OPINION SUBJECT EXTRACTION DEVICE, LEARNING DEVICE, AND PROGRAM

By using a trained model in the opinion analysis device to analyze the word vector sequence, detect and match the interval of the opinion object, the problem of difficulty in predicting the opinion object in the prior art is solved, and efficient extraction of the difficult-to-expected opinion object is achieved.

JP7664705B2Active Publication Date: 2025-04-18NIPPON HOSO KYOKAI
View PDF 5 Cites 0 Cited by

Patent Information

Application Number
JP2021003420
Authority / Receiving Office
JP · JP
Patent Type
Patents
Current Assignee / Owner
Filing Date
2021-01-13
Publication Date
2025-04-18
Estimated Expiration
2041-01-13

AI Technical Summary

Technical Problem

When using machine learning methods to automatically analyze opinions, it is difficult to predict opinions objects in advance, resulting in the inability to effectively extract those opinions objects that are not easily expected.

Method used

By designing a device that contains the detection of opinion target paragraphs and predicts opinion breaks, the device uses a trained model to analyze the word vector sequence in the sentence, detects the opinion object interval, and calculates the matching degree of opinion object intervals through the matching model.

Benefits of technology

The device can automatically extract object intervals related to opinions without presuming opinion objects, improving the accuracy and flexibility of opinion extraction.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 0007664705000005
    Figure 0007664705000005
  • Figure 0007664705000006
    Figure 0007664705000006
  • Figure 0007664705000007
    Figure 0007664705000007
Patent Text Reader

Abstract

To provide an opinion extracting device, etc., capable of extracting an opinion object linked with an opinion even under a circumstance in which it is difficult to assume the opinion object in advance.SOLUTION: An opinion object extracting device includes a segment detecting unit and a check-up unit. The segment detecting unit extracts, on the basis of a word vector string corresponding to a sentence, an opinion object segment using a learnt opinion object extracting model that represents the likelihood of the opinion object segment within the word vector string, and extracts an opinion stated part segment using a learnt opinion stated part extracting model representing the likelihood of the opinion stated part segment within the word vector string. The check-up unit obtains, on the basis of an opinion object segment expression based on the word vector string and the extracted opinion object segment, and an opinion stated part segment expression based on the word vector string and the extracted opinion stated part segment, the check-up level between the opinion object segment expression and the opinion stated part segment using a learnt check-up model.SELECTED DRAWING: Figure 1
Need to check novelty before this filing date? Find Prior Art

Citation Information

Patent Citations

  • Multioutput DC / DC converter

    JP1986092166A

  • antioxidant

    JP1988095287A

  • Reputation relation extracting device, its method, and program

    JP2009140048A

  • Evaluation expression extraction method, evaluation expression extraction device and evaluation expression extraction program

    JP2009187048A

  • Complaint retrieval device, complaint retrieval method, and program of the same

    JP2012174104A