JP3614055B2 - Summary sentence creation method and apparatus, and storage medium storing summary sentence creation program - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a summary sentence creation method and apparatus, and a storage medium storing a summary sentence creation program. In particular, in text processing, a summary of a plurality of e-mails, home pages, shop guidance sentences, etc. is displayed as a list. In addition, the present invention relates to a summary sentence creation method and apparatus for summarizing the contents in about one line, and a storage medium storing a summary sentence creation program.
[0002]
For details, analyze sentence structure from input text, select summary words from extracted important words, select sentence templates based on thesaurus knowledge, and generate summary sentences according to them. The present invention relates to a summary sentence creation method and apparatus, and a storage medium storing a summary sentence creation program.
[0003]
[Prior art]
Conventionally, when displaying a list of documents, a title is displayed or the first sentence of the document is displayed.
In addition, there is a method of making an excerpt sentence that selects and collects important sentences when creating a summary sentence. This method first checks all the nouns contained in the sentence, and the noun at the top rank is considered the most important keyword of the sentence. Make a ranking. Then, low-ranked sentences are discarded according to the specified summary rate. As described above, conventionally, an abstract word from which an important sentence is extracted with respect to the summarization technique is selected by a statistical technique such as the tf × idf method.
[0004]
[Problems to be solved by the invention]
However, there are many untitled documents on the homepage, and the contents cannot be expressed only by the top of the document. In particular, when a long sentence is displayed in about one line, it may be interrupted and the content may not be understood.
In addition, since the keyword has the same weight regardless of where it appears, the word appearing in the title and the word in the supplementary explanation at the end have the same importance. Since the title is considered to be an expression that summarizes the sentence, it is necessary to increase the importance of the word extracted from the title, and it is necessary to decrease the importance of the word extracted from the supplementary explanation.
[0005]
When displaying a list of search results, it is effective to unify the items to be displayed so that differences in the contents of each document can be understood. However, in the above conventional statistical method, keywords are selected based on the meaning of words. Since it has not been extracted, it is not possible to extract keywords from the viewpoints of “target person”, “handling product”, and “price”.
When some information and a summary are displayed at the same time, it is not necessary to put the contents overlapping with the information displayed other than the summary in the summary. However, such a determination is not possible with the above-described document list display method in the prior art. For example, when searching for “welfare services for the elderly”, it is clear that the target is “elderly”, so it is not a summary of “sending helpers to the elderly” but “send helpers” Is enough. In addition, when the title and the summary are displayed at the same time, it is not necessary to include the content overlapping the title in the summary. However, such a determination is impossible with the conventional technology.
[0006]
For example, for a title of “helper dispatch”, the user should not display a summary of “send a helper” but other information such as additional help, helper's work contents, and whether a fee will be charged. Useful.
In addition, when searching for “retirement home” and searching for a plurality of services such as “special elderly nursing home” and “nursing nursing home”, if a summary is individually created for each guidance service, It is not a useful summary. Compare the contents of all the searched services, highlight the differences in each service, and emphasize the differences, "I care for bedridden elderly people and elderly people with dementia", "I care for elderly people living alone and low-income earners" Therefore, it is necessary to improve the user service selectivity.
[0007]
Furthermore, in the above conventional technique, since summary information cannot be changed dynamically based on the displayable area on the device, for example, when the guide area is a 40-character sentence, the summary sentence is edited with a fixed logic. In such a case, only a part of the summary sentence can be guided, or the summary becomes too simple and there is a problem that there is too little information.
[0008]
The present invention has been made in view of the above points, and an object thereof is to provide a summary sentence creation method and apparatus capable of creating a summary sentence only with useful words, and a storage medium storing a summary sentence creation program. And
[0009]
[Means for Solving the Problems]
The present invention (Claim 1)In a summary sentence generator for generating a summary sentence of a documentA summary sentence generation method,
In sentence structure analysis means,
When a document composed of a plurality of sentences including information indicating the sentence format is input by the sentence input means (step 1),
A sentence structure analysis process for extracting a plurality of sentences and sentence types of each sentence from the input document using a structure definition rule in which correspondence between sentence format information and sentence types is defined in advance (step 2) )When,
In service important word extraction means,
For each analysis item determined in advance, an extraction rule in which the relationship between the dependency relationship of the extracted important word, the attribute of the extracted important word and the certainty factor of the extracted important word attribute is defined in advance, Service important word extraction process (step 3) for extracting the certainty of the important word group by one or more important words of each sentence and the important word group of each sentence from the sentence;
In summary word selection control means,
The sentence type in which each important word appears, the attribute of the important word, and the selection method in which the conversion method of the importance is determined from the certainty, and the sentence type of each sentence, the important word group of each sentence, and the important word of each sentence. Find the importance of key words in each sentence from the confidence of the group,
When the important word group of each sentence matches the word input from the outside or the important word group extracted in the service important word extraction process from the sentence of the specified sentence type in the input document , Reduce the importance of matched key words,
For each analysis item, using a thesaurus knowledge in which a correspondence between a word and a concept of a word is defined in advance, the important word group of each sentence is grouped for each having the same concept, and the important word group having the same concept The importance of the important word group having the same concept is calculated by adding the importance of the important word group of each sentence constituting the important word group having the same concept,
For each analysis item, a summary word selection control process (step 4) for selecting, as the summary word group of each analysis item, the one having the highest importance among the important word groups having the same concept;
In the template selection means,
For each analysis item, the template of the summary word of each analysis item is selected from the summary word group of the analysis item by using a template selection rule in which a combination of the attribute of each summary word and the template of the summary sentence constituting the summary word group is defined in advance. A template selection process (step 5) for selecting
In the summary sentence generation means,
For each analysis item, a summary sentence generation process (step 6) is performed in which a summary word group for each analysis item is generated by embedding a summary word group for each analysis item in the summary sentence template for each analysis item selected.
[0010]
The present invention (Claim 2)A summary sentence generation method in a summary sentence generation apparatus for generating a summary sentence of a document,
In sentence structure analysis means,
When a document consisting of a plurality of sentences including information indicating the sentence format is input by the sentence input means, input is made using a structure definition rule in which the correspondence between the information indicating the sentence format and the sentence type is defined in advance. A sentence structure analysis process for extracting a plurality of sentences and sentence types from each sentence
In service important word extraction means,
For each predetermined analysis item, using an extraction rule in which the relationship between the dependency relationship of the extracted important word and the attribute of the extracted important word and the certainty factor of the extracted important word attribute is defined in advance. Service important word extraction process to extract the certainty of important words by one or more important words of each sentence and important words from each sentence from sentences,
In summary word selection control means,
The sentence type in which each important word appears, the attribute of the important word, and the selection method in which the conversion method of the importance is determined from the certainty, and the sentence type of each sentence, the important word group of each sentence, and the important word of each sentence. Find the importance of the key word group of each sentence from the confidence of the group,
Match if the key word group of each sentence matches completely or partially with the key word group of each sentence of the other document extracted in the sentence structure analysis process and service key word extraction process from other documents Reduced the importance of important word groups
For each analysis item, using a thesaurus knowledge in which the correspondence between words and word concepts is defined in advance, the important word groups of each sentence are grouped into the important word groups having the same concept for each having the same concept. The importance of the important word group having the same concept is obtained by adding the importance of the important word group of each sentence constituting the important word group having the same concept,
For each analysis item, a summary word selection control process of selecting the most important word group having the same concept as the summary word group of each analysis item,
In the template selection means,
For each analysis item, the template of the summary word of each analysis item is selected from the summary word group of the analysis item by using a template selection rule in which a combination of the attribute of each summary word and the template of the summary sentence constituting the summary word group is defined in advance. A template selection process for selecting
In the summary sentence generation means,
For each analysis item, a summary sentence generation process for generating a summary sentence of each analysis item by embedding the summary word group of each analysis item in the summary sentence template of each analysis item selected,
Do.
[0011]
The present invention (claims)3)A summary sentence generation method in a summary sentence generation apparatus for generating a summary sentence of a document, comprising:
In sentence structure analysis means,
When a document consisting of a plurality of sentences including information indicating the sentence format is input by the sentence input means, input is made using a structure definition rule in which the correspondence between the information indicating the sentence format and the sentence type is defined in advance. A sentence structure analysis process for extracting a plurality of sentences and sentence types from each sentence
In service important word extraction means,
For each predetermined analysis item, the relationship between the dependency relationship of the important noun to be extracted, the correspondence between the important noun to be extracted and the certainty of the important noun to be extracted, and the dependency relationship between the important verb to be extracted and the important verb to be extracted Service importance to extract the confidence level of the important word group of each sentence and one or more important nouns and important verbs of each sentence and the important word group of each sentence using the extraction rule with which the association with Word extraction process,
In summary word selection control means,
Sentence type of each sentence, important noun of each sentence, importance of each sentence Find the importance of the important word group of each sentence from the certainty of the word group,
Important nouns and important verbs in the important word group of each sentence are extracted from other documents in the sentence structure analysis process and service important word extraction process. If it is exactly the same, the important word group in which the important noun and the important verb exactly match is rejected, and if not, the important noun of each sentence that matches the important noun of each sentence in other documents is rejected. Reject important nouns in each important word group from the important word group,
For each analysis item, using a thesaurus knowledge in which the correspondence between words and word concepts is defined in advance, the important word groups of each sentence are grouped into those having the same concept to form an important word group having the same concept. Adding the importance of the important word group of each sentence constituting the important word group having the same concept to obtain the importance of the important word group having the same concept,
For each analysis item, a summary word selection control process of selecting the most important word group having the same concept as the summary word group of each analysis item,
In the template selection means,
For each analysis item, the template of the summary word of each analysis item is selected from the summary word group of the analysis item by using a template selection rule in which a combination of the attribute of each summary word and the template of the summary sentence constituting the summary word group is defined in advance. A template selection process for selecting
In the summary sentence generation means,
For each analysis item, a summary sentence generation process for generating a summary sentence of each analysis item by embedding the summary word group of each analysis item in the summary sentence template of each analysis item selected,
I do.
[0012]
The present invention (Claim 4)A summary sentence generation method in a summary sentence generation apparatus for generating a summary sentence of a document,
In sentence structure analysis means,
When a document composed of a plurality of sentences including information indicating the sentence format is input by the sentence input unit, information indicating the sentence format and a structure definition rule in which a correspondence between the sentence types is defined in advance are used. A sentence structure analysis process for extracting a plurality of sentences and sentence types of each sentence from the input document;
In service important word extraction means,
For each predetermined analysis item, using an extraction rule in which the relationship between the dependency relationship of the extracted important word and the attribute of the extracted important word and the certainty factor of the extracted important word attribute is defined in advance. Service important word extraction process to extract the certainty of important words by one or more important words of each sentence and important words from each sentence from sentences,
In summary word selection control means,
The sentence type in which each important word appears, the attribute of the important word, and the selection method in which the conversion method of the importance is determined from the certainty, and the sentence type of each sentence, the important word group of each sentence, and the important word of each sentence. Find the importance of the key word group of each sentence from the confidence of the group,
Extract words that characterize the input sentence from the sentence input from the outside, user intention inference knowledge defined in advance,
For each sentence, if the word related to the extracted word is included in the important word group of each sentence, increase the importance of the important word group including the related word,
For each analysis item, using a thesaurus knowledge in which the correspondence between the word and the concept of the word is defined in advance, the important word group of each sentence is grouped together for each having the same concept to be an important word group having the same concept, The importance of the important word group having the same concept is calculated by adding the importance of the important word group of each sentence constituting the important word group having the same concept,
For each analysis item, a summary word selection control process of selecting the most important word group having the same concept as the summary word group of each analysis item,
In the template selection means,
For each analysis item, the template of the summary word of each analysis item is selected from the summary word group of the analysis item by using a template selection rule in which a combination of the attribute of each summary word and the template of the summary sentence constituting the summary word group is defined in advance. A template selection process for selecting
In the summary sentence generation means,
For each analysis item, a summary sentence generation process for generating a summary sentence of each analysis item by embedding a summary word group of each analysis item in a summary sentence template of each selected analysis item is performed.
[0016]
Figure2 is a diagram illustrating the principle of the present invention.
The present invention (claims)5)Of the documentA summary sentence generation device for generating a summary sentence,
A sentence input means 10 for inputting a document composed of a plurality of sentences including information indicating a sentence format;
A sentence structure analysis means 20 for extracting a plurality of sentences and sentence types of each sentence from an input document using a structure definition rule 101 in which correspondence between sentence format information and sentence types is defined in advance; ,
For each predetermined analysis item, by using an extraction rule 102 in which an association between a dependency relationship of an extracted important word, an attribute of the extracted important word and a certainty factor of the extracted important word attribute is defined in advance, Service important word extraction means 30 for extracting the confidence level of the important word group of each sentence from one sentence or one or more important words of each sentence and the important word group of each sentence;
Using the selection rule 103 in which the conversion method of importance is pre-defined from the sentence type in which the important word appears, the attribute of the important word, and the certainty factor, the sentence type of each sentence, the important word group of each sentence, and the importance of each sentence Find the importance of the important word group of each sentence from the certainty of the word group,
When the important word group of each sentence matches the word inputted from the outside or the important word group extracted by the service important word extracting means 30 from the sentence of the sentence type specified in the inputted document in advance. Lower the importance of the matched key words group,
For each analysis item, the thesaurus knowledge 104 in which the correspondence between the words and the word concepts is defined in advance is used, and the important words having the same concept are grouped together for each sentence having the same concept. The importance of the important word group having the same concept is obtained by adding the importance of the important word group of each sentence constituting the important word group having the same concept as a group,
Summary word selection control means 40 that selects, for each analysis item, the most important word group having the same concept as the summary word group of each analysis item;
For each analysis item, by using a template selection rule 105 in which a set of attributes of summary words and a template of summary sentences constituting the summary word group is defined in advance, the summary sentence of each analysis item is extracted from the summary word group of analysis items. Template selection means 50 for selecting a template;
For each analysis item, there is a summary sentence generation means 60 for generating a summary sentence for each analysis item by embedding a summary word group for each analysis item in the summary sentence template for each analysis item selected.
[0017]
The present invention (claims)6)A summary sentence generation device for generating a summary sentence of a document,
A sentence input means for inputting a document composed of a plurality of sentences including information indicating a sentence format;
Sentence structure analysis means for extracting a plurality of sentences and sentence types of each sentence from an input document using a structure definition rule in which correspondence between information indicating a sentence format and sentence types is defined in advance;
For each analysis item determined in advance, an extraction rule in which the relationship between the dependency relationship of the extracted important word, the attribute of the extracted important word and the certainty factor of the extracted important word attribute is defined in advance, A service important word extracting means for extracting a certainty level of an important word group by one or more important words of each sentence and an important word group of each sentence from the sentence;
The sentence type in which each important word appears, the attribute of the important word, and the selection method in which the conversion method of the importance is determined from the certainty, and the sentence type of each sentence, the important word group of each sentence, and the important word of each sentence Find the importance of the key word group of each sentence from the confidence of the group,
Match if the key word group of each sentence completely or partially matches the key word group of each sentence of the other document extracted from the other document by the sentence structure analysis means and the service key word extraction means Reduced the importance of important word groups
For each analysis item, using a thesaurus knowledge in which the correspondence between words and word concepts is defined in advance, the important word groups of each sentence are grouped into the important word groups having the same concept for each having the same concept. The importance of the important word group having the same concept is obtained by adding the importance of the important word group of each sentence constituting the important word group having the same concept,
Summary word selection control means for selecting, for each analysis item, the most important word group having the same concept as the summary word group of each analysis item;
For each analysis item, the template of the summary word of each analysis item is extracted from the summary word group of the analysis item by using a template selection rule in which a combination of the attribute of each summary word and the template of the summary sentence constituting the summary word group is defined in advance. A template selection means for selecting
For each analysis item, there is provided summary sentence generation means for generating a summary sentence for each analysis item by embedding a summary word group for each analysis item in a summary sentence template for each analysis item selected.
[0018]
The present invention (claims)7)A summary sentence generation device for generating a summary sentence of a document,
A sentence input means for inputting a document composed of a plurality of sentences including information indicating a sentence format;
Sentence structure analysis means for extracting a plurality of sentences and sentence types of each sentence from an input document using a structure definition rule in which correspondence between information indicating a sentence format and sentence types is defined in advance;
For each predetermined analysis item, the relationship between the dependency relationship of the important noun to be extracted, the correspondence between the important noun to be extracted and the certainty of the important noun to be extracted, and the dependency relationship between the important verb to be extracted and the important verb to be extracted Service importance to extract the confidence level of the important word group of each sentence and one or more important nouns and important verbs of each sentence and the important word group of each sentence using the extraction rule with which the association with Word extraction means;
Sentence type of each sentence, important noun of each sentence, and importance of each sentence Find the importance of the important word group of each sentence from the certainty of the word group,
Important nouns and important verbs in important word groups of each sentence are extracted from other documents by sentence structure analysis means and service important word extraction means. If it is exactly the same, the important word group in which the important noun and the important verb are completely matched is rejected. If it is not exactly the same, the important noun of each sentence in the other document is matched. Reject important nouns in each important word group from the important word group,
For each analysis item, using a thesaurus knowledge in which the correspondence between words and word concepts is defined in advance, the important word groups of each sentence are grouped into those having the same concept to form an important word group having the same concept. Adding the importance of the important word group of each sentence constituting the important word group having the same concept to obtain the importance of the important word group having the same concept,
Summary word selection control means for selecting, for each analysis item, the most important word group having the same concept as the summary word group of each analysis item;
For each analysis item, the template of the summary word of each analysis item is extracted from the summary word group of the analysis item by using a template selection rule in which a combination of the attribute of each summary word and the template of the summary sentence constituting the summary word group is defined in advance. A template selection means for selecting
For each analysis item, there is provided summary sentence generation means for generating a summary sentence for each analysis item by embedding a summary word group for each analysis item in a summary sentence template for each analysis item selected.
[0019]
The present invention (claims)8)A summary sentence generation device for generating a summary sentence of a document,
A sentence input means for inputting a document composed of a plurality of sentences including information indicating a sentence format;
A sentence structure analyzing means for extracting a plurality of sentences and a sentence type of each sentence from the input document using information indicating the format of the sentence and a structure definition rule in which the correspondence between the sentence types is defined in advance; ,
For each analysis item determined in advance, an extraction rule in which the relationship between the dependency relationship of the extracted important word, the attribute of the extracted important word and the certainty factor of the extracted important word attribute is defined in advance, A service important word extracting means for extracting a certainty level of an important word group by one or more important words of each sentence and an important word group of each sentence from the sentence;
HeavyThe sentence type of each sentence, important word attributes, and the conversion method of importance based on the certainty factor are used to select the sentence type of each sentence, the important word group of each sentence, and the important word of each sentence. Find the importance of key words in each sentence from the confidence of the group,
Extract words that characterize the input sentence from the sentence input from the outside, user intention inference knowledge defined in advance,
For each sentence, if the word related to the extracted word is included in the important word group of each sentence, increase the importance of the important word group including the related word,
For each analysis item, using a thesaurus knowledge in which the correspondence between the word and the concept of the word is defined in advance, the important word group of each sentence is grouped together for each having the same concept to be an important word group having the same concept, The importance of the important word group having the same concept is obtained by adding the importance of the important word group of each sentence constituting the important word group having the same concept,
Summary word selection control means for selecting, for each analysis item, the most important word group having the same concept as the summary word group of each analysis item;
For each analysis item, the template of the summary word of each analysis item is extracted from the summary word group of the analysis item by using a template selection rule in which a combination of the attribute of each summary word and the template of the summary sentence constituting the summary word group is defined in advance. A template selection means for selecting
For each analysis item, there is provided summary sentence generation means for generating a summary sentence for each analysis item by embedding a summary word group for each analysis item in a summary sentence template for each analysis item selected.
[0020]
The present invention (claims)9)A storage medium storing a summary sentence generation program for generating a summary sentence of a document,
A storage medium storing a program for causing a computer to execute processing for realizing the method according to claim 1.
[0028]
UpAs described above, in the present invention, the text structure is analyzed, the text type is determined, the attribute of the word appearing in the text is determined based on the dependency relationship for each extracted text type, and the text type is determined. The importance is determined by the word attributes. Then, an important word is extracted from the text according to the importance, a template corresponding to the extracted important word is selected, and a summary sentence is created. As described above, the present invention analyzes the type of text in consideration of the appearance position of the word by the structural analysis of the text, so that it is possible to specify that the word appearing at the head of the title or body is important. . It is also possible to specify that link information should be ignored.
[0029]
Also, when there are multiple guidance targets when determining the importance, the importance of words that are expressed in the same way is reduced with respect to the important words extracted for each text, and the user input information is Analyzes and performs processing similar to the expression existing in the input information of the user to reduce the importance, and analyzes the content of the text by default, such as the title, the same as the expression that exists in the default text content By applying a process that lowers the importance of the word that expresses, items that overlap with the title, items specified in the inspection condition, and items that are common to other text can be deleted from the summary sentence, so only useful words are summarized It becomes possible to make a sentence.
[0030]
In other words, it compares all the text contents searched and emphasizes the difference between the texts, for example, “care for bedridden elderly people and demented elderly people”, “care for elderly people living alone and low-income earners. It is possible to improve the selectivity of the user's service by highlighting and displaying the difference.
In addition, when there is a common expression with a plurality of other sentences, it is possible to collectively generate one summary sentence for the plurality of sentences.
[0031]
In addition, display information can be changed in accordance with user input information, and a summary sentence including information that would be useful to guide the user can be created. For example, if you find that the user is interested in the price of the service by searching for "cheap nursing home", you can create a summary sentence that the user wants by putting the charge information in the summary sentence It becomes.
[0032]
Further, since the size of the summary sentence to be dynamically displayed can be changed based on a display area predetermined for each device, it is possible to display accurate information for the user. For example, considering a guide area of 40 characters, if the summary sentence is long, the word is aggregated, and if the summary sentence is short, the display information of the summary sentence is increased. Information can be displayed.
[0033]
DETAILED DESCRIPTION OF THE INVENTION
FIG. 3 shows the configuration of the summary sentence creation apparatus of the present invention.
The summary sentence creating apparatus shown in FIG. 1 includes an input unit 10, a document structure analysis unit 20, a service important word extraction unit 30, a summary word selection control unit 40, a template selection unit 50, a summary sentence generation unit 60, an output unit 70, a structure. It comprises a definition 101, an extraction rule 102, a selection rule 103, a thesaurus knowledge 104, and a template selection rule 105.
[0034]
The input unit 10 inputs text and input information from the user. In the present invention, a corpus including tag information will be described as an example of input text. However, by setting knowledge, various classifications such as text can be made based on sentence structure information (arrangement, font, etc.). Therefore, it is only necessary to perform processing for analyzing text layout, fonts, and the like necessary for that purpose.
[0035]
The document structure analysis unit 20 includes a text title, a body, a table, etc.ofThe text type / structure is analyzed based on the structure definition 101 that defines the text type based on the appearance position and tag information. Specifically, in the input text, the appearance position of the sentence to be analyzed is taken into consideration, and as the appearance position, a serial number indicating what number the sentence is for each type is given. In the case of newspaper articles, etc.DifferentIt can be used when the importance level is to be set.
[0036]
The service important word extraction unit 30 performs morphological analysis for each text structure extracted by the document structure analysis unit 20, and follows the extraction rule 102 that defines the certainty of words appearing in the text based on the dependency relationship and the part of speech information. For the words in the text, the word attributes, confidence, and dependency information are set. In the present invention, the following method is adopted.
[0037]
(1) Perform morphological analysis on the input text and analyze the dependency relationship.
{Circle around (2)} The text dependency relation and the extraction rule 102 are collated, and a certainty factor is set for the collated thing. As an example of the matching method, there is a pattern matching or the like.
[0038]
The summary selection control unit 40 selects, for the words extracted by the service important word extraction unit 30, a selection rule 103 in which importance is defined based on text type, word attribute, and certainty, and a thesaurus in which the concept of the word is defined. Information of concepts closer to the knowledge 104 is integrated, importance is calculated, and control of determining important words based on importance and dependency relationships is controlled. The template selection unit 50 selects a template for creating a summary sentence by referring to the template selection rule 105 defined for the important word group based on the important word extracted by the summary selection control unit 40. To do.
[0039]
The summary sentence generation unit 60 creates a summary sentence by applying an important word to the template selected by the template selection unit 50.
The output unit 70 outputs the created summary sentence.
Here, the configuration of the summary word selection control unit 40 will be described.
FIG. 4 shows the configuration (part 1) of the summary word selection control unit of the present invention.
[0040]
The summary word selection control unit 40 shown in the figure includes a summary word selection unit 41 and an information suppression unit 42.
The summary word selection unit 41 selects, for the words extracted by the service important word extraction unit 30, a selection rule 103 that defines importance based on text type, word attributes, and certainty, and a thesaurus that defines the concept of the word. Information of concepts closer to the knowledge 104 is integrated, importance is calculated, important words are determined from the importance and dependency, and transferred to the information suppression unit 42.
[0041]
The information suppression unit 42 refers to the suppression word storage unit 106 that stores the result of analyzing the user's input by morphological analysis or the like and the result of analyzing the guidance content that is guided by default. The importance level of a word that performs the same expression as the expression that exists in the default expression and the expression that is guided by default is reduced, or a process of cutting by filtering is performed.
[0042]
Another configuration of the summary word selection control unit 40 will be described.
FIG. 5 shows the configuration (part 2) of the summary word selection control unit of the present invention.
The summary word selection control unit 40 shown in the figure includes a summary word selection unit 41 and a summary word comparison unit 43.
The summary word selection unit 41 refers to the summary word storage unit 107 that stores a result of a series of processing performed on each text when there are a plurality of texts for the determined important word. When a word having the same expression appears in a plurality of texts, the word is transferred to the summary word comparison unit 43.
[0043]
As a result, the summary word comparison unit 43 compares the attributes of similar words, analyzes the content that is guided by default, such as the title, and the importance of important words that perform the same expression as the expression that exists in the default guidance content. Decrease the degree or suppress the appearance of case information and part-of-speech information.
In addition, in the above example, when the summary word selection unit 41 acquires a word that expresses the same expression in a plurality of texts, it is transferred to the summary word comparison unit 43 in the above example. It is also possible to transfer and perform the above-described processing to obtain a template, and the summary sentence generation unit 60 can generate one summary sentence regarding a plurality of texts. That is, a summary sentence of a plurality of similar texts is created.
[0044]
Next, a case where the intention of the user is inferred and the importance of the expression related to the information that the user wants is increased will be described.
FIG. 6 shows the configuration (part 3) of the summary word selection control unit of the present invention.
The summary word selection control unit 40 shown in the figure includes a summary word selection unit 41, a user intention determination unit 44, a selection rule 103, a thesaurus knowledge 104, and user intention inference knowledge 108.
[0045]
The user intention determination unit 44 extracts a word that the user corresponding to the important word is intended by referring to the user intention inference knowledge 108 with respect to the important word included in the information input by the user, Transfer to summary word selector 41.
Thereby, the summary word selection part 41 gives importance to the word extracted in the user intention determination part 44. FIG. At this time, when there are a plurality of important words, the word having the highest importance is set as the important word.
[0046]
Next, a description will be given of a case where a predetermined guide area is determined on the display screen and editing is performed so that the summary text fits in the area.
FIG. 7 shows the configuration of the summary sentence generation unit of the present invention.
The summary sentence generated by the summary sentence generation unit 60 is transferred to the summary sentence editing unit 65. The summary sentence editing unit 65 refers to the display area information 109, compares the display section size, display information amount, display information attribute, and summary sentence shown in the guidance area, and the summary sentence is displayed in the guidance area. If the amount of information to be displayed is exceeded, the summary sentence is subjected to word aggregation, and if the summary sentence is smaller than a predetermined threshold of the amount of information shown in the guidance area, the display information (summary sentence) is increased. Do.
[0047]
【Example】
Embodiments of the present invention will be described below with reference to the drawings.
Hereinafter, a description will be given based on the configuration of FIGS.
First, rules and the like referred to in each component shown in FIG. 3 will be described.
FIG. 8 shows an example of the structure definition of one embodiment of the present invention.
[0048]
The structure definition 101 shown in the figure is used to extract the appearance location, importance, important word, and important word attribute of a word, and is referred to by the sentence structure analysis unit 20. It consists of an applicable document form, a rule, and a sentence type, and the sentence type selected from the rule is a type of content entered in addition to a table, a figure, a menu, etc., which is a type of sentence structure. Text, title, procedure, contact information, links, etc. can also be used.
[0049]
FIG. 9 shows an example of an extraction rule according to an embodiment of the present invention.
The extraction rule 102 shown in the figure defines the certainty of a word appearing in a sentence based on the dependency relationship and part-of-speech information, and is referenced by the service important word extraction unit 30 to determine the dependency relationship, attribute, and belief. Consists of degrees. As an example of the description of the extraction rule 102, the form of Japanese Patent Application No. 10-53889 can be used. In this method, by treating words belonging to the same representative classification node as the same, it is possible to determine the importance of the keyword based on the frequency even for a short sentence.
[0050]
Also, based on the case grammar, it is possible to use the place case, frequency, etc. of the verb. Furthermore, it is possible to describe a rule of defining 0 as the certainty factor and not extracting it.
FIG. 10 shows an example of a selection rule according to an embodiment of the present invention.
In the selection rule 103 shown in the figure, the importance is defined based on the sentence type, the word attribute, and the certainty factor, and is referred to by the summary word selection control unit 40. It consists of a selection rule and an importance. The importance is calculated as the importance for each word by performing four arithmetic operations on the certainty and integrating the results.
[0051]
FIG. 11 shows an example of thesaurus knowledge according to an embodiment of the present invention.
The thesaurus knowledge 104 shown in the figure defines the concept of words and is referred to by the summary word selection control unit 40. In addition to the meaning and hierarchical relationship of a word, the degree of abstraction and concreteness of the word itself are defined. In the case of verbs, case information is defined. FIG. 12 shows an example of a template according to an embodiment of the present invention.
[0052]
Template shown in the figureSelection ruleA summary word 105 is applied by the template selection unit 50 and used by the summary sentence generation unit 60. It consists of analysis items, extracted important words, conditions, and a plurality of template groups corresponding to these.
The operation of this embodiment will be described below.
[First embodiment]
FIGS. 13 and 14 show examples of HTML documents according to the first embodiment of the present invention, which are a series of HTML documents. The operation when this HTML document is input will be described below.
[0053]
FIG. 15 is a flowchart showing the operation of the first embodiment of the present invention.
Step 101) The sentence structure analysis unit 20 applies the structure definition 101 to the tag information and document information of the HTML document input from the input unit 10, assigns a sentence number to each sentence of the document, and assigns a sentence type. . Sentence numbers shall be divided in sentence units or, in the case of tables, in table attribute units. When products are arranged in parallel, they are divided by product. The analysis result of the input document is shown in FIG. In the example of the figure, sentence types are divided into “menu”, “title”, “text”, “table”, “procedure”, “contact”, and “link”, and a sentence number is assigned to each sentence type. .
[0054]
Step 102) Next, the service important word extraction unit 30 performs morphological analysis for each sentence number of each sentence type, and then processes the information by, for example, the method disclosed in Japanese Patent Laid-Open No. 10-269827. Based on the information, verb case information is extracted based on the extraction rule 102. The method disclosed in Japanese Patent Application Laid-Open No. 10-2669827 performs a morpheme / syntax analysis of a query sentence, divides it into word units, and if the inquiry sentence includes occupation name information, a knowledge base conversion table Refer to the above to convert the profession endings into verbs, and use the product name as the product name for the parts other than the endings of the profession name information. Search for. On the other hand, when the product name and verb are included in the inquiry sentence, the database is inferred from the product name and verb directly using the related knowledge and searching the database without referring to the conversion table. .
[0055]
There are various methods for summarizing the analysis results, but in this example, the “verb corresponding to the product and the subject corresponding to the verb”, “about the subject”, “about the procedure”, “about the contact” Four items were analyzed. In the case of information written in parallel, a certainty factor is given to each.
An example extracted by the service important word extraction unit 30 is shown in FIG.
[0056]
Step 103) Next, the summary word selection control unit 40 defines the importance using the selection rule 103 from the sentence type and the certainty shown in FIG. 17 acquired from the service important word extraction unit 30. An example of definition of importance by the summary word selection control unit 40 is shown in FIG.
Step 104) Further, in the summary word selection control unit 40, the thesaurus knowledge 104 is used to perform the process of merging records with the same concept unification and the same notation and combining the common concepts into one. At this time, the display omitted such as “etc.” is complemented.
[0057]
The result is shown in FIG.
Step 105) Further, the summary word selection control unit 40 selects a record having high importance for each analyzed item. The selected results are shown in FIG.
Step 106) The summary word selection control unit 40 supplements the information based on the case structure of the verb obtained as a result of the morphological analysis. In this example, it is understood that the verb “product and verb corresponding to the product, the subject corresponding to the verb”, “benefits / subsidies”, the part of speech is a noun and the meaning is human information. The information is supplemented by using the target person “About the target person”. The verb “Procedure”, “Submission”, is not supplemented because it shows that the part of speech is a noun and the meaning does not take human information.
[0058]
A supplemented example is shown in FIG.
Step 107) Next, the template selection unit 50 refers to the template 105 based on the selected word and selects an appropriate template. In this example, "About the product and the verb corresponding to the product, the target person corresponding to the verb"
“Verb” “product” to “subject”. ]
Select the template and “About Procedure”
“Please“ verb ”“ product ”to“ location case ”. 』
Select a template called "About Contacts"
“For details, contact“ Contact Department Name ”(telephone“ phone number ”). 』
Select the template.
[0059]
Step 108) Next, the summary sentence generation unit 60 processes elements to be applied to the template. Considering the generality of parts of speech using thesaurus knowledge 104, if an abstract expression is used, replace it with a concrete expression if possible. When the number of items is large, the latter half is cut and replaced with the expression “etc.”.
In this case, the information “Products and Verbs Corresponding to Commodities, Target Persons Corresponding to Verbs”
“Gas leak alarm, automatic fire detector, movement detector, automatic digester, mattress, posture changer”
And then
“Gas leak alarm, automatic fire detector, movement detector, etc.”
Converted to. Also,
“A bed-ridden elderly, elderly living alone, elderly with dementia”
But,
“A bedridden, living alone, elderly with dementia”
Converted to.
[0060]
Step 109) The summary sentence generator 60 selects a summary sentence to be output. In this case, depending on the importance,
“Products and Verbs Corresponding to Products, Target Persons Corresponding to Verbs”
Is selected.
Step 110) The output unit 70 outputs a summary sentence. In this case,
“We provide gas subsidies, automatic fire detectors, mobile detectors, etc., living alone, providing benefits and subsidies for elderly with dementia”
Is output.
[0061]
[Second Embodiment]
Next, in this embodiment, in the case of step 103 in the flowchart of the first embodiment described above, the information to be guided to the user and the user's input are always compared, and the importance is lowered, based on the configuration of FIG. I will explain.
For example, if you always display the title,
[Title: Home helper dispatch
Body: Home helpers are dispatched to families with people who are over 65 years old who have difficulty in daily life because of physical or mental disabilities.
[0062]
The main services are laundry, cleaning, shopping, meal preparation, personal care and consultation. However, you will be responsible for some expenses depending on the income from the previous year. As a result of using morphological analysis results,
From the title,
"Product: Home helper (person) Corresponding verb: Dispatch importance: 9",
From the text,
"Product: Home helper (person) Corresponding verb: Dispatch Importance: 6",
Products: Laundry, cleaning, shopping, meal preparation, personal care, consultation, etc. (Things) Importance: 6,
Product: Cost (thing) Verb: Burden Importance 1 ”
It becomes. If exactly the same thing as the title is extracted as a combination of the product and the verb, the importance level of the information is set to 0, and then the process proceeds to Step 104 merging. After that, proceed as a template selection rule.
“If the product is not a verb and the meaning of the product is“ thing ”, then“ product (thing) ”is done”
As a result,
“Washing, cleaning, shopping, etc.”
The summary sentence is obtained.
[0063]
In the present embodiment, when the title is necessarily displayed, the importance of the word in the title is lowered, but the same applies to the user input, and the importance of the word in accordance with the user input can be lowered.
[Third embodiment]
In the present embodiment, an example in which a plurality of pieces of service information are compared in step 103 of the above-described first embodiment to increase the importance of expressions different from others will be described with reference to FIG.
[0064]
The input sentence shown in FIG. 22 is input as the input sentence, and the contents shown in FIG. 23 to FIG. 23 (a) are obtained from the analysis processing in steps 101 and 102, and the contents shown in FIG. Is extracted.
Here, (1) in FIG. 23 (a) and (4) in FIG. 23 (b), (3) in FIG. 23 (a) and (6) in FIG. Decrease the degree greatly. In this example, it is “−4”. Further, (2) in FIG. 23 (a) and FIG. 23 (b) (5) are expressed in which only one record is different among the record contents, and therefore the degree of importance is lowered. In this example, “−2” is set.
[0065]
As a result, the importance is set as shown in FIG. 24, and then the process proceeds to merging in step 104 based on the information in each of the text 1 and text 2.
After that, the process proceeds and the template selection rule is that the verb is “deposit”, the product has the meaning of “person”, and the “location case” and “time case”
“I will leave the“ subject ”at“ time ”at“ place ”
As a result,
From body 1
“I will leave demented elderly at a nursing home at night”
From text 2,
`` I will leave elderly people with dementia in nursing homes during the day ''
Is obtained.
[0066]
[Fourth embodiment]
In step 103 of the third embodiment described above, information on a plurality of services is compared to reduce the importance of the same expression as the other, and the appearance of case information and part-of-speech information is suppressed for the corresponding record. An example will be described with reference to FIGS.
As in the third embodiment described above, when the input shown in FIG. 22 is performed, the results shown in FIG.
[0067]
Here, compare the text 1 and text 2 with the same “verb” and “product”, and if the contents are all the same, the importance of the record is rejected as “0”, and even if the contents are partially If they are different, the verb information is left, only the case information that is different is left, and other case information and part-of-speech information are discarded.
as a result,
In text 1,
(2) Verb: Keeping time: Daytime Importance: 4
Set
In text 2,
(5) Verb: Keeping time: Night Importance: 4
Set. Thereafter, the merge processing in step 104 is performed based on the information of the text 1 and the text 2.
[0068]
After that, proceed as a template selection rule.
“If there is no“ product ”and“ verb ”is stored, and there is a“ time case ”, it is stored in“ time ”. ]
As a result,
From body 1
`` I will keep it at night ''
From text 2,
"We will keep it during the day"
The summary sentence is obtained.
[0069]
[Fifth embodiment]
In the present embodiment, when the summary word selection unit 41 acquires words that perform the same expression in a plurality of texts based on the configuration shown in FIG. An example of generating one summary sentence will be described. That is, one summary sentence is generated from a plurality of texts by comparing information of a plurality of services and collecting the same expressions as others.
[0070]
In FIG. 5, it is assumed that the summary word selection unit 41 refers to the summary word storage unit 107 that stores a series of processing results performed on each text for each text and acquires the following contents.
From the text 1 stored in the summary word storage unit 107,
(1) Product: Care (thing) Verb: Difficult Target: Demented elderly Importance: 6
(2) Product: Demented elderly (person) Verb: Keeping place: Nursing home Time: Daytime importance: 6
(3) Product: Burden Verb: Reduction Target: Caregiver Importance: 6
From text 2,
(1) Product: Care (thing) Verb: Difficult Target: Demented elderly Importance: 6
▲ 2 ▼ Product: Demented elderly (person) Verb: Keeping Location: Nursing home Time: Night Importance: 6
(3) Product: Burden Verb: Reduction Target: Caregiver Importance: 6
Is extracted.
[0071]
Here, the summary word comparing unit 43 compares the body 1 and the body 2 in which the verb and the product match, and if the contents are all the same, the summary word selecting unit 41 summarizes the record contents into a plurality of sentences. The template selection unit 50 is notified of the following contents so as to be adopted as
Product: Demented elderly (People) Verb: Keep Location: Nursing home
In the template selection unit 50, referring to the template selection rule 105,
“If the item is“ person ”, the verb is“ deposited ”, and there is a place case,“ item (person) ”will be kept in“ location ”. ]
This rule is applied.
[0072]
In the summary sentence creation unit 60, using this template,
“I will leave elderly people with dementia in nursing homes”
One summary sentence is generated.
[Sixth embodiment]
In the present embodiment, an example in which the user's intention is inferred from the input information input by the user and the importance of the expression related to the information desired by the user is increased will be described with reference to FIG.
[0073]
When the summary word selection control unit 40 acquires the important word extracted by the important word extraction unit 30, the user intention determination unit 44 searches the user intention inference knowledge 108 based on the important word.
In this embodiment, the input information input by the user is
"I want you to take care of my grandpa cheaply"
When searching the database,
“Title: Nursing care service
Body: We will take care of elderly people with dementia at home in nursing homes.
[0074]
Number of implementations: The service can be used up to 30 times a year.
Offer time: From 9:00 am to 5:00 pm
User burden: 1,050 yen / time
Is retrieved as a search result and processed by the syntax analysis unit 20 and the keyword extraction unit 30,
・ About the product and the verb corresponding to the product, the target person for the verb:
(1) Product: Demented elderly (person) Verb: Keep Location: Nursing home Importance: 6
・ About the frequency of product implementation:
(2) Number of times: 30 times a year Importance: 2
・ About product time:
(3) Time: 9 am-5 pm Importance: 2
・ About the price of the product:
(4) Price: 1050 yen / time Importance: 2
Can be extracted. Here, the user intention determination unit 44 refers to the intention reasoning knowledge 108 based on the input information from the user “I want you to take care of my grandpa cheaply”. In this case, since the keyword “cheap” is included in the input information, it is determined that price information is important for the user, and the importance of the price information is calculated.
[0075]
Here, the importance of the information corresponding to the input information from the user is tripled
・ About the price of the product:
(4) Price: 1050 yen / time Importance: 6
It becomes.
As a result, the summary word selection unit 41
(1) Product: Demented elderly (person) Verb: Keep Location: Nursing home Importance: 6
(2) Number of times: 30 times a year Importance: 2
(3) Time: 9 am-5 pm Importance: 2
(4) Price: 1050 yen / time Importance: 6
Is extracted. These pieces of information are transferred to the template selection unit 50 at the next stage.
[0076]
In the template selection unit 50, as the template selection rule 105, the verb is “deposit”, the product has the meaning of “person”, “location case”, “time case”, “number of times”, “price”, “ If you have time information,
“We keep the“ target person ”at“ place ”at“ price ”for“ number of times ”and“ time ”. 』
Apply the rule. Thereby, the summary sentence generation unit 60 uses the template,
“Demented elderly” will be kept at “Nursing Home” at “1050 yen / time” between “30 times a year” and “9 am-5pm”. ]
Is generated and output.
[0077]
[Seventh embodiment]
Next, in the present embodiment, an example will be described in which the summary sentence generated by the summary sentence generation unit 60 is edited based on predetermined display area information 109 based on FIG.
First, as summary candidate sentences in the summary sentence generator 60,
“Demented elderly people are deposited at a nursing home at 1050 yen / time 30 times a year, from 9 am to 5 pm. ]
Is created.
[0078]
Here, referring to the display area information 109, since the total number of characters is 46, it is assumed that the condition “within 30 characters” is not satisfied, and the above summary candidate sentence is processed.
Here, the summary sentence editing unit 65 is selected by the template selection unit 50.
“We keep the“ target person ”at“ place ”at“ price ”for“ number of times ”and“ time ”. 』
See the importance of each element in the template,
“Subject”: Demented elderly Importance: 6
“Number of times”: 30 times a year Importance: 2
"Time": 9 am-5 pm Importance: 2
"Price": 1050 yen / time Importance: 6
"Place": Retirement home Importance: 6
From the “number of times” and “time” with low importance, “time” with a large number of characters is deleted. as a result,
“Subject”: Demented elderly Importance: 6
“Number of times”: 30 times a year Importance: 2
"Price": 1050 yen / time Importance: 6
"Place": Retirement home Importance: 6
Here, when the template is selected again by the template selection unit 50, the verb is “deposit”, the product has the meaning of “person”, the location case, the time case, the number of times, and the price information Is
“We will keep the“ target person ”at“ location ”with“ number of times ”and“ price ”. ]
In the summary sentence generation unit 60, the template is again selected.
“Demented elderly people are deposited at nursing homes at 1050 yen / time 30 times a year.”
A 34-character summary sentence candidate is created. Similarly, referring to the display area information 109, the condition “within 30 characters” is not satisfied, and the summary sentence candidate is processed again. Here, it is assumed that the “number of times” is deleted based on the above-described importance. As a result,
“I will leave demented elderly at a nursing home at ¥ 1050 / time.”
(Read: Chiho Seikosha will be sent to you by your teacher)
Get. This summary sentence candidate becomes and satisfies the contents of the display area information 109.
[0079]
The above example is an example of processing because the number of characters is too large. However, when the number of characters is small, a plurality of sentences are selected, or the information omitted when processing in the first embodiment is used. It is possible to perform processing such as The determination that the number of characters is small can be made by setting a threshold value such as when the number of characters determined in the display area information 109 is 80% or less.
[0080]
In the above embodiment, the description is based on the flowchart shown in FIG. 15. However, the flowchart shown in FIG. 15 is constructed as a program, and a disk device connected to a computer used as a summary creation device or a floppy disk. The present invention can be easily realized by storing it in a portable storage medium such as a CD-ROM and installing it when carrying out the present invention.
[0081]
The present invention is not limited to the above-described embodiments, and various modifications and applications can be made within the scope of the claims.
[0082]
【The invention's effect】
As described above, according to the present invention, the word appearing at the beginning of the title or the body is recognized as important by analyzing the type of text in consideration of the appearance position of the word. Since the items specified in the inspection condition and items common to other guidance sentences can be deleted from the summary sentence, the summary sentence can be created using only useful words.
[0083]
Further, since it is possible to emphasize differences between a plurality of texts, it is possible to provide an effective summary sentence for selecting information.
Further, by inferring the user's intention based on the input information from the user, a summary sentence meeting the user's needs can be created.
In addition, when there is a common expression among a plurality of texts, common information between the texts can be extracted and representative summary sentences, not individual summary sentences, can be created together.
[0084]
In addition, when there is a display area restriction or display time restriction on the device, it is possible to change the summary sentence dynamically according to the area and time without losing the main part. Can present an accurate summary sentence.
In this way, in the present invention, the mobile terminal with a limited display information area is used to analyze the default guidance information such as the title and the input of the user and suppress the appearance of words that perform the same expression in the summary sentence. This is effective for creating a summary sentence in a search engine or a yellow page service that searches for useful information when a lot of search results are obtained.
[0085]
In order to emphasize and display differences between sentences such as a plurality of guidance information, the user's information can be increased in choice by displaying it together with the search result in a search engine or a yellow page service.
[Brief description of the drawings]
FIG. 1 is a diagram for explaining the principle of the present invention.
FIG. 2 is a principle configuration diagram of the present invention.
FIG. 3 is a configuration diagram of a summary sentence creation device of the present invention.
FIG. 4 is a block diagram (No. 1) of a summary word selection control unit of the present invention.
FIG. 5 is a configuration diagram (part 2) of the summary word selection control unit of the present invention;
FIG. 6 is a configuration diagram (part 3) of the summary word selection control unit of the present invention;
FIG. 7 is a configuration diagram of a summary sentence generation unit of the present invention.
FIG. 8 is an example of a structure definition according to an embodiment of the present invention.
FIG. 9 is an example of an extraction rule according to an embodiment of the present invention.
FIG. 10 is an example of a selection rule according to an embodiment of the present invention.
FIG. 11 is an example of thesaurus knowledge according to an embodiment of the present invention.
FIG. 12 is an example of a template according to an embodiment of the present invention.
FIG. 13 is an example (part 1) of an HTML document according to the first embodiment of this invention;
FIG. 14 is an example (No. 2) of the HTML document according to the first embodiment of this invention;
FIG. 15 is a flowchart showing the operation of the first exemplary embodiment of the present invention.
FIG. 16 shows the first embodiment of the present invention.Document structureIt is an example of an analysis result.
FIG. 17 is an example of a result extracted by a service important word extraction unit according to the first exemplary embodiment of the present invention;
FIG. 18 is an example of importance setting according to the first embodiment of this invention.
FIG. 19 is an example of concept unification / merging / common concept unification processing by the summary word selection control unit of the first exemplary embodiment of the present invention;
FIG. 20 is an example of record selection processing with high importance by the summary word selection control unit according to the first embodiment of this invention;
FIG. 21 is an example of information supplementation in the summary word selection control unit of the first embodiment of the present invention;
FIG. 22 is an example of an input sentence according to the third embodiment of this invention.
FIG. 23 is a result of the type assignment / important word extraction of the third example of the present invention.
FIG. 24 is an example of assigning importance in the third embodiment of the present invention.
FIG. 25 is an example of the result of type assignment / keyword extraction in the fourth embodiment of the present invention;
[Explanation of symbols]
10Document input means,Input section
20 sentence structure analysis means, sentence structure analysis part
30 key word extraction means, service key word extraction unit
40 summary word selection means, summary word selection control unit
41 Summary word selector
42 Information suppression part
43 Summary Word Comparison Unit
44 Summary word selector
50 Template selection means, template selection section
60 Summary sentence generator, summary sentence generator
65 Summary sentence editor
70 Output section
101 Structure definition
102 Extraction rules
103 Selection rules
104 Thesaurus knowledge
105 Template selection rules
106 Suppression word storage
107 Summary word storage
108 User intention reasoning knowledge
109 Display area information

Claims (9)

A summary sentence generation method in a summary sentence generation apparatus for generating a summary sentence of a document , comprising:
In sentence structure analysis means,
When a document consisting of a plurality of sentences including information indicating the sentence format is input by the sentence input means,
A sentence structure analysis process for extracting a plurality of sentences and sentence types of each sentence from the input document using a structure definition rule in which a correspondence between information indicating a sentence format and a sentence type is defined in advance;
In service important word extraction means,
For each predetermined analysis item, by using an extraction rule in which a correlation between a dependency relationship of an extracted important word and an attribute of the extracted important word and a certainty factor of the extracted important word attribute is defined in advance, A service important word extraction process that extracts the certainty of the important word group of each sentence from one sentence and one or more important words of each sentence and the important word group of each sentence;
In summary word selection control means,
Using a selection rule in which an importance conversion method is defined in advance from a sentence type in which an important word appears, an attribute of the important word, and a certainty factor, a sentence type of each sentence, an important word group of each sentence, and each sentence The importance of the important word group of each sentence is calculated from the certainty of the important word group of
The important word group of each sentence matches the word inputted from the outside or the important word group extracted in the service important word extraction process from the sentence of the sentence type specified in the inputted document in advance. To reduce the importance of matched key words,
For each of the analysis items, using a thesaurus knowledge in which a correspondence between a word and a concept of the word is defined in advance, the important word group of each sentence is grouped for each having the same concept, and the important word having the same concept And determining the importance of the important word group having the same concept by adding the importance of the important word group of each sentence constituting the important word group having the same concept,
A summary word selection control process for selecting, for each analysis item, a keyword group having the greatest importance among the important word groups having the same concept as a summary word group of each analysis item;
In the template selection means,
For each analysis item, a summary sentence of each analysis item is extracted from the summary word group of the analysis item using a template selection rule in which a set of attributes of the summary words and a template of the summary sentence constituting the summary word group is defined in advance. A template selection process for selecting a template,
In the summary sentence generation means,
For each analysis item, a summary sentence generation process for generating a summary sentence of each analysis item by embedding a summary word group of each analysis item in the summary sentence template of each selected analysis item;
The summary sentence generation method characterized by performing . A summary sentence generation method in a summary sentence generation apparatus for generating a summary sentence of a document,
In sentence structure analysis means,
When a document composed of a plurality of sentences including information indicating the sentence format is input by the sentence input means, the structure definition rule in which the correspondence between the information indicating the sentence format and the sentence type is defined in advance is used. A sentence structure analysis process of extracting a plurality of sentences and sentence types of each sentence from the input document;
In service important word extraction means,
For each predetermined analysis item, by using an extraction rule in which a correlation between a dependency relationship of an extracted important word and an attribute of the extracted important word and a certainty factor of the extracted important word attribute is defined in advance, A service important word extraction process that extracts the certainty of one or more important words of each sentence from each sentence and the certainty of the important words of each sentence;
In summary word selection control means,
Using a selection rule in which an importance conversion method is defined in advance from a sentence type in which an important word appears, an attribute of the important word, and a certainty factor, a sentence type of each sentence, an important word group of each sentence, and each sentence The importance of the important word group of each sentence is obtained from the certainty of the important word group of
Important word group of the sentence respectively, complete with each sentence of the key word group of other documents extracted in said service key word extraction process and the sentence structure analysis process from another document or partially if match Lower the importance of the matched key words group,
For each of the analysis items, using a thesaurus knowledge in which a correspondence between a word and a concept of the word is defined in advance, the important word group of each sentence is grouped for each having the same concept, and the important word having the same concept And determining the importance of the important word group having the same concept by adding the importance of the important word group of each sentence constituting the important word group having the same concept,
For each analysis item, a summary word selection control process for selecting the most important word group having the same concept as the summary word group for each analysis item;
In the template selection means,
For each analysis item, a summary sentence of each analysis item is extracted from the summary word group of the analysis item using a template selection rule in which a set of attributes of the summary words and a template of the summary sentence constituting the summary word group is defined in advance. A template selection process for selecting a template,
In the summary sentence generation means,
For each analysis item, a summary sentence generation process for generating a summary sentence of each analysis item by embedding a summary word group of each analysis item in the summary sentence template of each selected analysis item;
The summary sentence generation method characterized by performing . A summary sentence generation method in a summary sentence generation apparatus for generating a summary sentence of a document,
In sentence structure analysis means,
When a document composed of a plurality of sentences including information indicating the sentence format is input by the sentence input means, the structure definition rule in which the correspondence between the information indicating the sentence format and the sentence type is defined in advance is used. A sentence structure analysis process of extracting a plurality of sentences and sentence types of each sentence from the input document;
In service important word extraction means,
For each predetermined analysis item, the relationship between the dependency relationship of the important noun to be extracted, the correspondence between the important noun to be extracted and the certainty of the important noun to be extracted, and the dependency relationship between the important verb to be extracted and the important verb to be extracted A service for extracting the confidence level of the important word group of each sentence and one or more important nouns and important verbs of each sentence and the important word group of each sentence from each sentence using an extraction rule in which the association with the sentence is defined in advance Important word extraction process,
In summary word selection control means,
Using the selection rule in which the conversion method of importance is pre-defined from the attribute type and the certainty of the sentence type in which the important noun appears and the certainty level, the sentence type of each sentence, the important noun of each sentence, and each of the above From the certainty of the key word group of the sentence, the importance of the key word group of each sentence is determined,
Important nouns and important verbs in the important word group of each sentence are important nouns of the important word group of each sentence of the other document extracted from the other document in the sentence structure analysis process and the service important word extraction process. If the key noun and the key verb are exactly the same, the key word group in which the key noun and key verb are completely matched is rejected. Reject important nouns in each important word group of each sentence from the important word group,
For each of the analysis items, using a thesaurus knowledge in which a correspondence between a word and a word concept is defined in advance, the important word group having the same concept by grouping the important word groups of each sentence into those having the same concept And adding the importance of the important word group of each sentence constituting the important word group having the same concept to obtain the importance of the important word group having the same concept,
A summary word selection control process for selecting, for each analysis item, a keyword group having the greatest importance among the important word groups having the same concept as a summary word group of each analysis item;
In the template selection means,
For each analysis item, a summary sentence of each analysis item is extracted from the summary word group of the analysis item using a template selection rule in which a set of attributes of the summary words and a template of the summary sentence constituting the summary word group is defined in advance. A template selection process for selecting a template,
In the summary sentence generation means,
For each analysis item, a summary sentence generation process for generating a summary sentence of each analysis item by embedding a summary word group of each analysis item in the summary sentence template of each selected analysis item;
The summary sentence generation method characterized by performing . A summary sentence generation method in a summary sentence generation apparatus for generating a summary sentence of a document,
In sentence structure analysis means,
When a document composed of a plurality of sentences including information indicating the sentence format is input by the sentence input unit, information indicating the sentence format and a structure definition rule in which a correspondence between the sentence types is defined in advance are used. , A sentence structure analysis process for extracting a plurality of sentences and sentence types of each sentence from the input document;
In service important word extraction means,
For each predetermined analysis item, by using an extraction rule in which the relationship between the dependency relationship of the extracted important word and the attribute of the extracted important word and the certainty factor of the extracted important word attribute is defined in advance, A service important word extraction process that extracts the certainty of the important word group of each sentence from one sentence and one or more important words of each sentence and the important word group of each sentence;
In summary word selection control means,
Using a selection rule in which the conversion method of importance is pre-defined from the sentence type in which the important word appears, the attribute of the important word, and the certainty factor, the sentence type of each sentence, the important word group of each sentence, and each sentence The importance of the important word group of each sentence is calculated from the certainty of the important word group of
Extracting words that characterize the inputted sentence from the sentence inputted from the outside, user intention reasoning knowledge defined in advance,
For each of the sentences, when a word related to the extracted word is included in the important word group of each sentence, increase the importance of the important word group including the related word,
For each of the analysis items, using a thesaurus knowledge in which a correspondence between a word and a word concept is defined in advance, the important word group of the sentences is grouped for each having the same concept, and the important word group having the same concept And calculating the importance of the important word group having the same concept by adding the importance of the important word group of each sentence constituting the important word group having the same concept,
A summary word selection control process for selecting, for each analysis item, a keyword group having the greatest importance among the important word groups having the same concept as a summary word group of each analysis item;
In the template selection means,
For each analysis item, a summary sentence of each analysis item is extracted from the summary word group of the analysis item using a template selection rule in which a set of attributes of the summary words and a template of the summary sentence constituting the summary word group is defined in advance. A template selection process for selecting a template,
In the summary sentence generation means,
For each analysis item, a summary sentence generation process for generating a summary sentence of each analysis item by embedding a summary word group of each analysis item in the summary sentence template of each selected analysis item;
The summary sentence generation method characterized by performing . A summary sentence generation device for generating a summary sentence of a document,
A sentence input means for inputting a document composed of a plurality of sentences including information indicating a sentence format;
A sentence structure analyzing means for extracting a plurality of sentences and a sentence type of each sentence from the input document using a structure definition rule in which a correspondence between information indicating a sentence format and a sentence type is predefined;
For each predetermined analysis item, by using an extraction rule in which a correlation between a dependency relationship of an extracted important word and an attribute of the extracted important word and a certainty factor of the extracted important word attribute is defined in advance, Service important word extraction means for extracting the confidence level of the important word group of each sentence from one sentence or one or more important words of each sentence and the important word group of each sentence;
Using a selection rule in which an importance conversion method is defined in advance from a sentence type in which an important word appears, an attribute of the important word, and a certainty factor, a sentence type of each sentence, an important word group of each sentence, and each sentence The importance of the important word group of each sentence is obtained from the certainty of the important word group of
The important word group of each sentence matches the word inputted from the outside or the important word group extracted by the service important word extracting means from the sentence of the sentence type designated in advance in the inputted document. To reduce the importance of matched key words,
For each analysis item, using a thesaurus knowledge in which a correspondence between a word and a word concept is defined in advance, the important word group of each sentence is grouped for each having the same concept, and the important concept having the same concept The importance of the important word group having the same concept is obtained by adding the importance of the important word group of each sentence constituting the important word group having the same concept as the word group,
Summary word selection control means for selecting, as the summary word group of each analysis item, the one having the greatest importance among the important word groups having the same concept for each analysis item;
For each analysis item, a summary sentence of each analysis item is extracted from the summary word group of the analysis item using a template selection rule in which a set of attributes of the summary words and a template of the summary sentence constituting the summary word group is defined in advance. A template selection means for selecting a template of
For each analysis item, summary sentence generation means for generating a summary sentence of each analysis item by embedding a summary word group of each analysis item in the summary sentence template of each selected analysis item;
The summary sentence generation device characterized by having. A summary sentence generation device for generating a summary sentence of a document,
A sentence input means for inputting a document composed of a plurality of sentences including information indicating a sentence format;
A sentence structure analyzing means for extracting a plurality of sentences and a sentence type of each sentence from the input document using a structure definition rule in which a correspondence between information indicating a sentence format and a sentence type is predefined;
For each predetermined analysis item, by using an extraction rule in which the relationship between the dependency relationship of the extracted important word and the attribute of the extracted important word and the certainty factor of the extracted important word attribute is defined in advance, Service important word extraction means for extracting the certainty of the important word group of each sentence from one sentence or one or more important words of each sentence and the important word group of each sentence;
Using a selection rule in which an importance conversion method is defined in advance from a sentence type in which an important word appears, an attribute of the important word, and a certainty factor, a sentence type of each sentence, an important word group of each sentence, and each sentence The importance of the important word group of each sentence is obtained from the certainty of the important word group of
When the important word group of each sentence matches completely or partially with the important word group of each sentence of the other document extracted from the other document by the sentence structure analyzing unit and the service important word extracting unit. Lower the importance of the matched key words group,
For each of the analysis items, using a thesaurus knowledge in which a correspondence between a word and a concept of the word is defined in advance, the important word group of each sentence is grouped for each having the same concept, and the important word having the same concept And determining the importance of the important word group having the same concept by adding the importance of the important word group of each sentence constituting the important word group having the same concept,
Summary word selection control means for selecting, as the summary word group of each analysis item, the one having the greatest importance among the important word groups having the same concept for each analysis item;
For each analysis item, a summary sentence of each analysis item is extracted from the summary word group of the analysis item using a template selection rule in which a set of attributes of the summary words and a template of the summary sentence constituting the summary word group is defined in advance. A template selection means for selecting a template of
For each analysis item, summary sentence generation means for generating a summary sentence of each analysis item by embedding a summary word group of each analysis item in the summary sentence template of each selected analysis item;
  The summary sentence generation device characterized by having. A summary sentence generation device for generating a summary sentence of a document,
A sentence input means for inputting a document composed of a plurality of sentences including information indicating a sentence format;
Sentence structure analysis means for extracting a plurality of sentences and sentence types of each sentence from the input document using a structure definition rule in which a correspondence between information indicating a sentence format and a sentence type is predefined;
For each predetermined analysis item, the relationship between the dependency relationship of the important noun to be extracted, the correspondence between the important noun to be extracted and the certainty of the important noun to be extracted, and the dependency relationship between the important verb to be extracted and the important verb to be extracted A service for extracting the confidence level of the important word group of each sentence and one or more important nouns and important verbs of each sentence and the important word group of each sentence from each sentence using an extraction rule in which the association with the sentence is defined in advance Important word extraction means,
Using the selection rule in which the conversion method of the importance from the sentence type and the important noun attribute and the certainty of the important noun appear in advance, the sentence type of each sentence, the important noun of each sentence, and each of the above From the certainty of the key word group of the sentence, the importance of the key word group of each sentence is determined,
Important nouns and important verbs in the important word group of each sentence are extracted from other documents by the sentence structure analyzing means and the service important word extracting means. If the key noun and the key verb are exactly the same, the key word group in which the key noun and key verb are completely matched is rejected. Reject important nouns in each important word group of each sentence from the important word group,
For each of the analysis items, using a thesaurus knowledge in which a correspondence between a word and a word concept is defined in advance, the important word group having the same concept by grouping the important word groups of each sentence into those having the same concept And adding the importance of the important word group of each sentence constituting the important word group having the same concept to obtain the importance of the important word group having the same concept,
Summary word selection control means for selecting, as the summary word group of each analysis item, the one having the greatest importance among the important word groups having the same concept for each analysis item;
For each analysis item, a summary sentence of each analysis item is extracted from the summary word group of the analysis item using a template selection rule in which a set of attributes of the summary words and a template of the summary sentence constituting the summary word group is defined in advance. A template selection means for selecting a template of
For each analysis item, summary sentence generation means for generating a summary sentence of each analysis item by embedding a summary word group of each analysis item in the summary sentence template of each selected analysis item;
  The summary sentence generation device characterized by having. A summary sentence generation device for generating a summary sentence of a document,
A sentence input means for inputting a document composed of a plurality of sentences including information indicating a sentence format;
Sentence structure analysis means for extracting a plurality of sentences and a sentence type of each sentence from the input document using information indicating a sentence format and a structure definition rule in which association with a sentence type is defined in advance When,
For each predetermined analysis item, by using an extraction rule in which a correlation between a dependency relationship of an extracted important word and an attribute of the extracted important word and a certainty factor of the extracted important word attribute is defined in advance, Service important word extraction means for extracting the confidence level of the important word group of each sentence from one sentence or one or more important words of each sentence and the important word group of each sentence;
Using a selection rule in which an importance conversion method is defined in advance from a sentence type in which an important word appears, an attribute of the important word, and a certainty factor, a sentence type of each sentence, an important word group of each sentence, and each sentence The importance of the important word group of each sentence is obtained from the certainty of the important word group of
Extracting words that characterize the inputted sentence from the sentence inputted from the outside, user intention reasoning knowledge defined in advance,
For each of the sentences, when a word related to the extracted word is included in the important word group of each sentence, increase the importance of the important word group including the related word,
For each of the analysis items, using a thesaurus knowledge in which a correspondence between a word and a concept of the word is defined in advance, the important word group having the same concept by grouping the important word groups of each sentence having the same concept And calculating the importance of the important word group having the same concept by adding the importance of the important word group of each sentence constituting the important word group having the same concept,
Summary word selection control means for selecting, as the summary word group of each analysis item, the one having the greatest importance among the important word groups having the same concept for each analysis item;
For each analysis item, a summary sentence of each analysis item is extracted from the summary word group of the analysis item using a template selection rule in which a set of attributes of the summary words and a template of the summary sentence constituting the summary word group is defined in advance. A template selection means for selecting a template of
For each analysis item, summary sentence generation means for generating a summary sentence of each analysis item by embedding a summary word group of each analysis item in the summary sentence template of each selected analysis item;
The summary sentence generation device characterized by having. A storage medium storing a summary sentence generation program for generating a summary sentence of a document,
5. A storage medium storing a summary sentence generation program, wherein a program for causing a computer to execute processing for realizing the method according to claim 1 is stored.

Images