CN-116186242-B - Descriptive text editing method and device

CN116186242BCN 116186242 BCN116186242 BCN 116186242BCN-116186242-B

Abstract

The embodiment of the disclosure discloses an editing method and device of a description text, wherein the method comprises the steps of obtaining a talkback set comprising a plurality of talkbacks based on historical information corresponding to a target object, extracting a plurality of sentence subsets from at least one explanation text corresponding to the target object based on the talkback points in the talkback set, determining a plurality of weight values corresponding to the talkback points based on user portraits corresponding to the target user, determining the sequence of the talkback points based on the weight values, sequencing the sentence sets corresponding to the talkback points of the target object according to the sequence to obtain the description text corresponding to the target user, achieving long text through the extracted sentence sets, determining the sequencing of the sentence subsets based on the weight values determined by the user portraits, enhancing interaction with the user, and achieving personalized generation of the corresponding description text for the target user.

Inventors

ZHENG KAIYU
SUN LIN
WANG HEQING

Assignees

如你所视(北京)科技有限公司

Dates

Publication Date: 20260505
Application Date: 20221227

Claims (8)

1. A method of editing descriptive text, comprising: Acquiring a talkback point set comprising a plurality of talkback points based on the history information corresponding to the target object; extracting a plurality of sentence subsets from at least one explanation text corresponding to the target object based on a plurality of talkpoints in the talkpoint set, wherein each sentence set corresponds to one talkpoint, each explanation text comprises a plurality of explanation sentences, and each sentence set comprises at least one explanation sentence; Determining a plurality of weight values corresponding to the plurality of talkback points based on the user portrait corresponding to the target user, and determining the sequence of the plurality of talkback points based on the weight values; According to the sequence, sorting the sentence sets corresponding to the plurality of talkback points of the target object to obtain a description text of the target user corresponding to the target object; The extracting, based on a plurality of talkback points in the talkback point set, from at least one explanation text corresponding to the target object to obtain a plurality of sentence subsets includes: Acquiring at least one explanation text corresponding to the target object; based on the speaking points corresponding to the speaking sentences included in the at least one speaking text, aggregating the plurality of speaking sentences to obtain a plurality of sets corresponding to the plurality of speaking points, wherein each set comprises at least one speaking sentence; For each set in the plurality of sets, determining each explanation sentence in the set as a first expression sentence or a second expression sentence through matching to obtain a first expression set and a second expression set, wherein the first expression set comprises at least one first expression sentence, and the second expression set comprises at least one second expression sentence; clustering the first expression set and the second expression set respectively to obtain a plurality of clustering sets, wherein each clustering set comprises at least one first expression sentence or at least one second expression sentence; And extracting an explanation sentence from each cluster set in the plurality of cluster sets respectively to obtain a plurality of sentence subsets.
2. The method according to claim 1, wherein the obtaining a set of lectures including a plurality of lectures based on the basic information corresponding to the target object includes: Performing entity identification on the historical information to determine a plurality of triples corresponding to the target object, wherein each triplet comprises an entity, an attribute corresponding to the entity and an attribute value corresponding to the attribute; determining a connection relationship among the triples based on the association relationship among the entities; And determining the talkback point set based on a plurality of triples with connection relations, wherein each entity corresponds to one talkback point.
3. The method of claim 2, further comprising, prior to determining the connection relationship between the plurality of triples based on the association relationship between the plurality of entities: processing basic information corresponding to the target object based on a preset rule to obtain a rule attribute corresponding to at least one entity included in the target object; and forming the triples with preset attribute values based on the entities of the rule attributes and the rule attributes, and obtaining at least one triplet.
4. A method according to any one of claims 1-3, wherein the aggregating the plurality of lecture sentences based on the lecture sentences corresponding to the lecture sentences included in the at least one lecture text comprises: Classifying the explanation sentences included in the explanation texts based on the plurality of the speaking points, and determining the speaking points corresponding to each explanation sentence in the at least one explanation sentence; And determining the set corresponding to each talkback point based on the talkback point corresponding to each explanation sentence.
5. A method according to any one of claims 1-3, wherein clustering the first expression set and the second expression set to obtain a plurality of cluster sets comprises: determining the trunk information of each first expression sentence through dependency syntactic analysis, and clustering the first expression sets based on the trunk information to obtain at least one clustering set, wherein each clustering set comprises the explanation sentences of at least one expression mode corresponding to the same trunk information; and carrying out vector analysis on each second expression sentence through a preset network model to obtain at least one sentence vector, and clustering the second expression set based on the at least one sentence vector to obtain at least one clustering set.
6. A method according to any one of claims 1-3, wherein the determining a plurality of weight values corresponding to the plurality of speaks based on the representation of the user corresponding to the target user, and determining an order of the plurality of speaks based on the weight values, comprises: assigning a plurality of weight values to the plurality of talkbacks based on the order of talkbacks in the history information; acquiring a user portrait corresponding to the target user, and adjusting at least one weight value in the plurality of weight values based on the user portrait; Determining an order of the plurality of talkback points based on the adjusted weight values.
7. The method of claim 6, further comprising, after determining the order of the plurality of talkpoints based on the adjusted weight values: and responding to the received input talkback sequence adjustment instruction, and adjusting the sequences corresponding to the talkback points based on the talkback sequence adjustment instruction to obtain the adjusted sequences.
8. An editing apparatus for descriptive text, comprising: A talk point obtaining module for obtaining the talk point based on the history information corresponding to the target object, obtaining a speaker set comprising a plurality of speakers; The sentence extraction module is used for extracting a plurality of sentence subsets from at least one explanation text corresponding to the target object based on a plurality of talkpoints in the talkpoint set, wherein each sentence set corresponds to one talkpoint, each explanation text comprises a plurality of explanation sentences, and each sentence set comprises at least one explanation sentence; The order determining module is used for determining a plurality of weight values corresponding to the plurality of speaking points based on the user portrait corresponding to the target user, and determining the order of the plurality of speaking points based on the weight values; the text determining module is used for sequencing the sentence sets corresponding to the plurality of speaking points of the target object according to the sequence to obtain a description text of the target user corresponding to the target object; The sentence extraction module includes: The text acquisition unit is used for acquiring at least one explanation text corresponding to the target object; The aggregation unit is used for aggregating the plurality of explanation sentences based on the speaking points corresponding to the explanation sentences included in the at least one explanation text to obtain a plurality of sets corresponding to the plurality of speaking points, wherein each set comprises at least one explanation sentence; The sentence extraction unit is used for determining that each explanation sentence in the collection is a first expression sentence or a second expression sentence according to each collection in the collection, so as to obtain a first expression set and a second expression set, wherein the first expression set comprises at least one first expression sentence, the second expression set comprises at least one second expression sentence, the first expression set and the second expression set are clustered respectively, so as to obtain a plurality of clustered collections, each clustered collection comprises at least one first expression sentence or at least one second expression sentence, and each clustered collection comprises at least one explanation sentence, so as to obtain a plurality of sentence subsets.

Description

Descriptive text editing method and device Technical Field The disclosure relates to the technical field of text generation, in particular to a descriptive text editing method and device. Background Text generation technology is applied to different professional fields, such as dialogue systems, news lecture generation and the like, and becomes one of important tasks for research in the field of natural language processing. The text generation is to make a machine learn grammar standards and expression habits from a large amount of natural language corpus, and finally automatically generate text content meeting the business target requirements. However, intelligent applications increasingly pay attention to the experience of man-machine interaction, and in order to enhance interactivity and interestingness with clients, text generation results are required to be more interactive and diverse. The generation method of the text is based on template type generation, is common in the early stage of business, falls to the ground rapidly and is suitable for short text generation. Depending on manual labeling, a great deal of rule and template knowledge needs to be built and maintained offline, and the labor cost is high. Expert experience is particularly important if it is applied in the vertical field, and these experiences are often very scarce, and it takes a lot of time and cost to develop their experience. Disclosure of Invention The present disclosure has been made in order to solve the above technical problems. The embodiment of the disclosure provides a descriptive text editing method and device. According to an aspect of the embodiments of the present disclosure, there is provided an editing method of descriptive text, including: Acquiring a talkback point set comprising a plurality of talkback points based on the history information corresponding to the target object; extracting a plurality of sentence subsets from at least one explanation text corresponding to the target object based on a plurality of talkpoints in the talkpoint set, wherein each sentence set corresponds to one talkpoint, each explanation text comprises a plurality of explanation sentences, and each sentence set comprises at least one explanation sentence; determining a plurality of weight values corresponding to the plurality of talkback points based on the user portrait corresponding to the target user, and determining the sequence of the plurality of talkback points based on the weight values; And sequencing the sentence sets corresponding to the plurality of talkback points of the target object according to the sequence to obtain the description text of the target user corresponding to the target object. Optionally, the obtaining a speaker set including a plurality of speakers based on the basic information corresponding to the target object includes: Performing entity identification on the historical information to determine a plurality of triples corresponding to the target object, wherein each triplet comprises an entity, an attribute corresponding to the entity and an attribute value corresponding to the attribute; Determining a connection relationship between the triples based on the association relationship between the entities; And determining the talkback point set based on the multiple triples with the connection relation, wherein each entity corresponds to one talkback point. Optionally, before determining the connection relationship between the triples based on the association relationship between the entities, the method further includes: processing basic information corresponding to the target object based on a preset rule to obtain a rule attribute corresponding to at least one entity included in the target object; and forming the triples with preset attribute values based on the entities of the rule attributes and the rule attributes, and obtaining at least one triplet. Optionally, the extracting, based on a plurality of talkpoints in the talkpoint set, from at least one explanation text corresponding to the target object to obtain a plurality of sentence subsets includes: Acquiring at least one explanation text corresponding to the target object; based on the speaking points corresponding to the speaking sentences included in the at least one speaking text, aggregating the plurality of speaking sentences to obtain a plurality of sets corresponding to the plurality of speaking points, wherein each set comprises at least one speaking sentence; and respectively extracting at least one explanation sentence from the plurality of sets to obtain a plurality of sentence subsets. Optionally, the aggregating the plurality of explanation sentences based on the lecture points corresponding to the explanation sentences included in the at least one explanation text includes: Classifying the explanation sentences included in the explanation texts based on the plurality of the speaking points, and determining the speaking points cor