APPARATUS AND METHOD FOR LEARNING NARRATIVE OF DOCUMENT, AND APPARATUS AND METHOD FOR GENERATING NARRATIVE OF DOCUMENT

• US 20180189387A1
• Filed: 12/22/2017
• Published: 07/05/2018
• Est. Priority Date: 12/29/2016
• Status: Active Grant

• Current Assignee: NCSoft Corporation
• Original Assignee: NCSoft Corporation
• Granted Patent: US 11,210,328 B2
• Time in Patent Office: Days
• CPC Class Codes: G06F 16/35 Clustering; Classification
• CPC Class Codes: G06F 16/358 Browsing; Visualisation therefor
• CPC Class Codes: G06F 16/381 using identifiers, e.g. barcodes, RFIDs for URLs G06F16/9554
• CPC Class Codes: G06F 16/93 Document management systems
• CPC Class Codes: G06F 40/205 Parsing
• CPC Class Codes: G06F 40/30 Semantic analysis

Abstract

Disclosed are an apparatus and method for learning a narrative of a document, and an apparatus and method for generating a narrative of a document. According to an embodiment of the present disclosure, the narrative learning method includes the steps of receiving a plurality of documents, generating a topic database which includes one or more topics and words related to each of the one or more topics from the plurality of documents, splitting each of the plurality of documents into one or more segments including one or more sentences by using the topic database, grouping the segments split from each of the plurality of documents into one or more clusters, and generating a cluster label for each of the one or more clusters.

Claims

1. A method for learning a narrative of a document performed in a computing device comprising one or more processors and a memory storing one or more programs to be executed by the one or more processors, the method comprising:

receiving a plurality of documents;

generating a topic database which comprises one or more topics and words related to each of the one or more topics from the plurality of documents;

splitting each of the plurality of documents into one or more segments comprising one or more sentences by using the topic database;

grouping the segments split from each of the plurality of documents into one or more clusters; and

generating a cluster label for each of the one or more clusters.

2. The method of claim 1, wherein the generating of the topic database comprises generating the topic database by using a topic modeling technique.

3. The method of claim 1, wherein the splitting comprises:

generating a topic vector for each sentence included in each of the plurality of documents by using the topic database; and

splitting each of the plurality of documents into the one or more segments by using the topic vector.

4. The method of claim 3, wherein the generating of the topic vector comprises generating the topic vector by using an appearance frequency of words that belong to each of the one or more topics in each of the sentences.

5. The method of claim 3, wherein the splitting comprises splitting the plurality of documents into one or more segments based on a topic vector similarity between adjacent sentences in each of the plurality of documents.

6. The method of claim 1, wherein the grouping comprises:

generating a segment vector for each of the split segments; and

grouping the split segments into the one or more clusters by using the segment vector.

7. The method of claim 6, wherein the grouping comprises grouping the split segments into one or more clusters based on a segment vector similarity between the split segments.

8. The method of claim 1, wherein the generating of the cluster label comprises generating the cluster label for each of the grouped clusters from the segments included in each of the grouped clusters.

9. A method for generating a narrative of a document performed in a computing device comprising one or more processors and a memory storing one or more programs to be executed by the one or more processors, the method comprising:

receiving a document;

splitting the document into one or more segments comprising one or more sentences by using a topic database which comprises one or more topics and words related to each of the one or more topics;

grouping each of the one or more segments into one of one or more clusters by using a cluster database which comprises at least one of a cluster label for each of the one or more clusters and segments previously collected and grouped into the one or more clusters; and

generating a label sequence for the one or more segments in accordance with the cluster labels for the grouped clusters.

10. The method of claim 9, wherein the splitting comprises:

generating a topic vector for each sentence included in the document by using the topic database; and

splitting the document into the one or more segments by using the topic vector.

11. The method of claim 10, wherein the generating of the topic vector comprises generating the topic vector by using an appearance frequency of words that belong to each of the one or more topics in each of the sentences.

12. The method of claim 10, wherein the splitting comprises splitting the document into one or more segments based on a topic vector similarity between adjacent sentences in the document.

13. The method of claim 9, wherein the grouping comprises:

generating a segment vector for each of the one or more segments; and

grouping each of the one or more segments into the one or more clusters by using the segment vector.

14. The method of claim 13, wherein the grouping comprises grouping each of the one or more segments into one or more clusters based on a segment vector similarity between each of the one or more segments and the previously collected segments.

15. An apparatus for learning a narrative of a document, the apparatus comprising:

an inputter configured to receive a plurality of documents;

a topic parser configured to generate a topic database comprising one or more topics and words related to each of the one or more topics from the plurality of documents;

a topic storage configured to store the topic database;

a splitter configured to split each of the plurality of documents into one or more segments comprising one or more sentences by using the topic database;

a clusterer configured to group the segments split from each of the plurality of documents into one or more clusters;

a labeler configured to generate a cluster label for each of the one or more clusters; and

a cluster storage configured to store a cluster database comprising at least one of the cluster label for each of the one or more clusters and the segments grouped into the one or more clusters.

16. The apparatus of claim 15, wherein the topic parser is further configured to generate the topic database by using a topic modeling technique.

17. The apparatus of claim 15, wherein the splitter is further configured to generate a topic vector for each of the sentences included in each of the plurality of documents by using the topic database and split each of the plurality of documents into the one or more segments by using the topic vector.

18. The apparatus of claim 17, wherein the splitter is further configured to generate the topic vector by using an appearance frequency of words that belong to each of the one or more topics in each of the sentences.

19. The apparatus of claim 17, wherein the splitter is further configured to split the plurality of documents into one or more segments based on a topic vector similarity between adjacent sentences in each of the plurality of documents.

20. The apparatus of claim 15, wherein the clusterer is further configured to generate a segment vector for each of the split segments and group the split segments into the one or more clusters by using the segment vector.

21. The apparatus of claim 20, wherein the clusterer is further configured to group the split segments into one or more clusters based on segment vector similarity between the split segments.

22. The apparatus of claim 15, wherein the labeler is further configured to generate the cluster label for each of the grouped clusters from the segments included in each of the grouped clusters.

23. An apparatus for generating a narrative of a document, the apparatus comprising:

an inputter configured to receive a document;

a topic storage configured to store a topic database comprising one or more topics and words related to each of the one or more topics;

a splitter configured to split the document into one or more segments comprising one or more sentences by using the topic database;

a cluster storage configured to store a cluster database comprising at least one of a cluster label for each of one or more clusters and segments previously collected and grouped into the one or more clusters;

a clusterer configured to group the one or more segments into the one or more clusters; and

a narrative generator configured to generate a label sequence for the one or more segments in accordance with the cluster labels for the grouped clusters.

24. The apparatus of claim 23, wherein the splitter is further configured to generate a topic vector for each sentence included in the document by using the topic database and split the document into the one or more segments by using the topic vector.

25. The apparatus of claim 24, wherein the splitter is further configured to generate the topic vector by using an appearance frequency of words that belong to each of the one or more topics in each of the sentences.

26. The apparatus of claim 24, wherein the splitter is further configured to split the document into one or more segments based on a topic vector similarity between adjacent sentences in the document.

27. The apparatus of claim 23, wherein the clusterer is further configured to generate a segment vector for each of the one or more segments and group each of the one or more segments into one or more clusters based on the segment vector.

28. The apparatus of claim 27, wherein the clusterer is further configured to group each of the one or more segments into one or more clusters based on a segment vector similarity between each of the one or more segments and the previously collected segments.