JP7091146B2 - Information processing equipment, information processing methods, and programs - Google Patents

Description

The present invention relates to an information processing apparatus, an information processing method, and a program.

Patent Document 1 describes a technique for extracting keywords (summary words) by using various information in a document, particularly information at a parsing result level under a unified framework.

Japanese Unexamined Patent Publication No. 8-44763

Here, in a sentence, whether the subject is important or the object is important depends on the sentence. For example, in the sentence "○○ TV has won the triple crown of audience rating", the subject "○○ TV" is an important summary word. On the other hand, in the sentence "○○ TV reported the election of Mr. △△", the subject "○○ TV" cannot be said to be an important summary word. In this way, even if the clause contains the same particle, the importance of the clause differs depending on the sentence.
However, in the technique described in Patent Document 1, the weight of the word contained in the phrase is determined based on the particle of the phrase, but whether the subject is important or the object is important depends on the sentence. .. Therefore, the technique described in Patent Document 1 has a problem that an appropriate summary word may not be extracted depending on the sentence.

The present invention has been made in view of the above points, and an object of the present invention is to provide an information processing apparatus, an information processing method, and a program capable of extracting more appropriate summary words from a sentence.

(1) The present invention has been made to solve the above-mentioned problems, and one aspect of the present invention includes an acquisition unit for acquiring a document composed of one or a plurality of sentences and each word included in the sentence. , A basic score assigning unit that weights according to its case element, a score updating unit that updates the weight of each word included in the sentence according to the predicate in the sentence, and a score updating unit that updates the weight of each word, and the document based on the weight of each word. It is an information processing apparatus including a summary word selection unit for selecting a summary word and an output unit for outputting the selected summary word.

(2) Further, one aspect of the present invention includes the information processing apparatus according to (1), which includes a summary word integration unit for summing the weights of words regarded as the same.

(3) Further, one aspect of the present invention is the information processing apparatus according to (1) or (2), and the score updating unit updates weights according to different rules according to the category of the document.

(4) Further, one aspect of the present invention is the information processing apparatus according to any one of (1) to (3), and the basic score giving unit is the subject of the sentence when the sentence is active. If the sentence is passive, the weight of the object of the sentence is increased.

(5) Further, one aspect of the present invention is the information processing apparatus according to (4), and when a specific predicate is included in a sentence, the score updating unit has a weight and an object of the subject of the sentence. Reverse the weight of the word.

(6) Further, one aspect of the present invention is the information processing apparatus according to (5), wherein the specific predicate is a verb meaning an act of providing information, and the score updating unit is a score update unit. When the specific predicate is included in a sentence, the weight of the information provided in the sentence is increased.

(7) Further, one aspect of the present invention is the information processing apparatus according to (6), and the specific predicate differs depending on the category of the document.

(8) Further, the present invention has been made to solve the above-mentioned problems, and one aspect of the present invention is a process in which the acquisition unit acquires a document consisting of one or a plurality of sentences, and a basic score is given. A process in which the unit weights each word contained in the document according to its case element, and a process in which the score updating unit updates the weight of each word contained in the sentence according to the predicate in the sentence. This is an information processing method including a process in which a summary word selection unit selects a summary word in the document based on the weight of each word, and a process in which an output unit outputs the selected summary word.

(9) Further, the present invention has been made to solve the above-mentioned problems, and one aspect of the present invention includes a step of a computer acquiring a document composed of one or a plurality of sentences and the above-mentioned sentence. A step of weighting each word according to its case element, a step of updating the weight of each word contained in the sentence according to the predicate in the sentence, and a summary of the document based on the weight of each word. It is a program for executing a step of selecting a word and a step of outputting the selected summary word.

According to the present invention, more appropriate abstracts can be extracted from a document.

It is a schematic block diagram which shows an example of the functional structure of the summary word generation apparatus which concerns on 1st Embodiment of this invention. It is a schematic diagram which shows the data structure and the data example of the basic score rule table stored in the basic score rule storage part which concerns on 1st Embodiment of this invention. It is a schematic diagram which shows the data structure and the data example of the specific predicate list table stored in the specific predicate list storage part which concerns on 1st Embodiment of this invention. It is a schematic diagram which shows the data structure and the data example of the score update rule table stored in the score update rule storage part which concerns on 1st Embodiment of this invention. It is a schematic block diagram which shows an example of the hardware composition of the summary word generation apparatus which concerns on 1st Embodiment of this invention. It is a figure for demonstrating the process which the abstract word generator which concerns on 1st Embodiment of this invention analyzes a morpheme and a syntax of a sentence. It is a figure for demonstrating the process which gives the score by the summary word generation apparatus which concerns on 1st Embodiment of this invention. It is a figure for demonstrating the process which gives the score by the summary word generation apparatus which concerns on 1st Embodiment of this invention. It is a figure for demonstrating the process which gives the score by the summary word generation apparatus which concerns on 1st Embodiment of this invention. It is a flowchart which shows an example of the summary word generation process executed by the summary word generation apparatus which concerns on 1st Embodiment of this invention. It is a schematic block diagram which shows an example of the functional structure of the summary word generation apparatus which concerns on 2nd Embodiment of this invention. It is a schematic diagram which shows the data structure and the data example of the category specific predicate list table stored in the category specific predicate list storage part which concerns on 2nd Embodiment of this invention. It is a figure for demonstrating the process which gives the score by the summary word generation apparatus which concerns on 2nd Embodiment of this invention. It is a figure for demonstrating the process which gives the score by the summary word generation apparatus which concerns on 2nd Embodiment of this invention. It is a flowchart which shows an example of the summary word generation process executed by the summary word generation apparatus which concerns on 2nd Embodiment of this invention.

(First Embodiment)
Hereinafter, the first embodiment of the present invention will be described with reference to the drawings.
FIG. 1 is a schematic block diagram showing an example of the functional configuration of the summary word generation device 1 according to the first embodiment of the present invention.
The summary word generation device 1 is an information processing device that generates and outputs a summary word of the input document from an input document composed of one or a plurality of sentences. The input document is, for example, a news text published on the Internet or the like. The summary word is an important word in an input document such as a keyword, and is a word that summarizes the input document. That is, the summary word generation device 1 generates a summary word that expresses the subject of the sentence in one word from the input document consisting of one or more sentences. This makes it possible to generate a message such as "News about XX" from a news article, for example. The summary word may be one word, a plurality of words, or a combination of a plurality of words.

The summary word generation device 1 is, for example, a server device to which a plurality of terminal devices can be connected, and outputs a summary word of an input document to the connected terminal devices. Alternatively, the summary word generation device 1 may be incorporated in each terminal device such as a mobile phone, a smartphone, a tablet terminal, or a personal computer. For example, when introducing news published on the Internet, the terminal device uses the summary word "○○" of the news acquired from the summary word generation device 1, and "is news about 〇〇." Etc. can be provided.

The summary word generation device 1 includes an input unit 11, a control unit 12, a storage unit 13, and an output unit 14.
The input unit 11 is an interface for inputting an input document, such as a keyboard, a microphone, and a network interface. The input unit 11 outputs the input input document to the control unit 12. The input unit 11 may acquire input text from a network or another device.

The control unit 12 controls the summary word generation device 1 in an integrated manner. The control unit 12 includes a morpheme / syntax analysis unit 121, a basic score giving unit 122, a score updating unit 123, a summary word integration unit 124, and a summary word selection unit 125.

The morpheme / syntax analysis unit 121 analyzes the morpheme and syntax of each sentence in the input document based on the analysis dictionary stored in the analysis dictionary storage unit 131, and obtains the part of speech and dependency information of each phrase in each sentence. It is given to the input document and output to the basic score giving unit 122.
The basic score giving unit 122 weights (hereinafter, also referred to as “score”) the words of each phrase included in the input document according to the case element. A phrase is, for example, a phrase in which a case particle is removed from a case element. Specifically, the basic score giving unit 122 assigns a score to each phrase word based on the basic score rule table stored in the basic score rule storage unit 132. For example, the basic score giving unit 122 may increase the weight of the subject of the sentence when the sentence is active, and may increase the weight of the object of the sentence when the sentence is passive. The basic score giving unit 122 outputs the score given to each word and the input document to the score updating unit 123.

The score updating unit 123 updates the weight of the word of each clause in each sentence according to the predicate of each sentence included in the input document. Specifically, the score update unit 123 is stored in the score update rule table stored in the score update rule storage unit 134 when the specific predicate in the specific predicate list stored in the specific predicate list storage unit 133 is a sentence predicate. Update the word score for each phrase in the sentence based on it. A specific predicate is, for example, a verb that means an act of providing information. The specific predicate is not limited to the verb meaning the act of providing information, and may be another verb. When the specific predicate is included in the sentence, the score updating unit 123 reverses the weight of the subject and the weight of the object of the sentence. That is, when the specific predicate is included in the sentence, the score updating unit 123 increases the weight of the information provided in the sentence. In other words, when the specific predicate is included in the sentence, the score update unit 123 updates the weighting of the information provided in the sentence. The score updating unit 123 outputs the updated score and the input document to the summary word integration unit 124.

When there are a plurality of sentences in the input document, the summary word integration unit 124 compares each sentence and adds up the scores for similar words that can be regarded as the same, and the summed up score and the input document are combined with the summary word selection unit 125. Output to.
The summary word selection unit 125 selects the summary word of the input document based on the score of each word, and outputs the selected summary word to the output unit 14. Specifically, the summary word selection unit 125 sorts all the words in descending order of the score, and selects the word with the highest score as the summary word. In addition, when there are a plurality of words having the highest score, the summary word selection unit 125 may select the word having the highest basic score first given by the basic score giving unit 122, or may be the first in the input document. You may choose a word or you may choose it randomly.

The storage unit 13 includes an analysis dictionary storage unit 131, a basic score rule storage unit 132, a specific predicate list storage unit 133, and a score update rule storage unit 134, and stores various information.

The analysis dictionary storage unit 131 stores an analysis dictionary for analyzing the morpheme and syntax of the document.
The basic score rule storage unit 132 stores a basic score rule table showing basic score rules when assigning a score.

The specific predicate list storage unit 133 stores a specific predicate list indicating a specific predicate.
The score update rule storage unit 134 stores a score update rule table showing a score update rule when updating the score.

The output unit 14 is an interface that outputs summary words, such as a display, a speaker, and a network interface. The output unit 14 outputs the abstract word input from the abstract word selection unit 125.

FIG. 2 is a schematic diagram showing a data structure and a data example of a basic score rule table stored in the basic score rule storage unit 132 according to the first embodiment of the present invention.
As shown in the figure, the basic score rule table has each item of sentence state, phrase case, and score. The voice of a sentence indicates whether the sentence is active or passive. The case of a phrase indicates a case particle of the phrase. The score indicates the weight attached to the word of the phrase.

In the illustrated example, the score corresponding to the sentence state "active voice" and the phrase case "ga case" is "1.0". In addition, the score corresponding to the sentence state "active voice" and the phrase case "wo case" is "0.8". In addition, the score corresponding to the sentence state "active voice" and the phrase case "second case" is "0.5". The score corresponding to the sentence state "active voice" and the phrase case "predicate" is "0.0". In addition, the score corresponding to the sentence state "passive voice" and the phrase case "ga case" is "0.8". The score corresponding to the sentence state "passive voice" and the phrase case "second case" is "1.0".

FIG. 3 is a schematic diagram showing a data structure and a data example of a specific predicate list table stored in the specific predicate list storage unit 133 according to the first embodiment of the present invention.
As shown in the figure, the specific predicate list table has a category and each item of the specific predicate list. A category is a category of documents. The specific predicate list is a list of specific predicates. A specific predicate is a verb that means an act of providing information by media such as newspapers, television, and the Internet. In the illustrated example, the specific predicate list corresponding to the category "all" includes predicates such as "post, introduce, report, ...".

FIG. 4 is a schematic diagram showing a data structure and a data example of a score update rule table stored in the score update rule storage unit 134 according to the first embodiment of the present invention.
As shown in the figure, the score update rule table has each item of sentence state, phrase case, and score. The voice of a sentence indicates whether the sentence is active or passive. The case of a phrase indicates a case particle of the phrase. The score indicates the change value of the score, "+" indicates an addition point, and "-" indicates a deduction point.

In the illustrated example, the score corresponding to the sentence state "active voice" and the phrase case "ga case" is "-0.5". In addition, the score corresponding to the sentence state "active voice" and the phrase case "wo case" is "+0.5". In addition, the score corresponding to the sentence state "active voice" and the phrase case "second case" is "-0.5". The score corresponding to the sentence state "active voice" and the phrase case "predicate" is "0.0". In addition, the score corresponding to the sentence state "passive voice" and the phrase case "ga case" is "+0.5". The score corresponding to the sentence state "passive voice" and the phrase case "second case" is "-0.5".

That is, in the score update rule of the illustrated example, if the predicate of the sentence is a specific predicate, the case element that is supposed to include the media name (case in the case of active voice, case in the case of passive voice). Is deducted, and the case elements that are supposed to contain the information transmitted by the media (Wo case in the case of the ability and Ga case in the case of the passive voice) are added.

FIG. 5 is a schematic block diagram showing an example of the hardware configuration of the summary word generation device 1 according to the first embodiment of the present invention.
The summary word generation device 1 includes a CPU 101, a storage medium interface unit 102, a storage medium 103, an input device 104, an output device 105, a ROM 106 (Read Only Memory), a RAM 107 (Random Access Memory), and auxiliary storage. A unit 108 and a network interface unit 109 are provided. The CPU 101, the storage medium interface unit 102, the input device 104, the output device 105, the ROM 106, the RAM 107, the auxiliary storage unit 108, and the network interface unit 109 are connected to each other via a bus.
The CPU 101 referred to here indicates a processor in general, and includes not only a device called a so-called CPU in a narrow sense but also a GPU, a DSP, and the like. Further, the CPU 101 referred to here is not limited to being realized by one processor, and may be realized by combining a plurality of processors of the same or different types.

The CPU 101 reads and executes a program stored in the auxiliary storage unit 108, ROM 106, and RAM 107, reads various data stored in the auxiliary storage unit 108, ROM 106, and RAM 107, and reads various data to the auxiliary storage unit 108, RAM 107. The summary word generation device 1 is controlled by writing. Further, the CPU 101 reads various data stored in the storage medium 103 via the storage medium interface unit 102, and writes various data in the storage medium 103. The storage medium 103 is a portable storage medium such as a magneto-optical disk, a flexible disk, and a flash memory, and stores various data.
The storage medium interface unit 102 is an interface for reading and writing the storage medium 103.

The input device 104 is an input device such as a mouse, a keyboard, a touch panel, a volume control button, a power button, a setting button, and an infrared receiver.
The output device 105 is an output device such as a display unit and a speaker.
The ROM 106 and the RAM 107 store programs and various data for operating each functional unit of the summary word generation device 1.
The auxiliary storage unit 108 is a hard disk drive, a flash memory, or the like, and stores programs and various data for operating each functional unit of the summary word generation device 1.
The network interface unit 109 has a communication interface and is connected to the network NW by wireless communication.

For example, the control unit 12 in the functional configuration of the summary word generation device 1 corresponds to the CPU 101, and the storage unit 13 corresponds to the ROM 106, the RAM 107, the auxiliary storage unit 108, or a combination thereof, and is an input unit. 11 corresponds to the input device 104 or the network interface unit 109, and the output unit 14 corresponds to the output device 105 or the network interface unit 109.

Subsequently, the operation of the summary word generation device 1 will be described in detail with reference to specific examples. First, the morpheme / syntax analysis unit 121 analyzes the morpheme and syntax of each sentence included in the input document.

FIG. 6 is a diagram for explaining a process in which the summary word generation device 1 according to the first embodiment of the present invention analyzes a morpheme and a syntax of a sentence. In the illustrated example, the input sentence is "○○ TV reported the election of Mr. △△." First, the morpheme / syntax analysis unit 121 divides the input sentence into the clauses "○○ TV ga", "△△ -san's", "winning" and "reported". Then, the morpheme / syntax analysis unit 121 classifies each phrase into the case elements "○○ TV ga", "△△ -san's", "winning" and the predicate "reported". The morpheme / parsing unit 121 assigns information on the case type (proprietary noun, sa-variable verb, etc.) and the case of the phrase to the case element, and assigns information on the word type and sentence state to the predicate. .. In the illustrated example, "○○ TV ga" is "ga-case / proper noun (company name)", "△△ -san's" is "no-case / proper noun (personal name)", and "winning". Is "wo case / general noun", and "reported" is "predicate / sa verb / active voice". Further, the morpheme / syntax analysis unit 121 analyzes the dependency information between clauses and expresses it in a tree structure. In the illustrated example, "○○ TV" is related to "reported", "△△ -san's" is related to "winning", and "winning" is related to "reporting".

Next, the basic score giving unit 122 and the score updating unit 123 assign scores to the words of each phrase. 7 to 9 are diagrams for explaining the process of assigning a score by the summary word generation device 1 according to the first embodiment of the present invention.

In the example shown in FIG. 7, the input document is "○○ TV has won the audience rating triple crown." That is, the input document does not indicate that the media provides information. First, as shown in FIG. 7 (A), the morpheme / parsing unit 121 states that "○○ TV ga" in the input document is "ga case / common noun (company name)" and "viewing rate triple crown". ”Is a“ wo case / common noun ”, and“ acquired ”is a“ predicate / sa verb / ability dynamic ”.

Subsequently, as shown in FIG. 7 (B), the basic score giving unit 122 adds the basic score “1. "0" is given, and the basic score "0.8" is given to the word "audience rating triple crown" of the wo case phrase "audience rating triple crown", and the predicate phrase "acquired" is given to the word "acquisition". A basic score of "0.0" is given.

Subsequently, the score updating unit 123 determines whether or not the predicate "acquired" is in the specific predicate list. Since the predicate "acquired" is not in the specific predicate list, the score update unit 123 sets all score updates, which are values to be updated from the basic score, to "0.0". That is, the score updating unit 123 does not update the score. As a result, the total score for each word remains the basic score. Therefore, the total score of the word "○○ TV" is "1.0", the total score of the word "audience rating triple crown" is "0.8", and the total score of the word "acquisition" is "0. 0 ". Finally, the summary word selection unit 125 selects the word "OO TV" having the highest total score as the summary word of the input document "OO TV has won the triple crown of audience rating." That is, if the sentence does not indicate that the media provides information, the summary word generation device 1 outputs the subject "○○ TV" as the summary word.

On the other hand, in the example shown in FIG. 8, the input document is "○○ TV reported the election of Mr. △△." That is, the input document indicates that the media provides information. First, as shown in FIG. 8 (A), in the morpheme / syntax analysis unit 121, "○○ TV ga" in the input document is "ga case / common noun (company name)", and "Mr. △△ was elected. It is analyzed that "o" is "wo case / common noun (personal name) + common noun" and "reported" is "predicate / sa verb / ability dynamic".

Subsequently, as shown in FIG. 8 (B), the basic score giving unit 122 adds the basic score “1. "0" is given, and the basic score "0.8" is given to the word "△△ -san's winning" of the wo-case phrase "△△ -san's winning", and the predicate phrase "reported" is given to the word "reporting". Is given a basic score of "0.0".

Subsequently, the score updating unit 123 determines whether or not the predicate "reported" is in the specific predicate list. Since the predicate "reported" is in the specific predicate list, the score update unit 123 adds the score update "-0.5" to the basic score "1.0" of "OO TV" based on the score update rule table. Then, the total score is updated to "0.5", and the score update "+0.5" is added to the basic score "0.8" of the word "△△ -san's winning" to make the total score "1.3". Update and update the total score to "0.0" by adding the score update "0.0" to the basic score "0.0" of the word "report".

As a result, before the update, the subject "○○ TV" had the highest basic score, but after the update, the object "Mr. △△ won" has the highest total score. Therefore, the summary word selection unit 125 selects the word "Mr. △△'s winning" with the highest total score as the summary word of the input document "○○ TV reported the winning of Mr. △△." That is, the summary word generation device 1 outputs "Mr. △△'s winning", which is the information reported by the media "○○ TV", as a summary word. Therefore, the summary word generation device 1 can generate and output a more appropriate summary word for the input document.

In the example shown in FIG. 9, the input document is "Mr. A handed a letter to Mr. B. Mr. B introduced the letter." First, as shown in FIG. 9 (A), the morpheme / syntax analysis unit 121 says that "Mr. A" in the first sentence of the input document "Mr. A handed a letter to Mr. B." "Professional noun (personal name)", "Mr. B" is "second case / proper noun (personal name)", "letter" is "wokaku / common noun", and "passed" is "predicate" Analyze that it is "/ verb / noun". Further, as shown in FIG. 9B, in the morpheme / syntax analysis unit 121, "Mr. B" in the second sentence "Mr. B introduced the letter" of the input document is "Ga case / proper noun". It is analyzed that "(person's name)", "that letter" is "wo-case / demonstrative noun + ordinary noun", and "introduced" is "predicate / syntax verb / ability dynamic".

Subsequently, as shown in FIG. 9 (C), the basic score giving unit 122 adds the basic score "Mr. A" to the word "Mr. A" in the predicate phrase "Mr. A" in the first sentence based on the basic score rule table. "1.0" is given, the basic score "0.5" is given to the word "Mr. B" of the second case phrase "Mr. B", and the basic score "Letter" is given to the word "letter" of the genitive phrase "letter". A score of "0.8" is given, and a basic score of "0.0" is given to the word "pass" of the predicate phrase "passed". Further, since the predicate "passed" in the first sentence is not in the specific predicate list, the score update unit 123 sets all the score updates in the first sentence to "0.0" and does not update the score in the first sentence. As a result, the total score for each word remains the basic score.

Further, as shown in FIG. 9 (D), the basic score giving unit 122 adds the basic score "Mr. B" to the word "Mr. B" in the second sentence based on the basic score rule table. "1.0" is given, the basic score "0.8" is given to the word "that letter" of the wo case phrase "that letter", and the basic score is given to the word "introduction" of the predicate phrase "introduced". Give "0.0".

Subsequently, the score update unit 123 updates the score to the basic score "1.0" of "Mr. B" based on the score update rule table because the predicate "introduced" in the second sentence is in the specific predicate list. Add -0.5 "to update the total score to" 0.5 ", update the score to the basic score" 0.8 "of the word" the letter ", add" +0.5 "to add the total score to" 1. Update to "3", add score update "0.0" to the basic score "0.0" of the word "introduction", and update the total score to "0.0".

Subsequently, as shown in FIG. 9 (E), the summary word integration unit 124 compares the first sentence and the second sentence, adds up the scores of the words "Mr. B" that can be regarded as the same, and totals the scores. Is set to "1.0", and the scores of the word "letter" are added up to make the total score "2.1". This increases the score of words that appear frequently in the input document. Finally, the summary word selection unit 125 selects the word "letter" having the highest total score as the summary word of the input document "Mr. A handed the letter to Mr. B. Mr. B introduced the letter." In this way, when the input document has a plurality of sentences, the summary word generator 1 sums up the scores of the words considered to be the same to select the summary word, so that the important words frequently appearing in the input document are summarized. Can be selected as a word.

FIG. 10 is a flowchart showing an example of a summary word generation process executed by the summary word generation device 1 according to the first embodiment of the present invention.
In step S101, the input document is input to the input unit 11.
In step S103, the morpheme / syntax analysis unit 121 analyzes the input document based on the analysis dictionary.

In step S105, the basic score giving unit 122 assigns a score to each sentence element included in the input document based on the basic score rule table. Sentence elements are the words of each phrase.
In step S107, the score updating unit 123 determines whether or not the specific predicate in the specific predicate list is used in the sentence included in the input document. When the specific predicate is used (step S107; YES), the summary word generator 1 proceeds to the process of step S109. On the other hand, when the specific predicate is not used (step S107; NO), the summary word generator 1 proceeds to the process of step S111.

In step S109, the score update unit 123 updates the score of the sentence element of the sentence in which the specific predicate is used, based on the score update rule table.

In step S111, when the input document has a plurality of sentences, the summary word integration unit 124 compares the sentences and adds up the scores of similar words that can be regarded as the same.
In step S113, the summary word selection unit 125 selects the word with the highest score as the summary word.
In step S115, the output unit 14 outputs a summary word. After that, the process ends.

As described above, the summary word generation device 1 according to the first embodiment has an input unit 11 (acquisition unit) for acquiring a sentence composed of one or a plurality of sentences, and each word included in the document has its rating element. A basic score giving unit 122 that weights according to the sentence, a score updating unit 123 that updates the weight of each word included in the sentence according to the predicate in the sentence, and a summary word of the document selected based on the weight of each word. It includes a summary word selection unit 125 and an output unit 14 that outputs the selected summary word.

With such a structure, it is possible to extract an appropriate summary word even when the weight of a clause containing the same particle changes depending on the predicate. That is, a more appropriate summary word can be extracted from the document.

Further, the summary word generation device 1 includes a summary word integration unit 124 that totals the weights of words that are regarded as the same. With such a structure, it is possible to increase the weight of words that frequently appear in a document.

Further, the basic score giving unit 122 increases the weight of the subject of the sentence when the sentence is dynamic, and the basic score giving unit 122 increases the weight of the object of the sentence when the sentence is passive. When a specific predicate is included in a sentence, the weight of the subject and the weight of the object of the sentence are reversed. More specifically, the specific predicate is a verb meaning an act of providing information, and the score updater 123, when the specific predicate is included in the sentence, weights the information provided in the sentence. Make it heavy. With such a structure, in the case of a sentence in which the media or the like provides information, the provided information can be extracted as a summary word.

(Second embodiment)
Hereinafter, a second embodiment of the present invention will be described with reference to the drawings.
Even sentences that use the same predicate may have different case elements that are appropriate summary words depending on the category of the input document. The category indicates the genre to which the sentence belongs (for example, performing arts, sports, politics, etc.). For example, in the sentence "Mr. OO announced her marriage" whose category is "Entertainment", the subject "Mr. OO" is appropriate as a summary word, while the sentence "Mr. OO" whose category is "Politics" is appropriate. In "The Prime Minister announced the △△ tax reduction.", The subject "△△ tax reduction" is appropriate as a summary word. Therefore, the present embodiment differs from the first embodiment in that the specific predicate differs depending on the category of the input document.

FIG. 11 is a schematic block diagram showing an example of the functional configuration of the summary word generation device 1A according to the second embodiment of the present invention. The summary word generation device 1A is an information processing device that generates and outputs a summary word of the document from an input document composed of one or a plurality of sentences. The summary word generation device 1A is, for example, a server device to which a plurality of terminal devices can be connected, and outputs a summary word of an input document to the connected terminal devices. Alternatively, the summary word generation device 1A may be incorporated in each terminal device such as a mobile phone, a smartphone, a tablet terminal, or a personal computer. Since the hardware configuration of the abstract word generation device 1A is the same as that of the first embodiment, the description thereof will be omitted. Hereinafter, the same components as those in the first embodiment are designated by the same reference numerals, and the description thereof will be omitted.

The summary word generation device 1A includes an input unit 11A, a control unit 12A, a storage unit 13A, and an output unit 14. The input unit 11A is an interface for inputting an input document together with its category, such as a keyboard, a microphone, and a network interface. The input unit 11A outputs the input input document and its category to the control unit 12A.

The control unit 12A controls the summary word generation device 1A in an integrated manner. The control unit 12A includes a morpheme / syntax analysis unit 121, a basic score giving unit 122, a score updating unit 123A, a summary word integration unit 124, and a summary word selection unit 125.

The score updating unit 123A updates the weight of each phrase in each sentence according to the predicate of each sentence included in the input document according to different rules according to the category of the input document. Specifically, the score updating unit 123A reads a specific predicate list according to the category of the input document from the category-specific specific predicate list storage unit 133A, and the specific predicate included in the read specific predicate list is a sentence predicate. In this case, the score of the word of each clause of the sentence is updated based on the score update rule table stored in the score update rule storage unit 134. The score updating unit 123A outputs the updated score and the input document to the summary word integration unit 124.

The storage unit 13A includes a dictionary storage unit 131 for analysis, a basic score rule storage unit 132, a category-specific specific predicate list storage unit 133A, and a score update rule storage unit 134, and stores various information. The category-specific specific predicate list storage unit 133A stores a specific predicate list indicating a specific predicate for each category.

FIG. 12 is a schematic diagram showing a data structure and a data example of a category-specific predicate list table stored in the category-specific predicate list storage unit 133A according to the second embodiment of the present invention.
As shown in the figure, the category-specific predicate list table has a category and each item of the specific predicate list. A category is a category of documents. The specific predicate list is a list of specific predicates.
In the illustrated example, the specific predicate list corresponding to the category "Entertainment" includes predicates such as "report, publish, ...". In addition, the specific predicate list corresponding to the category "sports" includes predicates such as "achieve, acquire, ...". In addition, the list of specific predicates corresponding to the category "politics" includes predicates such as "publish, announce, ...". The specific predicate list of each category may include the same predicate in a plurality of categories.

Subsequently, the operation of the summary word generation device 1A will be described in detail with reference to a specific example. 13 and 14 are diagrams for explaining the process of assigning a score by the summary word generator 1A according to the second embodiment of the present invention.

In the example shown in FIG. 13, the input document is "Mr. XX announced her marriage." The category is performing arts. First, as shown in FIG. 13 (A), in the morpheme / parsing unit 121, "○○ san ga" is "ga case / common noun (personal name)" and "marriage" is "wo" in the input document. It is analyzed that it is a case / common noun and that "announced" is a "predicate / sa verb / ability dynamic".

Subsequently, as shown in FIG. 13 (B), the basic score giving unit 122 adds the basic score “1. A basic score of "0.8" is given to the word "marriage" of the wo case phrase "marriage", and a basic score of "0.0" is given to the word "announcement" of the predicate phrase "announced". Is given.

Subsequently, the score updating unit 123A determines whether or not the predicate "announced" is in the specific predicate list of the input document category "entertainment". Since the predicate "announced" is not in the specific predicate list of the category "entertainment", the score update unit 123A sets the score update of all words to "0.0" and does not update the score of each word. As a result, the total score for each word remains the basic score. Therefore, the summary word selection unit 125 selects the word "Mr. XX" having the highest total score as the summary word of the input document "Mr. XX announced the marriage."

On the other hand, in the example shown in FIG. 14, the input document is "Prime Minister XX has announced a tax reduction of △△", and the category is politics. First, as shown in FIG. 14 (A), the morpheme / parsing unit 121 states that "○○ Prime Minister ga" is "ga case / common noun (personal name)" in the input document, and "△△ tax reduction". Is "wo case / common noun", and "announced" is "predicate / sa verb / ability dynamic".

Subsequently, as shown in FIG. 14 (B), the basic score giving unit 122 adds the basic score “1. "0" is given, and the basic score "0.8" is given to the word "△△ tax reduction" of the phrase "△△ tax reduction", and the word "announcement" of the predicate phrase "announced". A basic score of "0.0" is given.

Subsequently, the score updating unit 123A determines whether or not the predicate "announced" is in the specific predicate list of the input document category "politics". Since the predicate "announced" is in the specific predicate list of the category "politics", the score update unit 123A updates the score to the basic score "1.0" of "Prime Minister XX" based on the score update rule table. Add "0.5" to update the total score to "0.5", add the score update "+0.5" to the basic score "0.8" of the word "△△ tax reduction", and add the total score to "1". Update to "0.3", add score update "0.0" to the basic score "0.0" of the word "announcement", and update the total score to "0.0".

As a result, the subject "Prime Minister" had the highest basic score before the update, while the object "△△ tax reduction" had the highest total score after the update. Therefore, the summary word selection unit 125 selects the word "△△ tax reduction" having the highest total score as the summary word of the input document "Prime Minister XX has announced the △△ tax reduction."

As described above, since the summary word generator 1A uses a different list of specific predicates according to the category of the input document, even if the sentence has the same predicate "to announce", the summary word suitable for each input document can be obtained. It can be generated and output.

FIG. 15 is a flowchart showing an example of the summary word generation process executed by the summary word generation device 1A according to the second embodiment of the present invention.
In step S201, the input document and its category are input to the input unit 11A.
Since the processing of steps S203 to S205 is the same as the processing of steps S103 to S105 in the first embodiment, the description thereof will be omitted.

In step S207, the score updating unit 123A determines whether or not the sentence included in the input document uses a specific predicate in the specific predicate list corresponding to the category of the input document. When the specific predicate is used (step S207; YES), the summary word generator 1A proceeds to the process of step S209. On the other hand, when the specific predicate is not used (step S207; NO), the summary word generator 1A proceeds to the process of step S211.

Since the processing of steps S209 to S215 is the same as the processing of steps S109 to S115 in the first embodiment, the description thereof will be omitted.

In the present embodiment, the specific predicate list is different depending on the category of the input document, but the present invention is not limited to this, and the score may be updated by a different score update rule according to the category of the input document.

As described above, the summary word generation device 1A according to the second embodiment has an input unit 11 (acquisition unit) for acquiring a document composed of one or a plurality of sentences, and each word included in the sentence has its rating element. The basic score giving unit 122 that weights according to the sentence, the score updating unit 123A that updates the weight of each word included in the sentence according to the predicate in the sentence, and the summary word of the document are selected based on the weight of each word. It includes a summary word selection unit 125 and an output unit 14 that outputs the selected summary word. The score update unit 123A updates the weight according to different rules according to the document category. More specifically, the particular predicate that updates the weights depends on the category of the document.

With such a structure, it is possible to extract an appropriate summary word even when the weight of a clause containing the same particle changes depending on the predicate. Further, even if the sentence uses the same predicate, the weight is updated according to the category of the document, so that the summary word suitable for the document can be extracted. That is, a more appropriate summary word can be extracted from the document.

Although the embodiments of the present invention have been described in detail with reference to the drawings, the specific configuration is not limited to the above, and various design changes and the like can be made without departing from the gist of the present invention. It is possible to do.

For example, in the above-described embodiment, the summary word generators 1 and 1A update the score according to the same score update rule when the predicate of the sentence is a specific predicate, but the score is not limited to this, and the score is updated for each specific predicate. Scores may be updated with different score update rules.

Further, in the above-described embodiment, the summary word generators 1 and 1A output one word as a summary word, but the present invention is not limited to this, and a predetermined plurality of words may be output in descending order of score.

Further, in the above-described embodiment, the case where the input document is in Japanese is shown as an example, but the present invention is not limited to this, and other languages may be used.

Further, the summary word generation devices 1 and 1A may accept the input of the evaluation for the output summary word and change the score update rule by machine learning based on the evaluation.

In addition, since the summary word generators 1 and 1A are often more important than the preceding sentence, when there are a plurality of sentences in the input document, a score is given so that the score of the preceding sentence is high. You may.

Further, in the above-described embodiment, in the score update rule, in the case of the ability, the points are deducted from the ga-rank and the second-rank, and the wo-rank is added. It suffices to deduct points that are likely to include media names such as Kara, and add points that are likely to include information transmitted by the media such as Tok and Nikku. Similarly, in the score update rule, in the case of passive voice, a point is added to the ga rating and a point is deducted from the second rating. It suffices to deduct points from the case elements and add points to the case elements that are likely to contain information transmitted by the media, such as the To case and the Nikku case.

It should be noted that one or more of the above-described embodiments and modifications, and a part or all of the above-mentioned embodiments and modifications may be combined to realize one aspect of the present invention.

It should be noted that the program operating on the summary word generators 1 and 1A in one aspect of the present invention may be one or the program so as to realize the functions shown in the above embodiments and modifications related to the one aspect of the present invention. It may be a program (a program that makes a computer function) that controls a plurality of processors such as a CPU (Central Processing Unit). Then, the information handled by each of these devices is temporarily stored in a RAM (Random Access Memory) at the time of processing, and then stored in various storages such as a flash memory and an HDD (Hard Disk Drive), if necessary. It may be read, corrected and written by a CPU or the like.

It should be noted that a part or all of the summary word generation devices 1 and 1A in each of the above-described embodiments and modifications may be realized by a computer provided with one or a plurality of processors. In that case, a program for realizing this control function may be recorded on a recording medium readable by a computer, and the program recorded on the recording medium may be read by a computer system and executed.

The term "computer system" as used herein is a computer system built into the summary word generators 1 and 1A, and includes hardware such as an OS and peripheral devices. Further, the "computer-readable recording medium" refers to a portable medium such as a flexible disk, a magneto-optical disk, a ROM, or a CD-ROM, and a storage device such as a hard disk built in a computer system.

Further, a "computer-readable recording medium" is a medium that dynamically holds a program for a short time, such as a communication line when a program is transmitted via a network such as the Internet or a communication line such as a telephone line. In that case, a program may be held for a certain period of time, such as a volatile memory inside a computer system serving as a server or a client. Further, the above program may be for realizing a part of the above-mentioned functions, and may be further realized for realizing the above-mentioned functions in combination with a program already recorded in the computer system.

Further, a part or all of the summary word generators 1 and 1A in each of the above-described embodiments and modifications may be realized as an LSI, which is typically an integrated circuit, or may be realized as a chipset. .. Further, each functional block of the summary word generation devices 1 and 1A in each of the above-described embodiments and modifications may be individually chipped, or a part or all of them may be integrated into a chip. Further, the method of making an integrated circuit is not limited to the LSI, and may be realized by a dedicated circuit and / or a general-purpose processor. Further, when an integrated circuit technology that replaces an LSI appears due to advances in semiconductor technology, it is also possible to use an integrated circuit based on this technology.

As described above, as one aspect of the present invention, each embodiment and modification have been described in detail with reference to the drawings, but the specific configuration is not limited to each embodiment and modification, and deviates from the gist of the present invention. It also includes design changes that are not included. Further, one aspect of the present invention can be variously modified within the scope of the claims, and the technical aspects of the present invention can also be obtained by appropriately combining the technical means disclosed in the different embodiments. Included in the range. Further, the elements described in each of the above-described embodiments and modifications are included, and a configuration in which elements having the same effect are replaced with each other is also included.

1,1A Predicate generator 11, 11A Input unit 12, 12A Control unit 121 Predicate / parsing unit 122 Basic score assignment unit 123, 123A Score update unit 124 Summary word integration unit 125 Predicate selection unit 13, 13A Storage unit 131 Parsing dictionary storage unit 132 Basic score rule storage unit 133 Specific predicate list storage unit 133A Category-specific specific predicate list storage unit 134 Score update rule storage unit 14 Output unit 101 CPU
102 Storage medium interface unit 103 Storage medium 104 Input device 105 Output device 106 ROM
107 RAM
108 Auxiliary storage unit 109 Network interface unit

Claims (9)

An acquisition unit that acquires a document consisting of one or more sentences,
A basic scoring unit that weights each word contained in the document according to its case element,
A score updater that updates the weight of each word included in the sentence according to the predicate in the sentence, and a score updater.
A summary word selection unit that selects the summary word of the document based on the weight of each word, and
An output section that outputs the selected summary word, and
Equipped with
The basic score giving unit changes the weight of at least one of the subject and the object of the sentence depending on whether the sentence is active or passive.
Information processing equipment. A summary word integration section that adds up the weights of words that are considered to be the same,
The information processing apparatus according to claim 1. The score updater updates the weights according to different rules depending on the category of the document.
The information processing apparatus according to claim 1 or 2. When the sentence is active, the basic score giving unit increases the weight of the subject of the sentence, and when the sentence is passive, the weight of the object of the sentence is increased.
The information processing apparatus according to any one of claims 1 to 3. When a specific predicate is included in a sentence, the score updater reverses the weight of the subject and the weight of the object of the sentence.
The information processing apparatus according to claim 4. The specific predicate is a verb that means the act of providing information.
The score updater weights the information provided in a sentence when the particular predicate is included in the sentence.
The information processing apparatus according to claim 5. The information processing apparatus according to claim 5 or 6, wherein the specific predicate differs depending on the category of the document. The process by which the acquisition unit acquires a document consisting of one or more sentences,
The process in which the basic score assigning unit weights each word contained in the document according to its case element, and
The process in which the score updater updates the weight of each word included in the sentence according to the predicate in the sentence, and
The process in which the abstract word selection unit selects the abstract word of the document based on the weight of each word, and
The process by which the output section outputs the selected summary word,
The process in which the basic score giving unit changes the weight of at least one of the subject and the object of the sentence depending on whether the sentence is active or passive.
Have,
Information processing method. The computer
An acquisition step to acquire a document consisting of one or more sentences,
A basic scoring step that weights each word contained in the document according to its case element, and
A score update step that updates the weight of each word contained in the sentence according to the predicate in the sentence, and
A summary word selection step that selects the summary word of the document based on the weight of each word, and
An output step that outputs the selected summary word, and
And run
A program for changing the weight of at least one of the subject and the object of the sentence depending on whether the sentence is active or passive in the basic score giving step .

Images