CN114661893A - Abstract generation method and device - Google Patents

Abstract

The present application provides a method and device for generating an abstract, including: determining the keywords in each dialogue sentence in the dialogue text; determining the correlation between different keywords and the appearance order of the keywords in the dialogue sentences; based on the appearance order , determine the degree of influence of keywords on the dialogue text; combine the correlation between different keywords and the degree of influence of keywords on the dialogue text to determine the degree of importance of the keywords; based on the importance of the keywords, generate a summary of the dialogue text, The abstract includes: dialogue sentences in the dialogue text where the keywords whose importance levels meet the conditions are located. The solution of the present application can generate summaries of dialogue texts more efficiently and accurately.

Description

Abstract generating method and apparatus

technical field

The present application relates to the technical field of text processing, and in particular, to a method and apparatus for generating abstracts.

Background technique

The dialogue text refers to a text composed of dialogue sentences chatting or communicating between at least two dialogue parties.

In order to understand the content of the dialogue text more conveniently and efficiently, it is often necessary to generate a summary of the dialogue text. For example, in the customer service scenario, the user and the customer service (such as human customer service or robot customer service, etc.) can exchange information such as question consultation through dialogue, and by extracting the summary of the dialogue text between the user and the customer service, it is possible to Concentrating the interactive information between the user and the customer service is helpful for the relevant personnel in the customer service to obtain the general content of the discussed problems or solutions.

In order to more accurately reflect the content of the dialogue text, it is necessary to extract the abstract of the dialogue text reasonably. Therefore, how to extract the abstract from the dialogue text more reasonably so that the extracted abstract can more accurately reflect the content of the dialogue text It is a technical problem that those skilled in the art urgently need to solve.

SUMMARY OF THE INVENTION

The present application provides an abstract generating method and apparatus.

Among them, a summary generation method includes:

Determine the keywords in each dialogue sentence in the dialogue text;

determining the degree of relevancy between different keywords and the order in which the keywords appear in the dialogue sentence;

determining the degree of influence of the keyword on the dialogue text based on the order of appearance;

Determine the importance of the keyword in combination with the correlation between different keywords and the degree of influence of the keyword on the dialogue text;

Based on the importance of the keywords, a summary of the dialogue text is generated, and the summary includes: dialogue sentences in the dialogue text where the keywords whose importance levels meet the conditions are located.

In a possible implementation manner, before generating the summary of the dialogue text, the method further includes:

Dividing the dialogue text into at least one dialogue partition, and dialogue sentences in different dialogue partitions represent different types of dialogue intentions;

The generating a summary of the dialogue text based on the importance of the keyword, including:

Based on the importance of the keywords, the target dialogue sentences used to form the abstract in the dialogue partition are determined, and the abstract formed by the target dialogue sentences in each dialogue partition is obtained.

In a possible implementation manner, determining the appearance order of keywords in the dialogue sentence includes:

For each keyword, determine each co-occurrence keyword of the keyword and the co-occurrence dialogue sentence in which the keyword and its co-occurrence keyword co-occur for the first time, and the co-occurrence keyword of the keyword is the same as the keyword Words appearing simultaneously with other keywords within a dialogue sentence;

An order of appearance between the keyword and its co-occurrence keyword within the co-occurrence dialogue sentence is determined.

Among them, an abstract generating device, including:

a keyword determination unit, used for determining keywords in each dialogue sentence in the dialogue text;

an association determination unit, configured to determine the degree of relevancy between different keywords and the order of appearance of the keywords in the dialogue sentence;

an influence determination unit, configured to determine the degree of influence of the keyword on the dialogue text based on the appearance order;

an importance determination unit, configured to determine the importance of the keyword in combination with the degree of correlation between different keywords and the degree of influence of the keyword on the dialogue text;

The abstract generating unit is configured to generate an abstract of the dialogue text based on the importance of the keywords, where the abstract includes: dialogue sentences in the dialogue text where the keywords whose importance meets the condition are located.

It can be seen from the above that in the present application, after the keywords of each dialogue sentence in the dialogue text are determined, not only the correlation between different keywords is determined, but also the appearance order of the keywords in the dialogue sentences is used to determine the keywords. The degree of influence on the dialogue text. On this basis, this application will combine the two dimensions of the correlation between keywords and the degree of influence of keywords on the dialogue text, and comprehensively determine the importance of keywords, so that the importance of keywords can be more reasonable and accurate. It reflects the importance of the keyword in the dialogue text. Therefore, combining the importance of the keyword can more reasonably determine the dialogue sentence used to generate the summary from the dialogue text, so that the summary can more accurately reflect the dialogue. key information in the text.

Description of drawings

In order to explain the technical solutions in the embodiments of the present application more clearly, the following briefly introduces the drawings that are used in the description of the embodiments. Obviously, the drawings in the following description are only the embodiments of the present application. For those of ordinary skill in the art, other drawings can also be obtained according to the provided drawings without any creative effort.

1 shows a schematic flowchart of a method for generating an abstract provided by an embodiment of the present application;

FIG. 2 shows another schematic flowchart of the abstract generation method provided by the embodiment of the present application;

FIG. 3 shows a schematic diagram of a directed graph provided by an embodiment of the present application;

FIG. 4 shows a schematic flowchart of determining the degree of correlation between keywords in an embodiment of the present application;

5 shows a schematic flowchart of determining the importance of keywords in the embodiment of the present application;

FIG. 6 shows another schematic flowchart of the abstract generation method provided by the embodiment of the present application;

FIG. 7 shows a schematic diagram of a dialogue partition divided by dialogue text in an embodiment of the present application;

FIG. 8 shows another schematic flowchart of the method for generating an abstract provided by an embodiment of the present application;

FIG. 9 shows another schematic diagram of a directed graph provided by an embodiment of the present application;

FIG. 10 shows a schematic diagram of a composition structure of an abstract generating apparatus provided by an embodiment of the present application;

FIG. 11 shows a schematic diagram of a composition structure of an electronic device provided by an embodiment of the present application.

Detailed ways

The solutions of the embodiments of the present application are suitable for generating summaries for any type of dialog text, so that the generated summaries can more reasonably and accurately reflect the key information of the dialog text.

The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, but not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the protection scope of the present application.

As shown in FIG. 1 , it shows a schematic flowchart of a method for generating an abstract provided in an embodiment of the present application. The method in this embodiment can be applied to any type of electronic device. Or a cluster or a distributed system composed of multiple electronic devices, for example, the electronic device may be a notebook computer, a desktop computer, or a server, etc., which is not limited.

The method of this embodiment may include:

S101: Determine the keywords in each dialogue sentence in the dialogue text.

Wherein, the dialogue text includes at least one dialogue sentence. Of course, in general, the dialogue text that needs to generate a summary will include multiple dialogue sentences. For example, the dialogue sentences in the dialogue text may be text sentences used for information consultation, solution or communication between different dialogue parties.

In this application, the dialogue text for which the abstract needs to be extracted may be dialogue text in any scenario. For example, the dialogue text may be the dialogue text of the interaction between the user and the customer service in the customer service, for example, the text including a plurality of dialogue sentences of the interaction between the buyer user and the seller customer service. For another example, the dialogue text may be a dialogue text of a unit or enterprise whose content involves communication related to work affairs.

Of course, the dialogue text may also have other possibilities, and this application does not limit the source and acquisition method of the dialogue text.

Wherein, a dialogue sentence may include at least one keyword.

The keywords in the dialogue sentences can be determined in a number of ways without limitation.

For example, a relatively simple way may be to perform word segmentation on the dialogue sentence, and use each segmented word from the dialogue sentence as a keyword of the dialogue sentence.

For another example, after segmenting the dialogue sentence, the remaining words after removing some stop words from each segmented word may be determined as the keywords of the dialogue sentence.

Of course, words with specific semantics in the dialogue sentence may also be used as keywords according to the specific scene involved in the dialogue text, etc., which will not be repeated here.

S102: Determine the degree of relevance between different keywords and the appearance order of the keywords in the dialogue sentence.

The correlation between different keywords in the text dialogue may include: the correlation between any two keywords in the dialogue text.

Among them, the degree of correlation between keywords only reflects the degree of correlation between two keywords. It can be understood that, there may be various ways to calculate the degree of correlation between two keywords, which is not limited in this application.

For a dialogue sentence, the appearance order between the keywords in the dialogue sentence may include: the appearance order between any two keywords in the dialogue sentence. For example, if the dialogue sentence includes the keyword A and the keyword B, the sequence in which the keyword A and the keyword B appear in the dialogue sentence can be determined.

It can be understood that the appearance order of two keywords in the same dialogue sentence also reflects a relationship between the two keywords, and the appearance order of the two keywords in the same dialogue sentence can be represented. Find out the relationship between the two keywords.

For example, if keyword A appears before keyword B in a dialogue sentence, it means that keyword B may be a keyword derived from keyword A, and keyword B may appear only because of the appearance of keyword A.

S103, based on the order of appearance, determine the degree of influence of the keyword on the dialogue text.

It can be understood that, for a keyword, the keyword may appear in multiple dialogue sentences, and combining the appearance order of the keywords in different dialogue sentences, the relationship between the keyword and other keywords can be obtained. The order in which the keywords appear, the influence relationship of the keyword on other keywords, and the influence relationship of other keywords on the keyword can be obtained.

And through one or both of the influence relationship of the keyword on other keywords and the influence relationship of other keywords on the keyword, the degree of influence of the keyword in the dialogue text can be reflected.

For example, considering that other keywords appearing after the keyword in the same dialogue sentence can be considered as other keywords caused by the keyword. For a keyword, the greater the number of other keywords elicited by the keyword, or the greater the number of other keywords eliciting the keyword, the more important the keyword is in the dialogue text. influence, the greater the degree of influence.

Based on this, in a possible implementation manner, for a keyword, the first number of other keywords located after the keyword and the number of other keywords located before the keyword are determined in combination with the appearance order of the keywords in different dialogue texts The second number of other keywords. Correspondingly, the degree of influence of the keyword on the dialogue text may be determined based on the first quantity and the second quantity corresponding to the keyword.

For example, the degree of influence of the keyword on the dialogue text may be determined based on the sum of the first quantity and the second quantity. Wherein, the greater the sum of the first quantity and the second quantity corresponding to the keyword, the greater the influence of the keyword on the dialogue text.

For another example, the degree of influence of the keyword on the dialogue text may be determined based on the larger of the first number and the second number. Wherein, the larger the value of the larger of the first quantity and the second quantity corresponding to the keyword, the greater the influence of the keyword on the dialogue text. For example, the larger value of the first quantity and the second quantity may be used as the influence value representing the influence degree of the keyword on the dialogue text.

It can be understood that, in the present application, the degree of influence of keywords can be expressed in the form of scores or influence grades, and the specific manner is not limited.

S104 , determining the importance of the keywords in combination with the correlation between different keywords and the degree of influence of the keywords on the dialogue text.

It is understandable that for a keyword, the correlation between the keyword and other keywords in the dialogue text can reflect the importance of the keyword in the dialogue text; The degree of influence of the keyword on the dialogue text obtained from the order of appearance of the words can also reflect the importance of the keyword from another dimension. Based on this, the present application comprehensively determines the importance of the keyword in combination with the correlation between keywords and the degree of influence of the keyword on the dialogue text, so that the importance of each keyword in the dialogue text can be more accurately and reasonably reflected sex.

There are various possibilities for determining the degree of importance of the keyword in combination with the degree of relevance and the degree of influence of the keyword on the dialogue text, which is not limited in this application.

For example, the first importance level value of the keyword may be determined based on the correlation between the keyword and other keywords; the second importance level value may be determined in combination with the degree of influence of the keyword on the dialogue text. Then, based on the weighted sum or product or the like of the first importance level value and the second importance level value, the importance level of the keyword is determined.

S105, based on the importance of the keywords, generate a summary of the dialogue text.

Wherein, the abstract includes: dialogue sentences in the dialogue text where the keywords whose importance levels meet the conditions are located.

For example, in a possible implementation manner, after determining the importance of each keyword in the dialogue text, for each dialogue sentence in the dialogue text, the dialogue sentence can be determined by combining the importance of the keywords in the dialogue sentence degree of importance. Correspondingly, at least one preceding dialogue sentence with a high degree of importance in the dialogue text may be selected as the summary of the dialogue text.

There are also various ways to determine the importance of the dialogue sentence in combination with the importance of the keywords in the dialogue sentence. For example, the importance of the dialogue sentence can be determined by combining the importance of all keywords in the dialogue sentence.

In an implementation manner, in order to enable a keyword with a high degree of importance to appear in the abstract, the present application may also use the degree of importance of the keyword with the highest degree of importance in the dialogue sentence as the degree of importance of the dialogue sentence. For example, if the importance of keywords is represented by the importance score, then the importance score corresponding to the keyword with the highest importance score in the dialogue sentence can be determined as the score of the dialogue sentence, so that the dialogue sentence with a higher score can be selected to generate A summary of the dialogue text.

It can be seen from the above that in the present application, after the keywords of each dialogue sentence in the dialogue text are determined, not only the correlation between different keywords is determined, but also the appearance order of the keywords in the dialogue sentences is used to determine the keywords. The degree of influence on the dialogue text. On this basis, this application will combine the two dimensions of the correlation between keywords and the degree of influence of keywords on the dialogue text, and comprehensively determine the importance of keywords, so that the importance of keywords can be more reasonable and accurate. It reflects the importance of the keyword in the dialogue text. Therefore, combining the importance of the keyword can more reasonably determine the dialogue sentence used to generate the summary from the dialogue text, so that the summary can more accurately reflect the dialogue. key information in the text.

It is understandable that for any two keywords, these two keywords may appear in multiple dialogue sentences at the same time, and the order of appearance of these two keywords in different dialogue sentences may be different. different. In order to avoid inconsistency in the order in which the two keywords appear in different conversational sentences, which may affect the degree of influence of the keywords on the dialogue text, this application can only focus on the dialogue sentences in which the two keywords appear at the same time for the first time. the order in which the words appear.

It is understandable that, in practical applications, for any two keywords, the appearance order of the two keywords in the dialogue sentence in which the two keywords appear at the same time for the first time can better characterize the relationship between the two keywords. degree of importance.

For ease of understanding, the solutions of the embodiments of the present application will be described below with reference to a possible implementation manner of determining the order of appearance. As shown in FIG. 2 , which shows another schematic flowchart of the method for generating an abstract provided in an embodiment of the present application, the method in this embodiment may include:

S201: Determine the keywords in each dialogue sentence in the dialogue text.

S202, determining the degree of correlation between different keywords.

For the above steps S201 and S202, reference may be made to the relevant introduction of the previous embodiment, and details are not repeated here.

S203 , for each keyword, determine each co-occurrence keyword of the keyword and a co-occurrence dialogue sentence in which the keyword and its co-occurrence keyword co-occur for the first time.

Among them, the co-occurrence keyword of a keyword is other keywords that appear in a dialogue sentence at the same time as the keyword. That is to say, for each keyword appearing in the same dialogue sentence, any two keywords are co-occurrence keywords with each other.

For example, if the dialogue sentence includes keyword A and keyword B, then keyword B is a co-occurrence keyword of keyword A, and keyword B is also a co-occurrence keyword of keyword A.

It can be understood that, for any two keywords that are keywords for each other, the dialogue sentence in which the two keywords appear at the same time is also the dialogue sentence containing the two keywords at the same time. Dialogue sentences in which keywords appear at the same time are called co-occurring dialogue sentences. In practical applications, these two keywords may co-occur in multiple dialogue sentences at the same time. However, in order to obtain the influence relationship between these two keywords more accurately, this embodiment may only focus on these two keywords. A co-occurring dialogue sentence in which each keyword co-occurs for the first time.

S204 , for any co-occurrence keyword corresponding to the keyword, determine the appearance order between the keyword and the co-occurrence keyword in the co-occurrence dialogue sentence corresponding to the keyword and the co-occurrence keyword.

It should be noted that the co-occurrence dialogue sentence mentioned here also refers to a co-occurrence dialogue sentence in which a keyword and its co-occurrence keyword appear at the same time for the first time.

It can be understood that, since the two keywords are co-occurring keywords, for any two keywords, it is only necessary to determine the two keywords in the dialogue sentence in which the two keywords appear at the same time for the first time. The order in which the words appear. Based on this, step S204 can also be regarded as, for any two keywords that are mutually co-occurring keywords, determining the order in which the two keywords appear in the co-occurrence dialogue sentence in which the two keywords appear for the first time.

S205 , for each keyword, determine the degree of influence of the keyword on the dialogue text based on the appearance order of the keyword and the co-occurrence keywords of the keyword in each co-occurring dialogue sentence corresponding to the keyword.

For example, in a possible implementation manner, for each keyword, it is determined that the keyword located in the A first number of other keywords following the keyword and a second number of other keywords preceding the keyword. On this basis, the degree of influence of the keyword on the dialogue text may be determined based on the first quantity and the second quantity corresponding to the keyword.

In another possible implementation manner, in order to more quickly and efficiently determine the order of appearance between keywords and their co-occurrence keywords, and to efficiently count the co-occurrence keys that appear before and after each keyword The quantity information of the words, the present application can also combine the directed graph to analyze the influence relationship between the keywords and other keywords, and determine the degree of influence of the keywords on the dialogue text.

Specifically, after determining each co-occurrence keyword of the keyword and the co-occurrence dialogue sentence in which the keyword and its co-occurrence keyword co-occur for the first time, a directed graph representing the association relationship between different keywords in the dialogue text can be constructed .

The directed graph includes: a plurality of nodes and directed edges between the plurality of nodes indicating a direction. Each node represents a keyword, and having the directed edge between two nodes indicates that the two keywords corresponding to the two nodes are co-occurrence keywords. The direction indicates the order in which the two keywords corresponding to the two nodes appear in their co-occurring dialogue sentences.

As shown in FIG. 3 , it shows a schematic diagram of a directed graph provided by an embodiment of the present application. In FIG. 3 , each circle represents a node, which may include node A, node B, node C, node D, node E, node F, and node G as shown in FIG. 3 .

Wherein, different nodes represent different keywords in the dialogue text. For example, node A may represent keyword 1, node B may represent keyword 2, and others are similar, and will not be repeated here.

It can be seen from Figure 3 that some nodes are connected by directed edges with directions, while some nodes do not have directed edges. Among them, if the two keywords represented by the two nodes are co-occurring keywords, then there is a directed edge between the two nodes, and the arrow direction of the directed edge indicates that the two keywords appear at the same time for the first time. The order of appearance in the co-occurrence dialogue sentence of , the direction of the directed edge is that the keywords with the earlier order of appearance point to the keywords with the later order of appearance.

For example, the keyword 1 and the keyword 2 corresponding to the node A and the node B appear simultaneously in the dialogue sentence S for the first time, and in the dialogue sentence S, the keyword 1 appears first and the keyword 2 appears later. Therefore, in Figure 3 It is necessary to point from node A representing keyword 1 to node B representing keyword 2, so that the keyword represented by node A can lead to the keyword represented by node B.

Combining with the directed graph in Figure 3, it can be seen that based on the directed graph, it is possible to clearly characterize whether the keywords represented by each node belong to co-occurrence keywords, and whether two keywords that are co-occurring keywords appear between each other. order.

Correspondingly, on the basis of constructing the directed graph, the present application can determine the degree of influence of the keywords represented by the nodes on the dialogue text according to the number and direction of directed edges between different nodes in the directed graph.

For example, based on a directed graph, the number of other nodes pointed to by each node can be counted, and the number is essentially the first number of other keywords that appear after the keyword represented by the node. Correspondingly, the number of other nodes pointing to the node may also be counted, and the number is the second number of other keywords that appear before the keyword represented by the node.

It can be understood that the first quantity and the second quantity can be more conveniently calculated by using the existing link relationship analysis function in combination with the directed graph.

For example, in a possible implementation manner, a Hyperlink-Induced Topic Search (HITS) algorithm may be used to count the first quantity and the second quantity corresponding to each node.

The HITS algorithm can be used to calculate two values of a page, one is Hub Scores, the other is Authority Scores, these two values are interdependent and affect each other. The so-called pivot value refers to the sum of the authority values of all outgoing links on the page pointing to the page. The authority value is the sum of the pivots on the page where all incoming links are located.

In this application, the pivot value and authority value can be calculated for each node in the directed graph with the help of the HITS algorithm, where the pivot of a node refers to the total number of other nodes pointed to by the node, that is, the first node corresponding to the keyword represented by the node. The authority value of a node is the total number of all nodes pointing to the node, that is, the second number corresponding to the keyword represented by the node.

It can be understood that the first quantity and the second quantity corresponding to the keywords represented by each node are determined by analyzing the directed graph based on HITS as an example. In practical applications, other algorithms can also be used to analyze the directed graph. to determine the first quantity and the second quantity of each keyword, which is not limited.

S206, determining the importance of the keywords in combination with the correlation between different keywords and the degree of influence of the keywords on the dialogue text.

S207, based on the importance of the keywords, generate a summary of the dialogue text.

The abstract includes: dialogue sentences in the dialogue text where the keywords whose importance levels meet the conditions are located.

For the above steps S206 and S207, reference may be made to the relevant introductions in the previous embodiments, and details are not repeated here.

In the above embodiments of the present application, there may be multiple possibilities for determining the degree of correlation between keywords, which is not limited in the present application.

In a possible implementation manner, in order to reflect the correlation between keywords more accurately and reliably, in addition to combining the similarity between the vectors of keywords, the present application can also combine keywords in dialogue The word frequency in the text and the number of occurrences of different keywords in the same dialogue sentence at the same time determine the correlation between keywords.

As shown in FIG. 4 , which shows a schematic flowchart of determining the correlation between keywords in the embodiment of the present application, the embodiment may include:

S401 , for each keyword, determine a word vector of the keyword based on the dialogue sentence in which the keyword appears for the first time in the dialogue text.

It is understandable that a keyword may exist in multiple dialogue sentences, but the dialogue sentence in which the keyword appears for the first time has the greatest impact on the keyword, and can best characterize the semantics and other characteristics of the keyword. Based on this, the present application determines the word vector of the keyword based on the dialogue sentence in which the keyword appears for the first time in the dialogue text, so that the word vector of the keyword can more accurately reflect the semantics and other characteristics of the keyword .

It can be understood that, based on the dialogue sentence in which the keyword appears for the first time, there may be various ways to determine the word vector of the keyword, which is not limited in this application.

In a possible implementation manner, in order to enable the word vector of the keyword to reflect the characteristics of the context in which the keyword is located, so that the similarity between the keywords calculated subsequently can reflect the context in which the keywords are located. Similarity, the present application can use the transformer-based Bidirectional Encoder Representations from Transformers (BERT) model and the dialogue sentence in which the keyword first appears to determine the word vector of the keyword.

For example, input the dialogue sentence in which the keyword appears for the first time into the BERT model, and obtain the encoding vector of the last layer of the BERT model, and the encoding vector is the word vector of the keyword.

S402, based on the word vector of the keyword, determine the vector similarity between different keywords.

Among them, the vector similarity is the similarity between keywords. In this application, the similarity between any two keywords is mainly concerned. Therefore, in step S402, the vector similarity between the two keywords needs to be calculated respectively.

There are also multiple possibilities for determining the similarity between keywords based on the word vector of the keywords, which is not limited.

For example, in a possible implementation manner, the cosine similarity may be used to calculate the similarity between the word vectors of the keywords. For example, for any two keywords, such as keyword t _i and keyword t _j , the following formula 1 can be used to calculate the vector similarity Sim(t _i ,t _j ) between the two keywords:

Among them, Emb(t _i ) is the word vector of the keyword t _i , and Emb(t _j ) is the word vector of the keyword t _j .

The keyword t _i and the keyword t _j represent any two keywords in the dialogue text, where the keyword t _i is the ith keyword in the dialogue text, and the keyword t _j is the jth keyword in the dialogue text, i and j are natural numbers greater than or equal to 1 and less than or equal to N, where N is the total number of keywords in the dialogue text.

S403: Determine the correlation between different keywords by combining the vector similarity between different keywords, the word frequency of keywords in the dialogue text, and the number of times different keywords appear simultaneously in a dialogue sentence.

The word frequency of a keyword in the dialogue text refers to the number of times the keyword appears in the dialogue text.

The number of simultaneous occurrences of different keywords in a dialogue sentence refers to the total number of times that for any two keywords, the two keywords appear in the same dialogue sentence at the same time. For example, if keyword A and keyword B appear in dialogue sentence 1, dialogue sentence 3 and dialogue sentence 7 at the same time, the number of times these two keywords appear in the same dialogue sentence at the same time is 3 times.

In this embodiment, when determining the correlation between keywords, the vector similarity between different keywords, the word frequency of keywords in the dialogue text, and the number of times different keywords appear in the same dialogue sentence are comprehensively considered, so as to be able to The correlation between keywords is more comprehensively analyzed, so that the obtained correlation can more accurately and comprehensively reflect the correlation between keywords.

Wherein, the specific method for calculating the correlation degree can be set as required, which is not limited.

In a possible implementation manner, the present application can calculate the correlation m _ij between the keyword t _i and the keyword t _j through the following formula 2:

Among them, TF(t _i ) is the word frequency of the keyword t _i in the dialogue text, TF(t _j ) is the word frequency of the keyword t _j in the dialogue text; TS _ij is the simultaneous occurrence of the keyword t _i and the keyword t _j The number of occurrences in the same dialogue sentence, TS _ji is the number of times that the keyword TS _ji and the keyword t _i appear in the same dialogue sentence at the same time. It can be understood that the values of TS _ij and TS _ji are the same.

In the above embodiments of the present application, there are also various possibilities for determining the importance of the keywords in combination with the degree of correlation between different keywords and the degree of influence of the keywords on the dialogue text. For ease of understanding, an implementation manner for determining the importance of a keyword is described below as an example.

As shown in FIG. 5 , which shows a schematic flowchart of an implementation of determining the importance of a keyword provided by an embodiment of the present application, the method of this embodiment may include:

S501 , for each keyword, determine the first dialogue sentence in which the keyword appears for the first time in the dialogue text and the sentence source of the first dialogue sentence.

In this embodiment, it is also considered that the dialogue sentence in which the keyword appears for the first time has a great influence on determining the importance of the keyword. Therefore, this application needs to determine the source of the sentence to which the dialogue sentence in which the keyword appears for the first time belongs. The importance of keywords.

Among them, for the convenience of distinction, in this application, the dialogue sentence in which the keyword appears for the first time is referred to as the first-occurrence dialogue sentence of the keyword.

The source of the sentence of the first dialogue sentence of the keyword can represent which dialogue party among the multiple dialogue parties involved in the dialogue text belongs to which the first dialogue sentence belongs.

For example, in the customer service scenario, the dialogue text is mainly the dialogue between the user and the customer service. Therefore, the dialogue parties involved in the dialogue text include the user and the customer service. Therefore, the source of the sentence can be one of the user and the customer service.

For another example, taking a business communication scenario as an example, multiple business parties may be involved in the dialogue text, and the source of the sentence may belong to one of the multiple business parties.

S502: Determine the benchmark importance of the keyword in combination with the source of the first dialogue sentence corresponding to the keyword and the correlation between different keywords.

It can be understood that the dialogue sentences of different sentence sources involved in the dialogue text have different characteristics, and when the first dialogue sentence corresponding to a keyword belongs to a different sentence source, based on the difference between the keyword and other keywords. Relevance can reflect the importance of the keyword will also vary.

For example, still taking the customer service scenario as an example, the longer the dialogue sentence sent by the user side, the lower the suitability of the dialogue sentence for generating a summary. Correspondingly, if the first dialogue sentence of a keyword comes from the user side, the longer the dialogue sentence to which the keyword belongs, the higher the correlation between the keyword and other keywords may be, but the keyword does not act as a For more important words in the dialogue text, the importance of the keyword will be relatively low.

On the contrary, for the dialogue sentences that appear on the customer service side, considering that the dialogue sentences sent by the customer service side are generally more refined and have less irrelevant information irrelevant to the essential content of the dialogue text, the longer the dialogue sentences from the customer service side, the Then the importance of the keyword that appears in the dialogue sentence for the first time can also reflect the dialogue information of the dialogue text. Therefore, the higher the correlation between the keyword and other keywords, the higher the importance of the keyword. .

In a possible implementation manner, in the case that the source of the sentence includes one of the customer service source and the user source, for any keyword, the correlation of the relevance between the keyword and other keywords can be determined. On this basis, the reference importance degree of the keyword is determined by combining the sentence source of the first dialogue sentence corresponding to the keyword and the sum of the relevancy degrees corresponding to the keyword.

Among them, if the first dialogue sentence corresponding to the keyword comes from the customer service, there is a negative correlation between the baseline importance of the keyword and the sum of the correlation degrees corresponding to the keyword.

If the first dialogue sentence corresponding to the keyword comes from the user, there is a positive correlation between the reference importance of the keyword and the sum of the relevancy degrees corresponding to the keyword.

For example, for the case where the first dialogue sentence corresponding to the keyword t _i comes from a customer service agent (Agent), the following formula 3 can be used to calculate the benchmark importance degree SScore(t _i |Agent) of the keyword t _i :

For the case where the first dialogue sentence corresponding to the keyword t _i comes from the user (User) side, the following formula 4 can be used to calculate the benchmark importance of the keyword t _i can be calculated by the following formula 4:

Further, in order to be able to compare the benchmark importance levels of different keywords, the benchmark importance levels of different keywords can also be normalized. Therefore, the benchmark importance levels of keywords can be the normalized benchmark importance levels. degree.

For example, it is assumed that the benchmark importance of the keyword t _i is SScore(t _i ) (for example, it can be SScore(t _i |Agent) calculated based on formula 3, or SScore(t _i |User) calculated based on formula 4 )), then it can be normalized by the following formula 5 to obtain the normalized benchmark importance degree softmax(t _i ):

Among them, t _l is the lth keyword in the dialogue text, l is a natural number greater than or equal to 1 and less than or equal to N, and N is the total number of keywords in the dialogue text.

S503 , determining the importance of the keyword in combination with the benchmark importance degree of the keyword and the degree of influence of the keyword on the dialogue text.

Among them, the higher the importance of the benchmark and the higher the degree of influence, the higher the importance of the keyword.

For example, in a possible implementation manner, the reference importance degree of the keyword may be corrected by using the degree of influence of the keyword on the dialogue text to obtain the importance degree of the keyword.

For example, it is assumed that the degree of influence of keywords on the dialogue text is represented by the influence score, and the larger the influence score is, the higher the influence degree is. And the benchmark importance is also the benchmark importance score, then the product of the benchmark importance score and the influence score can be determined as the importance score representing the importance of the determined keyword.

It can be understood that the above is an example of a possible situation. In practical applications, the benchmark importance score corresponding to the benchmark importance degree of the keyword and the impact score corresponding to the impact degree can also be summed, and the result obtained by the summation can be obtained. As an importance score indicating the importance of the keyword. Of course, there may also be other possibilities for determining the importance of the keyword, which is not limited.

In the embodiment of FIG. 5 , in the composition of determining the importance of the keyword, it is considered that the first-occurring dialogue sentence in which the keyword appears for the first time has a greater impact on the keyword, and the source of the first-occurring dialogue sentence is comprehensively considered and The degree of correlation between a keyword and other keywords, comprehensively determine the basic importance of the keyword, so that the degree of correlation between the keyword and other keywords and the importance of the keyword can be more comprehensively and accurately determined.

It is understandable that in general, when different dialogue parties interact with dialogue sentences, there are usually multiple parts such as question consultation and answering involved, and the main intentions expressed by different parts will be different. Based on this, the present application can further divide the dialogue text into at least one dialogue partition, and dialogue sentences in different dialogue partitions represent different types of dialogue intentions.

Considering that dialogue sentences in different dialogue partitions belong to different dialogue intentions, the importance of dialogue sentences in different dialogue partitions to the key information representing the dialogue text will also be different. Based on this, in order to generate a more reasonable summary of the dialogue text, so that the generated summary can express the gist of the dialogue text more comprehensively and accurately, the present application can also determine the dialogue sentences used to compose the summary from each dialogue partition. .

Specifically, based on the importance of the keywords, the target dialogue sentences used for composing the abstract in the dialogue partition can be determined, and the abstract consisting of the target dialogue sentences in each dialogue partition can be obtained.

In a possible implementation manner, considering that the importance levels of the dialogue sentences in different dialogue partitions are different, the number of target dialogue sentences for generating summaries determined by the present application in different dialogue partitions may also be different.

The solution of the present application will be introduced below in conjunction with this possible implementation manner. As shown in FIG. 6 , which shows another schematic flowchart of the method for generating an abstract provided in an embodiment of the present application, the method in this embodiment may include:

S601: Determine the keywords in each dialogue sentence in the dialogue text.

S602, determine the correlation between different keywords and the appearance order of the keywords in the dialogue sentence.

S603, based on the order of appearance, determine the degree of influence of the keyword on the dialogue text.

S604: Determine the importance of the keywords in combination with the correlation between different keywords and the degree of influence of the keywords on the dialogue text.

For the above steps S601 to S604, reference may be made to the relevant introduction in the previous embodiment, and details are not repeated here.

S605: Divide the dialogue text into at least one dialogue partition.

Among them, the dialogue sentences in different dialogue partitions represent different categories of dialogue intentions.

For example, the semantics of each dialogue sentence in the dialogue text can be combined to determine the dialogue intention represented by the semantics of the dialogue sentence, and the dialogue sentences in the dialogue sentences with the same dialogue intention are divided into at least one dialogue partition.

It can be understood that the dialogue sentences in the dialogue texts of consultation and chat are carried out according to a specific dialogue logic, and the dialogue intentions represented by the logical components involved in the dialogue logic of the dialogue text are different. Therefore, The dialogue text can be divided into at least one logical partition according to the dialogue logic, and each logical partition is a dialogue partition.

For example, the dialogue of multiple dialogue parties generally starts with greeting, then asking questions, followed by problem discussion or confirmation of details, and finally problem solving. Therefore, the dialogue text can be divided into four dialogue sections. The intent of one dialogue partition is to say hello, which includes dialogue sentences related to the greeting; the intent of the second dialogue partition is the question title, which consists of one or more dialogue sentences related to asking the question and identifying the problem at hand; The intention of the dialogue partition can be to communicate the details of the problem, which can include at least one dialogue sentence to identify the specific details of the problem and explore possible directions of the problem; the intention of the fourth dialogue partition can be to determine the cause of the problem, which can include related to the problem. Conversational statements about the cause of the problem and the identified solution.

Of course, considering the greeting in the dialogue text and the goodbye sentence used to end the dialogue at the end of the dialogue text is of little significance for determining the subject meaning expressed by the text dialogue, the parts related to greeting and goodbye in the dialogue text can be removed.

for example:

FIG. 7 is a schematic diagram of a plurality of dialogue partitions divided by dialogue text.

The sentence in FIG. 7 is a dialogue sentence, and the length of the box where the sentence is located indicates the length of the dialogue sentence.

In the dialogue text shown in FIG. 7 , there is at least one dialogue sentence 701 related to greeting at the beginning, and this part of the dialogue sentence can be removed. In FIG. 7 , a cross is used to indicate deletion. In addition, the last dialogue sentence in the dialogue text is an end dialogue sentence 702 indicating goodbye, etc. to end the dialogue, and this part of the content can also be deleted.

On this basis, the dialogue text in Figure 7 is divided into three parts, namely three dialogue partitions. The three dialogue partitions are the first dialogue partition 703 located at the top, the second dialogue partition 704 located in the middle, and the third dialogue partition 705 located at the bottom.

It is understandable that since different dialogue parties conduct dialogue and communication step by step according to dialogue logic in the dialogue process, dialogue sentences belonging to the same dialogue intention are generally adjacent and continuous dialogue sentences. For example, as can be seen from FIG. 7 , at least one dialogue sentence in each dialogue partition is at least one continuous dialogue sentence.

For example, in FIG. 7 , the first dialogue section 703 may be a part for elaborating questions, and the part includes the first to sixth dialogue sentences after greeting, as shown in sentences 1 to 6 in FIG. 7 . Correspondingly, the main interaction between Statement 1 and Statement 6 is to determine the issues that need to be discussed or consulted.

The second dialogue partition 704 in FIG. 7 may be the question detail communication, which may include statement 7 to statement 8.

While the third dialogue section 705 is the problem solving section, this section may include statement 9 and statement 10.

Of course, the above is just an example of a possible realization of the dialogue partitions for dividing dialogue texts. In practical applications, the number of dialogue partitions that can be divided into dialogue texts and the partitioning method can be set as required. Add restrictions.

It can be understood that the execution order of this step S605 and any one of the previous steps S601 to S604 is not limited to the sequence number restriction in FIG. 6 . In practical applications, it is also possible to perform this step S605 first and then perform the previous steps S601 to S604. The step in the step S605 can also be performed while the step is any one of the steps 601 to S604.

S606: For each dialogue partition, determine the target number of target dialogue sentences corresponding to the dialogue partition from within the dialogue partition based on the target number of sentences to be extracted from the dialogue partitions with different dialogue intentions and in combination with the importance of the keywords.

The method of extracting the target number of target dialogue sentences from the dialogue partition is similar to the previous process of determining dialogue sentences from dialogue texts.

For example, for each dialogue sentence, the importance of the dialogue sentence can be determined based on the importance of each keyword in the dialogue sentence (for the specific implementation method, please refer to the previous introduction). On this basis, for each dialogue partition, combine the dialogue The number of targets corresponding to the partition can be used to extract the number of dialogue sentences with higher importance in the dialogue partition.

Wherein, the number of targets to be extracted for dialogue partitions with different dialogue intentions may be different or the same, and may be specifically set as required.

Considering that the dialogue sentences in the dialogue partitions with different intentions and the dialogue text need to express the subject content differently, that is, the importance levels are different, therefore, the present application can set different target numbers for dialogue partitions with different dialogue intentions. For example, considering that the dialogue sentences in the dialogue partition related to question formulation can reflect the topic direction of the dialogue text, the target number corresponding to the dialogue partition related to question extraction (for example, the first dialogue partition in Figure 7) can be set to relatively more.

For example, the example of FIG. 7 is still used for description. Assume that the number of targets that can be extracted from the dialogue partition of the question extraction intention is 3, and the number of targets corresponding to the dialogue partition representing the detailed description of the problem is 2, and the number of targets for the dialogue partition related to problem solving can be 1. . Then, it is necessary to extract 3 target dialogue sentences with higher importance from the first dialogue partition in Figure 7. Similarly, extract 2 target dialogue sentences from the second dialogue partition, and extract 2 target dialogue sentences from the third dialogue partition. A target dialogue sentence is extracted from the partition.

S607, based on the target dialogue sentences in each dialogue partition, generate a summary of the dialogue text.

For example, the target dialogue sentences in each dialogue partition can be combined into a summary according to the sequence of dialogue partitions and the order of target dialogue sentences in each dialogue partition.

In this embodiment, the dialogue text is divided into at least one dialogue partition with different dialogue intentions, and then the dialogue sentences used to generate the summary are extracted from the dialogue partition, which is beneficial to enable the summary to cover sentences representing different intentions, and to enable the summary to More fully reflects the content of the dialogue text.

In addition, since the present application can sequentially generate dialogue summaries from the dialogue sentences extracted from each dialogue partition, the dialogue summaries can also reflect the dialogue logic of the dialogue text.

In order to facilitate the understanding of the solution of the present application, the solution of the present application is described below by taking the method of generating a dialogue summary in combination with a directed graph as an example.

As shown in FIG. 8 , which shows another schematic flowchart of the method for generating an abstract provided in an embodiment of the present application, the method in this embodiment may include:

S801: Determine the keywords in each dialogue sentence in the dialogue text, and obtain N keywords in the dialogue text.

S802: Divide the dialogue text into at least one dialogue partition.

Each dialogue partition includes at least one dialogue sentence, and different dialogue partitions contain different dialogue sentences.

Among them, the dialogue sentences in different dialogue partitions represent different categories of dialogue intentions.

The sequence of the above steps S801 and S802 is not limited to that shown in FIG. 8 .

S803, for any two keywords in the dialogue text, determine the degree of correlation between the two keywords.

It can be understood that, by calculating the correlation between two keywords in the dialogue text, an N*N co-occurrence matrix M can be obtained, and the elements in the co-occurrence matrix are the keyword t _i and the keyword t _j . The degree of correlation m _ij between different elements corresponds to the degree of correlation between different pairs of keywords, and the keywords include two keywords in the dialogue text.

For the calculation of the correlation degree, reference may be made to the relevant introduction of the foregoing formula 2, and details are not repeated here.

S804, for any two keywords in the dialogue text, determine the co-occurrence dialogue sentence in which the two keywords appear at the same time for the first time, and obtain the appearance order of the two keywords in the co-occurrence dialogue sentence.

S805 , construct a directed graph representing the association relationship between different keywords in the dialogue text based on the correlation between two keywords in the dialogue text and the appearance order of any two keywords in the co-occurring dialogue sentences.

The directed graph includes: a plurality of nodes and directed edges indicating a direction between the plurality of nodes, each node represents a keyword, and a directed edge between two nodes indicates the relationship between the two nodes. The degree of relevance is not zero, and the direction of the directed edge between the two nodes represents the order in which the two keywords corresponding to the two nodes appear in the co-occurring dialogue sentence.

It is understandable that, according to the previous formula 2, if two keywords do not appear in the same dialogue sentence at the same time, the correlation between the two keywords will be zero. Therefore, the correlation can be combined to determine the two keywords. Whether there is a directed edge between the nodes corresponding to the word.

Meanwhile, in this embodiment, the weight value of a directed edge between two nodes in the directed graph may be the correlation between two keywords represented by the two nodes. Specifically, the weight value (or correlation) of the directed edge can be determined in combination with the co-occurrence matrix M.

It can be understood that, in order to reflect the dialogue partition from which the dialogue sentence where the keyword appears for the first time, the directed graph constructed in this application can also indicate which dialogue partition the keyword represented by the node comes from. As shown in Figure 9, the directed graph divides the dialogue partitions to which the keywords represented by each node belong. For example, the keywords represented by node A and node B are from the first dialogue partition. Similar to other nodes, they are not repeated here. Repeat.

S806, for each node in the directed graph, use the HITS algorithm to determine the authority value and the pivot value of the node, and determine the larger of the authority value and the pivot value of the node as the keyword represented by the node. The text's degree of influence value.

S807: For each node in the directed graph, determine the first occurrence of the dialogue sentence in which the keyword represented by the node appears for the first time in the dialogue text and the sentence source of the first occurrence of the dialogue sentence.

S808: Determine the reference importance of the keyword represented by the node in combination with the source of the first dialogue sentence and the correlation on the directed edge between the node and other nodes in the directed graph.

Wherein, for the process of determining the benchmark importance degree of the keyword, reference may be made to the relevant introduction of the previous embodiment, and details are not described herein again.

It should be noted that in this embodiment, the reference importance of keywords is determined based on the combination of directed graphs, but it can be understood that the combination of directed graphs only more clearly shows the degree of correlation between keywords and other keywords. , but the correlation between keywords is directly used without being combined with the directed graph, and the determination of the benchmark importance of keywords is also applicable to this embodiment.

S809: Determine the importance degree score of the keyword represented by the node in combination with the reference importance degree of the keyword represented by the node and the influence degree value of the keyword represented by the node on the dialogue text.

For example, taking the benchmark importance level as the benchmark importance level value as an example, assuming that the benchmark importance level value of the keyword t _i is obtained from the previous formulas 3 to 5, the benchmark importance level value is softmax(t _i ), then the importance level score of the keyword t _i S(t _i ) can be obtained by the following formula 6:

S(t _i )=softmax(t _i )*(1+max(HITS _A (t _i ),HITS _H (t _i ))) (Formula 6);

Among them, HITS _A (t _i ) is the authority value corresponding to the keyword t _i , and HITS _H (t _i ) is the pivot value corresponding to the keyword t _i .

S810, for any dialogue sentence in the dialogue text, determine the importance degree score corresponding to the keyword with the highest importance degree score in the dialogue sentence as the score of the dialogue sentence.

S811 , for each dialogue partition, according to the target number of sentences to be extracted for the dialogue partition, extract from the dialogue partition the former target number of target dialogue sentences with a higher importance score.

S812 , combine the target dialogue sentences extracted from each dialogue partition into an abstract of the dialogue text according to the sequence of each dialogue partition and the sequence of the target dialogue sentences extracted in the dialogue partition.

Corresponding to an abstract generating method provided by an embodiment of the present application, an embodiment of the present application also provides an abstract generating apparatus.

As shown in FIG. 10, it shows a schematic diagram of the composition and structure of an embodiment of an abstract generating apparatus provided by an embodiment of the present application. The apparatus in this embodiment may include:

A keyword determining unit 1001, configured to determine keywords in each dialogue sentence in the dialogue text;

an association determination unit 1002, configured to determine the degree of relevancy between different keywords and the order of appearance of the keywords in the dialogue sentence;

an influence determination unit 1003, configured to determine the degree of influence of the keyword on the dialogue text based on the appearance order;

An importance determination unit 1004, configured to determine the importance of the keyword in combination with the correlation between different keywords and the degree of influence of the keyword on the dialogue text;

The abstract generating unit 1005 is configured to generate an abstract of the dialogue text based on the importance of the keywords, where the abstract includes: dialogue sentences in the dialogue text where the keywords whose importance meets the condition are located.

In a possible implementation, the device further includes:

a dialogue partitioning unit, used for dividing the dialogue text into at least one dialogue partition before the summary generating unit generates the abstract of the dialogue text, and dialogue sentences in different dialogue partitions represent different types of dialogue intentions;

The abstract generating unit is specifically configured to determine, based on the importance of the keyword, target dialogue sentences in the dialogue partition for composing the abstract, and obtain an abstract consisting of the target dialogue sentences in each dialogue partition.

In another possible implementation manner, the abstract generating unit includes:

The sentence determination unit is configured to determine the target number of target dialogue sentences from the dialogue partition based on the target number of sentences to be extracted from the dialogue partitions with different dialogue intentions and in combination with the importance of the keyword.

In yet another possible implementation, the association determination unit includes:

A co-occurrence determination unit, configured to determine, for each keyword, each co-occurrence keyword of the keyword and a co-occurrence dialogue sentence in which the keyword and its co-occurrence keyword co-occur for the first time, and the co-occurrence of the keyword Keywords are other keywords that appear in a dialogue sentence at the same time as the keyword;

An order determination unit, configured to determine the appearance order of the keyword and its co-occurrence keyword in the co-occurrence dialogue sentence.

In yet another possible implementation manner, the influence determination unit includes:

A quantity determination subunit, configured to determine a first quantity of other keywords located after the keyword and a second quantity of other keywords located before the keyword based on the appearance order;

An influence determination subunit, configured to determine the degree of influence of the keyword on the dialogue text based on the first quantity and the second quantity corresponding to the keyword.

In yet another possible implementation manner, the sequence determination unit includes:

a directed graph construction unit, used for constructing a directed graph representing the association relationship between different keywords in the dialogue text;

Wherein, the directed graph includes: a plurality of nodes and a directed edge indicating a direction between the plurality of nodes, each node represents a keyword, and the directed edge between two nodes indicates that the two The two keywords corresponding to the nodes are mutually co-occurring keywords, and the direction of the directed edge between the two nodes represents the appearance order of the two keywords corresponding to the two nodes in the co-occurrence dialogue sentence;

The influence determination unit is specifically configured to determine the degree of influence of the keywords represented by the nodes on the dialogue text according to the number and direction of directed edges between different nodes in the directed graph.

In yet another possible implementation manner, the importance determination unit includes:

a source determination subunit, configured to, for each keyword, determine the first dialogue sentence in which the keyword appears for the first time in the dialogue text and the sentence source of the first dialogue sentence;

A benchmark determination subunit, configured to determine the benchmark importance of the keyword in combination with the source of the first dialogue sentence corresponding to the keyword and the correlation between different keywords;

The importance degree determination subunit is used for determining the importance degree of the keyword in combination with the reference degree of importance of the keyword and the degree of influence of the keyword on the dialogue text.

In another possible implementation manner, the source party of the sentence determined by the source party determining subunit includes one of a customer service source and a user source;

The benchmark determines subunits, including:

and a determining subunit for determining the sum of the relevancy degrees between the keyword and other keywords;

The benchmark degree determination subunit is used to determine the benchmark importance degree of the keyword in combination with the sentence source of the first dialogue sentence corresponding to the keyword and the sum of the relevancy degrees corresponding to the keyword; wherein, if the key The first dialogue sentence corresponding to the word comes from the customer service, and the reference importance of the keyword is negatively correlated with the sum of the correlation degrees corresponding to the keyword; if the first dialogue sentence corresponding to the keyword comes from the user, the There is a positive correlation between the benchmark importance of the keyword and the sum of the relevancy degrees corresponding to the keyword.

In yet another possible implementation, the association determination unit includes:

A vector determination subunit, configured to, for each of the keywords, determine the word vector of the keyword based on the dialogue sentence in which the keyword appears for the first time in the dialogue text;

a similarity determination subunit, used for determining the vector similarity between different keywords based on the word vectors of the keywords;

The correlation determination subunit is used to combine the vector similarity between different keywords, the word frequency of the keywords in the dialogue text and the number of occurrences of different keywords that appear in a dialogue sentence at the same time, to determine the difference. correlation between keywords.

In another aspect, the present application also provides an electronic device, as shown in FIG. 11 , which shows a schematic diagram of the composition structure of the electronic device, the electronic device can be any type of electronic device, and the electronic device at least includes processor 1101 and memory 1102;

Wherein, the processor 1101 is configured to execute the abstract generating method in any one of the above embodiments.

The memory 1102 is used to store programs required by the processor to perform operations.

It can be understood that the electronic device may further include a display unit 1103 and an input unit 1104 .

Of course, the electronic device may also have more or less components than those shown in FIG. 11 , which is not limited.

In another aspect, the present application also provides a computer-readable storage medium, where at least one instruction, at least one program, code set or instruction set is stored in the computer-readable storage medium, the at least one instruction, the at least one A piece of program, the code set or the instruction set is loaded and executed by the processor to implement the abstract generation method described in any one of the above embodiments.

The present application also proposes a computer program comprising computer instructions stored in a computer-readable storage medium. When the computer program runs on the electronic device, it is used to execute the abstract generating method in any one of the above embodiments.

It should be noted that the various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. For the same and similar parts among the various embodiments, refer to each other Can. Meanwhile, the features described in each embodiment in this specification can be replaced or combined with each other, so that those skilled in the art can realize or use the present application. As for the apparatus type embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and for the relevant part, please refer to the partial description of the method embodiment.

Finally, it should also be noted that in this document, relational terms such as first and second are used only to distinguish one entity or operation from another, and do not necessarily require or imply these entities or that there is any such actual relationship or sequence between operations. Moreover, the terms "comprising", "comprising" or any other variation thereof are intended to encompass a non-exclusive inclusion such that a process, method, article or device comprising a list of elements includes not only those elements, but also includes not explicitly listed or other elements inherent to such a process, method, article or apparatus. Without further limitation, an element qualified by the phrase "comprising a..." does not preclude the presence of additional identical elements in the process, method, article, or device that includes the element.

The above description of the disclosed embodiments enables any person skilled in the art to make or use the present application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, this application is not intended to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

The above are only the preferred embodiments of the present application. It should be pointed out that for those skilled in the art, without departing from the principles of the present application, several improvements and modifications can also be made, and these improvements and modifications should also be regarded as The protection scope of this application.

Claims (10)

1. A summary generation method comprises the following steps:

determining keywords in each pair of spoken sentences in the dialog text;

determining the correlation degree between different key words and the appearance sequence between the key words in the dialogue sentences;

determining the influence degree of the keywords on the dialog text based on the appearance sequence;

determining the importance degree of the keywords by combining the correlation degrees among different keywords and the influence degree of the keywords on the dialog text;

generating a summary of the dialog text based on the importance of the keyword, the summary comprising: and the conversation sentence in which the key word with the importance degree meeting the condition is positioned in the conversation text.

2. The method of claim 1, prior to said generating the summary of the dialog text, further comprising:

dividing the dialogue text into at least one dialogue partition, wherein dialogue sentences in different dialogue partitions represent different types of dialogue intents;

generating the abstract of the dialog text based on the importance degree of the keyword comprises the following steps:

and determining the target dialogue sentences used for composing the abstract in the dialogue partitions based on the importance degrees of the keywords to obtain the abstract composed of the target dialogue sentences in each dialogue partition.

3. The method of claim 2, wherein determining the target dialog sentences for composing the summary within the dialog partition based on the importance of the keyword comprises:

and determining the target quantity of the target dialog sentences from the dialog partitions based on the target quantity of the sentences required to be extracted by the dialog partitions with different dialog intentions and combining the importance degrees of the keywords.

4. The method of claim 1, determining an order of occurrence between keywords within a conversational sentence comprising:

for each keyword, determining co-occurrence keywords of the keyword and a co-occurrence dialogue sentence in which the keyword and the co-occurrence keyword thereof commonly occur for the first time, wherein the co-occurrence keywords of the keyword are other keywords which simultaneously occur in one dialogue sentence with the keyword;

determining an order of occurrence between the keyword and its co-occurring keywords within the co-occurring dialog sentence.

5. The method of claim 1 or 4, wherein determining the degree of influence of the keyword on the dialog text based on the order of occurrence comprises:

determining a first number of other keywords that are located after the keyword and a second number of other keywords that are located before the keyword based on the order of occurrence;

and determining the influence degree of the keywords on the dialog text based on the first quantity and the second quantity corresponding to the keywords.

6. The method of claim 4, said determining an order of occurrence between the keyword and its co-occurring keyword within the co-occurring dialog statement, comprising:

constructing a directed graph representing the incidence relation between different keywords in the dialog text;

wherein the directed graph comprises: the method comprises the following steps that directional edges with directions are indicated among a plurality of nodes and a plurality of nodes, each node represents a keyword, the directional edges are arranged between the two nodes to represent that the two keywords corresponding to the two nodes are co-occurrence keywords, and the directions of the directional edges between the two nodes represent the occurrence sequence of the two keywords corresponding to the two nodes in a co-occurrence dialogue statement;

the determining the influence degree of the keyword on the dialog text based on the appearance sequence comprises:

and determining the influence degree of the keywords represented by the nodes on the dialog text according to the number and the direction of directed edges between different nodes in the directed graph.

7. The method of claim 1, wherein determining the importance of the keyword by combining the relevance between different keywords and the influence degree of the keyword on the dialog text comprises:

aiming at each keyword, determining a first-appearing dialogue sentence with the keyword appearing for the first time in the dialogue text and a sentence source party of the first-appearing dialogue sentence;

determining the reference importance degree of the keywords by combining the sentence source party of the first-time dialog sentence corresponding to the keywords and the correlation degree between different keywords;

and determining the importance degree of the keyword by combining the reference importance degree of the keyword and the influence degree of the keyword on the dialog text.

8. The method of claim 7, the source of the statement comprising one of customer service origin and user origin;

determining the reference importance degree of the keyword by combining the sentence source side of the first-occurring dialog sentence corresponding to the keyword and the correlation degree between different keywords, wherein the step comprises the following steps of:

determining a correlation sum of correlations between the keyword and other keywords;

determining the reference importance degree of the keyword by combining the sentence source party of the first-occurring dialog sentence corresponding to the keyword and the correlation sum corresponding to the keyword;

if the first-time dialog statement corresponding to the keyword comes from customer service, the reference importance degree of the keyword is negatively correlated with the sum of the correlation degrees corresponding to the keyword;

and if the first-occurring dialog sentence corresponding to the keyword is from the user, positively correlating the reference importance degree of the keyword with the sum of the correlation degrees corresponding to the keyword.

9. The method of claim 1, determining relevance between different keywords, comprising:

for each keyword, determining a word vector of the keyword based on a dialog sentence in the dialog text, in which the keyword appears for the first time;

determining vector similarity between different keywords based on word vectors of the keywords;

and determining the correlation degree between different keywords by combining the vector similarity between different keywords, the word frequency of the keywords in the dialog text and the occurrence frequency of the keywords in a dialog sentence.

10. A digest generation apparatus comprising:

a keyword determining unit, configured to determine keywords in each pair of spoken sentences in the dialog text;

the association determining unit is used for determining the correlation between different keywords and the appearance sequence between the keywords in the dialogue sentences;

an influence determining unit configured to determine a degree of influence of the keyword on the dialog text based on the appearance order;

the importance determining unit is used for determining the importance of the keywords by combining the relevance among different keywords and the influence degree of the keywords on the dialog text;

a summary generation unit, configured to generate a summary of the dialog text based on the importance degree of the keyword, where the summary includes: and the conversation sentence in which the key word with the importance degree meeting the condition is positioned in the conversation text.

Images