences from the text, wherein each cause-effect sentence includes a cause expression and an effect expression;
• acquire information indicating a reference expression serving as a reference for analyzing a degree of similarity;
• calculate, for the each cause-effect sentence, a cause similarity degree that is the degree of similarity between the cause expression included in the cause-effect sentence and the reference expression and an effect similarity degree that is the degree of similarity between the effect expression included in the cause-effect sentence and the reference expression, wherein at least one of the cause similarity degree and the effect similarity degree is calculated using a similarity degree model based on a vector representation;
• select, from among the plurality of cause-effect sentences, a cause-effect sentence in which one of the cause similarity degree and the effect similarity degree is above a similarity threshold, and the other of the cause similarity degree and the effect similarity degree is below the similarity threshold;
• assign a priority on the basis of the cause similarity degree and the effect similarity degree to the cause-effect sentence of which the cause similarity degree and the effect similarity degree are calculated by the at least one processor;
• and
• assign the priority to a cause-effect sentence having an effect similarity degree that is above the similarity threshold and a cause similarity degree that is below the similarity threshold from among the plurality of cause-effect sentences, in an ascending order of the cause similarity degree.

A cause-effect sentence analysis device comprising:

• at least one processor configured to:
• obtain a text from a database and extract a plurality of cause-effect sentences from the text, wherein each cause-effect sentence includes a cause expression and an effect expression;
• acquire information indicating a reference expression serving as a reference for analyzing a degree of similarity;
• calculate, for the each cause-effect sentence, a cause similarity degree that is the degree of similarity between the cause expression included in the cause-effect sentence and the reference expression and an effect similarity degree that is the degree of similarity between the effect expression included in the cause-effect sentence and the reference expression, wherein at least one of the cause similarity degree and the effect similarity degree is calculated using a similarity degree model based on a vector representation;
• select, from among the plurality of cause-effect sentences, a cause-effect sentence in which one of the cause similarity degree and the effect similarity degree is above a similarity threshold, and the other of the cause similarity degree and the effect similarity degree is below the similarity threshold;
• assign a priority on the basis of the cause similarity degree and the effect similarity degree to the cause-effect sentence of which the cause similarity degree and the effect similarity degree are calculated by the at least one processor;
• and
• assign the priority to a cause-effect sentence having a cause similarity degree that is above the similarity threshold and an effect similarity degree that is below the similarity threshold from among the plurality of cause-effect sentences, in an ascending order of the effect similarity degree.