JP3363501B2 - Text search device - Google Patents

Description

Description: BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a text (here, a text refers to a document composed of one sentence or a plurality of related sentences). The present invention relates to a text search device that extracts a keyword of a question sentence to be input, searches for a text including the keyword, and outputs the extracted text. 2. Description of the Related Art In recent years, with the progress of computer technology and the increase in the capacity of document storage devices, databases having a large amount of data and text have become widespread, and there have been many opportunities to handle the databases. . In these databases, a large amount of data is stored in a storage device such as a hard disk in advance, and it is general to search the data based on a keyword. As a search based on a keyword in such a text search device, for example, Japanese Patent Laid-Open No. 2-2
As disclosed in Japanese Patent No. 458, a morphological analysis of a given text is performed to extract a keyword, a search of a text stored in a storage device is performed based on the extracted keyword, and the keyword matches. A method of outputting text to be used is used. An example of a search method using a conventional text search device will be described with reference to FIGS. FIG. 5 is a schematic block diagram of a conventional text search apparatus. In FIG. 1, reference numeral 1 denotes a question input unit into which a question sentence including a keyword is input; Reference numeral 3 denotes a keyword memory for storing the keywords extracted by the keyword extracting unit 2, reference numeral 4 denotes a document memory in which a text to be searched has already been stored, and reference numeral 5 denotes a keyword memory based on the keywords stored in the keyword memory 3. A text extraction unit for extracting a relevant text from the document memory 4, an output unit 6 for outputting the text extracted by the text extraction unit 5, and a control unit 7 for controlling the entire text search device. FIG. 6 shows a search result from the output unit 6 in the conventional text search apparatus shown in FIG. 5, and shows the output contents. Hereinafter, as an example of the operation of the text search apparatus shown in FIG. 5, a case will be described in which the question sentence "I want to know about Saturn's ring" is input. First, when a question sentence "I want to know about Saturn's ring" is input to the question input unit 1, the keyword extraction unit 2 performs a morphological analysis to extract a keyword. The question sentence input to the question input section 1 is "Saturn / no / ring / about / know / want /."
Represents a break by morphological analysis. Decomposed into morphemes as in ()). Here, since nouns are used as keywords, “Saturn” and “ring” are extracted as nouns. Therefore, the keyword extraction unit 2 extracts “Saturn” and “ring” as keywords and stores them in the keyword memory 3. The content of the search request in this question text is related to “Ring of Saturn”, and the text that is most requested is one in which the keywords “Saturn” and “ring” are related. When keywords are stored in the keyword memory 3, the text extracting unit 5 sequentially extracts texts including the keywords from the texts stored in the document memory 4. Table 1 shows a part of various texts stored in the document memory 4. [Table 1] The text extracting unit 5 sequentially searches for text from the contents of the document memory 4. The procedure is described below. The text extracting unit 5 determines whether or not the keyword "Saturn" is included in the text in the document memory 4 for all texts. First, the text "Saturn is said to be a symbol of Saturn" (hereinafter referred to as text 1) is extracted. As you continue your search, you will find the text "The Saturn's ring is one of the most beautiful in the solar system." The search continued, "Saturn is surrounded by satellites, and the ring is being deprived of energy from the satellites."
(Hereinafter referred to as text 4) is extracted. To continue, "How does the ring work on Saturn?"
That. ) Is extracted. Further, the text “Saturn is one of the planets of the solar system” (hereinafter referred to as text 7) is extracted. As described above, text 1, text 3,
Text 4, text 6, and text 7 are extracted. The text 2 and the text 5 are not extracted because the keyword “Saturn” is not included. Subsequently, the text extraction unit 5 determines whether or not the extracted keyword 1, text 3, text 4, text 6, and text 7 include the next keyword "ring". Text 1, text 3, text 4, and text 6 include the keyword “ring”, but text 7 does not include the keyword “ring”. For this reason, text 1, text 3, text 4, and text 6 are extracted from the text extraction unit 5 and transmitted to the output unit 6, and as shown in FIG. 4 and text 6 in this order. As described above, conventionally, as a result of a search using a keyword, texts including all types of keywords have been output from the output unit 6 in the order stored in the document memory 4. With the above configuration, it is possible to extract a text including a keyword, but since the text is output in the order stored in the document memory 4, It was not always output from a text that accurately represented the content of the search request intended by the question sentence. The present invention has been made in view of the above problems, and based on a result of parsing an extracted text,
It is an object of the present invention to provide a text search apparatus that gives priority to the order of outputting texts and outputs the texts in order from the text that accurately represents the content of the search request. In order to solve the above problems, a text search apparatus according to the present invention comprises a document memory storing a plurality of texts, and a text memory stored in the document memory. A question input unit for inputting a question sentence required for extracting a keyword, a keyword extraction unit for extracting a plurality of types of keywords from the question sentence input to the question input unit, and a keyword And then
A text extraction unit for extracting the text including the keyword from the document memory, an output unit for outputting the text extracted from the text extraction unit, and a syntax analysis for the text extracted by the text extraction unit. And a syntactic analysis judging unit for applying the key.
Judgment based on the distance between keywords extracted by the word extraction unit
The keywords in the order in which the distance between the keywords is short.
Determine the priority order for the above section to output the above text
Things. [0022] DETAILED DESCRIPTION OF THE INVENTION The present invention, keywords with the configuration described above, by using the keywords of the question sentence, the extracted text
Is determined based on the distance between the keywords, and the distance between the keywords is
By giving priorities to texts extracted in short order, it is possible to provide a text search device that automatically prioritizes and outputs texts that accurately represent search requests. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS A description will now be given, with reference to FIGS. 1 to 4, of a text search apparatus according to one embodiment of the present invention. I do. FIG. 1 is a schematic configuration diagram of a text search apparatus according to the present invention, FIGS. 2 and 3 are flowcharts showing a processing flow of a parsing determination unit 8 which is a feature of the present invention, and FIG.
It is an output result at the time of implementing the present invention. The present invention is different from the conventional example in that a syntax analysis judging unit 8 is provided between the text extracting unit 5 and the output unit 6. The parsing is performed on the extracted question text and the text extracted by the text extracting unit 5, and the text extracted by the text extracting unit 5 is rearranged based on the parsing result. It has a function to determine the priority of the text to be output. An outline of the processing in the syntax analysis determination unit 8 will be described with reference to the flowcharts shown in FIGS. At S1, the syntax of the question sentence input to the question input unit 1 is analyzed, and the syntax information of the keyword extracted by the keyword extraction unit 2, that is, which sentence and which phrase contains the keyword, Or, check the dependency relationship between keywords. In S2, one of the texts extracted by the text extraction unit 5 is sent to the syntax analysis determination unit 8
To read. In S3, morphological analysis is performed from the text read in S2 to extract keywords, and syntax information of the keywords in the read text (in which sentence or in which phrase the keywords are included, or between keywords). (Such as dependency relationships). In S4, S6, and S8, the priorities in the output unit 6 are determined based on the syntax information of the keywords extracted in S3. In the text
It is determined whether all types of keywords exist in one sentence. If all types of keywords exist in one sentence, the process proceeds to S6, and if all types of keywords do not exist in one sentence. To S5. In S5, the priority order when all types of keywords do not exist in one sentence is determined. When all types of keywords do not exist in one sentence, it is determined that there is almost no relationship between the keywords. (Hereinafter referred to as ID) is determined to be the fourth (ID = 4) (hereinafter, writing ID = n,
The priority indicates the n-th priority. ). Next, in S6, the text extraction unit 5
It is determined whether or not all types of keywords are present in one phrase in the text extracted in step. If all types of keywords are present in one phrase, the process proceeds to S8, where all types of keywords are 1 If it does not exist in one phrase, the process proceeds to S7. In S7, a priority is determined when all types of keywords do not exist in one phrase, and when all types of keywords do not exist in one phrase, it is determined that there is not much relationship between the keywords, ID
= 3. In S8, it is determined whether or not the relationship between the keywords in the text extracted by the text extraction unit 5 and the relationship between the keywords in the question sentence input to the question input unit 1 are the same. If the relationships are the same, the process proceeds to S10. If the relationships are different,
Proceed to S9. In S9, the priority order is determined when the relationship between the keywords in the text extracted by the text extraction unit 5 and the relationship between the keywords in the question sentence input to the question input unit 1 are different. In this case, it is determined that the content is close to the content of the search request but does not match,
= 2 is determined. In S10, the priority order is determined when the relationship between the keywords in the text extracted by the text extraction unit 5 and the relationship between the keywords in the question sentence input to the question input unit 1 are the same. In this case, it is determined that the content matches the content of the search request, and ID = 1 is determined. In S11, the above-mentioned S2 to S10 are executed for all the texts extracted by the text extraction unit 5.
It is determined whether or not the processing has been completed. If the processing has been completed for all the extracted texts, S1
Proceed to 2. In S12, based on the IDs determined in S2 to S10, the texts are rearranged in descending order of priority (lower IDs). In S13, it is determined whether or not the rearranged texts have the same priority. If there is one having the same priority, the process proceeds to S14, and there is no text having the same priority. In this case, the process proceeds to S17. In steps S14 to S16, if there is the same priority, the priority is determined by another method.
In S14, it is determined whether or not any of the keywords extracted in S3 has not been used in the determination of the priority in S4 to S10. , S3 are performed again on the text read in S3, and the priorities are determined. In S15 and S16, the above S2 to S1
If the priority is the same even after the processing of step 4, the priority is determined based on the distance between the keywords (the number of characters existing between a certain keyword and another keyword). In S17, based on the priorities determined in the processing of S2 to S16, the texts with the highest priority are transmitted to the output unit 6 in order. Hereinafter, as an example, a search related to the question sentence "I want to know about Saturn's ring" will be described with reference to Table 1, FIG. 2 and FIG. The flow from the question sentence input to the question input unit 1 to the text extraction by the text extraction unit 5 is the same as the flow of the text extraction of the conventional example, and the description is omitted here. A method of assigning priorities to the text extracted by the text extraction unit 5 in the syntax analysis determination unit 8 will be described with reference to FIGS. First, in S1 of FIG. 2, the question input unit 1
A morphological analysis of the question sentence "I want to know about Saturn's ring." Is performed, and "Saturn" and "ring" are extracted as keywords, and further syntax analysis is performed. The keywords "Saturn" and "Ring" are the same sentence,
It is determined that the keywords are in the same phrase and that the keywords are related. This result is stored in the syntax analysis determination unit 8 as a result of the syntax analysis of the input question sentence. Next, the text extracted by the text extraction unit 5 in S2 is sequentially read by the syntax analysis determination unit 8, and morphological analysis is performed in S3. In S3, first, a morphological analysis of the text 1 is performed, and "Saturn / about / say / ba /, / the / middle / ring / was / saturn / of / symbol / say / . ". Next, an ID is determined by syntax analysis based on the result of the morphological analysis.
It is determined whether the keywords “Saturn” and “ring” are in the same sentence for text 1. Since the keyword of the text 1 exists in the same sentence, the process proceeds to S6. In S6, it is determined whether a keyword exists in the same phrase. Since the keywords "Saturn" and "Ring" of text 1 do not exist in the same phrase, the process proceeds to S7, where ID =
After the determination is 3, the process proceeds to S11. In S11,
For all the texts extracted by the text extraction unit 5, it is checked whether or not the IDs have been determined by the syntax analysis. If the processing has been completed for all the texts extracted by the text extraction unit 5, the process proceeds to S12, and if not completed, the process returns to S2 and repeats. An example of parsing text 3, text 4, and text 6 will be described below. Subsequently, the morphological analysis of the text 3 is performed in S3. Text 3 is "Saturn / no / ring / wa /,
/ In the solar system / in / most / beautiful / things / one / one /. The analyzed text 3 is also subjected to S4, S6, S8, and S10 in the same manner as the text 1. Since the syntax analysis result of the text 3 is the same as the syntax analysis result of the question sentence, , ID = 1 is repeated, and the same is repeated for the third text extracted by the text extraction unit 5. Text 4 is described as "Saturn / ha /, / satellite // surrounding / surrounding. /ing/. / And / ring / has /, / satellite / from / energy / deprived / deprived /. The morphological analysis is performed as described above, and the ID is determined through the processing of S4 and S5, and the text 4 has ID = 4. / Work / does / does / no / maybe /? And the morphological analysis is performed as described above, and ID = 2 is determined through the processing of S4, S6, S8, and S9. As described above, the IDs are determined by parsing all the texts extracted by the text extracting unit 5. When the ID is finally determined by the syntax analysis, the output order is determined from those having the highest priority by the syntax analysis in S12 of Fig. 3. In S12, those having the same priority exist. If so, the process proceeds to S13, and if not, in S17, the output order is specified to the output unit 6. In the present embodiment, text 1 has ID = 3 and text 3 has ID = 3. 1, text 4 has ID = 4, and text 6 has ID = 2. Therefore, in step S17, the output unit 6 outputs text 3, text 6, text 1, and text 4 in this order. Is forced.
FIG. 4 shows the results. In this embodiment, the case where there is no one having the same priority is described. However, when there are a plurality of ones having the same priority, as shown in S14 of FIG. It is also possible to make a determination based on keywords that exist other than the keywords that have been used. Also, S1
If the same is obtained even after performing step 4, it is possible to determine the priority order based on the distance between keywords (such as the number of characters between keywords) in the extracted text, as shown in S15 and S16. In this embodiment, the case where the keyword extracted by the keyword extraction unit 2 is two has been described as an example. However, the case where there are three or more keywords is basically the same. In the case where there are three, it is also possible to use the result of parsing with three keywords, or the result of parsing for a keyword with a high frequency of appearance in a question sentence. According to the present invention, as is apparent from the above description, the text extracted by the keyword is output from the output unit.
Since the priority of the output order can be determined,
It is possible to sequentially output a search request from a text that accurately represents the search request.

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a schematic configuration diagram of a text search device including a syntax analysis determination unit according to the present invention. FIG. 2 is a flowchart illustrating a processing flow of the syntax analysis determination unit according to the present invention. FIG. 4 is a flowchart showing a flow of processing of a parsing determination unit according to the present invention. FIG. 4 is a diagram showing an output result when the present invention is implemented. FIG. 5 is a schematic configuration diagram of a conventional text search device. Diagram showing output result according to conventional example [Description of reference numerals] 1 Question input unit 2 Keyword extraction unit 3 Keyword memory 4 Document memory 5 Text extraction unit 6 Output unit 7 Control unit 8 Syntax analysis determination unit

Continuation of the front page (58) Field surveyed (Int.Cl. ⁷ , DB name) G06F 17/30 350 G06F 17/30 170 G06F 17/30 330 G06F 17/27 JICST file (JOIS)

Claims (1)

(57) [Claims] (1) A document memory storing a plurality of texts, and a question input unit for inputting a question sentence required for extracting the text stored in the document memory. A keyword extracting unit for extracting a plurality of types of keywords from the question sentence input to the question input unit, and extracting a text including the keywords from the document memory based on the keywords extracted by the keyword extracting unit. text extraction unit that includes an output unit for outputting the text extracted from the text extracting unit, on the extracted text in the text extraction unit, a syntactic analysis unit for performing parsing, the, above constituting sentence The analysis determination unit is extracted by the keyword extraction unit.
The keyword is determined based on the distance between the keywords.
The text is output from the output unit in the order of the shortest distance.
A text characterized by determining the priority order to be output
Strike search device.