JP3612769B2 - Information search apparatus and information search method - Google Patents

Description

[0001]
[Industrial application fields]
The present invention relates to an information search apparatus and an information search method for searching for information such as a document that matches a search request.
[0002]
[Prior art]
Conventionally, as a method for searching for desired information from a large amount of information stored in a database or the like, a keyword is assigned to each data in advance, and information having a keyword that matches a search key input by a user is searched. Have been developed.
[0003]
This method is called an index method or an index method. Examples of the indexing method include a hash method and a digital search method. For details, refer to “The Artof Computer Programming, Volume 3 / Sorting and Searching” (Donald E. Knud, Addison-Wesley Publishing Company, 73). Searching and the like.
[0004]
On the other hand, in recent years, a technique has been developed in which a keyword is not assigned in advance, but information including a search key input by a user can be directly searched. This method is called full text search or full text search. There are BM method, AC method, FSA method and so on as a method for realizing this method. For details, see Chapter 2, Chapter 3, etc. of "Text Search Processor" (written by Takahashi Kosuke, The Institute of Electronics, Information and Communication Engineers, 1991). Has been described.
[0005]
However, since there are few cases where information with a certain search key is uniquely specified, the information searched by one search key is usually plural, and when the amount of stored information becomes large, the number of candidates is Become very much. Therefore, it is a common practice to reduce the number of candidates by designating a plurality of search keys and outputting information including all of the search keys. However, many of the search results obtained in this way are different from the information that the user wanted, and the user has to spend time selecting the desired information. It was.
[0006]
Conventionally, for example, when it is desired to search a newspaper article about the development of a workstation, a search key is designated as in “workstation & development”. However, some of the results from this search included articles about developing workstations, articles about developing certain software on workstations, and even "workstations" and " Articles that contain keywords such as “development” but have no semantic relationship between these words are also included. Users had to read all these many articles and find the one they wanted.
[0007]
In order to eliminate such useless search results, for example, Japanese Patent Application Laid-Open No. 1-112331 describes a technique for efficiently performing keyword search by adding importance to keywords. However, if the importance of the keywords is the same, for example, if you want to search for an article about a user developing a workstation as described above, an article about developing a software on a workstation is Even though the meaning is different, it is still presented to the user as the same search result.
[0008]
Further, in such a search, there is a possibility that omission occurs only with information that completely matches the search key. For this reason, in recent years, a search key is expanded by referring to a synonym dictionary called a thesaurus and a search is performed based on the search key. As such a search device, for example, there is a document search device described in JP-A-62-248032. Thus, the use of synonyms also increases the number of search result candidates, and the task of selecting desired information is becoming more difficult.
[0009]
For example, in Japanese Patent Application Laid-Open Nos. 62-191924 and 62-191925, when searching using a plurality of search keys, an OR search is performed to prevent omission, and search results A search device is described in which a degree of matching between a keyword and a search key is calculated and output in descending order of the degree of matching so that a search result that is as close to the user's intention as possible is displayed preferentially. ing.
[0010]
However, even if such a synonym dictionary is used or a technique such as performing an OR search to calculate the degree of coincidence is used, for example, a newspaper article that has developed a workstation and software on the workstation. It will be mixed with the newspaper article that developed. Therefore, the user has to sequentially refer to unnecessary search result information, and the work of finding desired information is complicated.
[0011]
[Problems to be solved by the invention]
The present invention has been made in view of the above-described circumstances, and an object thereof is to provide an information search apparatus and an information search method that make it easy to find desired information from search results.
[0012]
[Means for Solving the Problems]
According to the present invention, in the information search device for searching for information that matches a search request, the information storage means for storing information including at least text, and the information using a plurality of words as search keys. An information search means for searching for information from the storage means, a relation extraction means for extracting a relationship between a plurality of words used as search keys from the information searched by the information search means, and a search result by the information search means; Display means for displaying a pair of relations extracted by the relation extraction means; and when the plurality of words are present in one sentence, the relation extraction means performs a morphological analysis on the one sentence to And when the plurality of words do not exist in one sentence, the distance between the sentence including each word and the sentence is calculated.
[0013]
Similarly, in the invention according to claim 9, in the information search method for searching for information that matches the search request, information including at least text is stored in the information storage means, and information search is performed using a plurality of words as search keys. The information is retrieved from the information storage means by means, the relation between the plurality of words used as the retrieval key is extracted from the retrieved information by the relation extraction means, and the search result and the extracted relation are displayed in pairs When the plurality of words are present in one sentence and the plurality of words are present in one sentence, the relation between the words is extracted by analyzing the one sentence, and when the plurality of words are not present in one sentence The distance between sentences containing each word is calculated.
[0014]
According to a second aspect of the present invention, in an information search device for searching for information that matches a search request, information storage means for storing information including at least text, and information using a plurality of words as search keys from the information storage means Information search means for searching for information, search result storage means for holding information searched from the information search means, and a relationship between a plurality of words used as search keys from information stored in the search result storage means A relation extraction means for extracting and storing in the search result storage means, and a display means for displaying the searched information held in the search result storage means and the extracted relation as a pair, When the plurality of words are present in one sentence, the relationship extracting means extracts a relationship between the words by performing morphological analysis on the one sentence, and when the plurality of words are not present in one sentence, It is characterized in calculating the distance between the sentence and the sentence containing the word respectively.
[0015]
Similarly, in the invention according to claim 10, in the information search method for searching for information that matches the search request, information including at least text is stored in the information storage means, and a plurality of words are stored in the information storage means. Information is searched by the information search means as a search key, the searched information is held in the search result storage means, and the relationship between a plurality of words used as the search key from the information stored in the search result storage means is determined. Extracted by the relationship extraction means and stored in the search result storage means, and the searched information held in the search result storage means and the extracted relationship are displayed on the display means as a pair, When the plurality of words are present in one sentence, the sentence is morphologically analyzed to extract a relationship between the words. When the plurality of words are not present in one sentence, each word is It is characterized in calculating the distance between the free sentences and sentence.
[0016]
According to a third aspect of the present invention, the information search device according to the first or second aspect further comprises synonym storage means for holding a list of words similar to a certain word, and the information search means includes Information is retrieved from the information storage means using a synonym obtained from the synonym storage means for a word as a search key.
[0017]
According to a fourth aspect of the present invention, in the information search device according to the first or second aspect, a translation in which a word in a certain language is input and a word equivalent to or similar to the word in each language is output together with the type of the language. A dictionary, the information search means gives the word given as a search key to the bilingual dictionary, searches information from the information storage means using the word obtained from the bilingual dictionary as a search key, and extracts the relation The means is characterized by extracting a plurality of words used as a search key for each language and a relationship between them according to the type of language in which the text in each searched information is described. is there.
[0018]
According to a fifth aspect of the invention, in the information search device according to the fourth aspect of the present invention, there is further provided relation relation storage means for holding relation correspondence information describing the relation between words in a certain language and the relation between words in another language. And the relationship extraction means includes, from a search request described using a certain language, a plurality of words given as search keys to the information search means according to the type of language in which the search request is described, and A plurality of words used as search keys in the text for each language according to the type of language in which the text in each information searched by the information search means is described, while extracting the relationship between words The relationship between words is extracted, and the relationship between words extracted from the search request and the relationship extracted from the information retrieved by the information retrieval unit are stored in the relationship correspondence storage unit Compared using the relationship correspondence information that has been, it is characterized in that to extract only the matching or consistent information.
[0019]
According to a sixth aspect of the present invention, in the information search device according to the fourth or eleventh aspect, the relationship extracting means is configured to determine the relationship based on a plurality of words extracted from each searched information and the relationship between them. The retrieved information is rearranged and output based on the clarity, the similarity between the plurality of words and the plurality of words given as the search key, or both.
[0020]
According to a seventh aspect of the present invention, in the information search device according to the fifth aspect, the relationship extraction means includes a relationship extracted from the search request and a relationship extracted from the information searched by the information search means. And the degree of coincidence is determined, and the retrieved information is rearranged based on the degree of coincidence.
[0021]
According to an eighth aspect of the present invention, in the information retrieval device according to any one of the fourth to thirteenth aspects, the language of the information in which the text of the information is described among the information retrieved by the information retrieval means. It is characterized by having a language comparison means that retrieves only information whose type matches the type of the designated language.
[0026]
[Action]
According to the present invention, in the inventions according to claim 1 and claim 9, the relationship extraction means uses a plurality of search keys searched from a plurality of search keys as a relationship between words used as search keys. When a word exists in one sentence, the relation between the words is extracted by analyzing the one sentence, and when a plurality of words do not exist in one sentence, a distance from the sentence including each word is calculated. . The relationship between the extracted words is displayed together with the searched information on the display means, thereby making it easy to find information desired by the user.
[0027]
Further, in the inventions according to claim 2 and claim 10, the relationship extracting means, based on the search results searched by the plurality of search keys, includes a plurality of words in one sentence as a relationship between the words used as the search keys. If there is a plurality of words in one sentence, the distance from the sentence containing each word is calculated and the search result Store in storage means. By displaying the search result stored in the search result storage means and the relationship between the words used as the search key in pairs on the display means, it is possible to easily find the information desired by the user. .
[0028]
In the invention according to claim 3, synonym storage means is further provided, and a synonym is obtained from the synonym storage means for a certain word, and this is used as a search key to perform a search on the information storage means. It is possible to prevent missing search results that occur only by matching.
[0029]
According to the fourth aspect of the present invention, it is possible to search for information including multiple languages by searching information from the information storage means using a word obtained from the bilingual dictionary as a search key. At this time, since the relationship extraction unit extracts the relationship between words for each language, even information including text described in different languages can easily find information having a desired content.
[0030]
According to the fifth aspect of the present invention, a plurality of words and a relationship between the words are extracted from the search request according to the type of language in which the search request is expressed. A language in which information is retrieved from the information storage means using a word obtained from the bilingual dictionary based on a plurality of words extracted from the retrieval request as a retrieval key, and the text in each retrieved information is described Depending on the type, the relationship between a plurality of words used as search keys in the text is extracted for each language. The relationship extracted from the retrieved information is compared with the relationship extracted from the retrieval request, and only information that does not match or contradict is extracted. At this time, by using the relationship correspondence information held in the relationship correspondence storage means, the relationship is compared in consideration of the correspondence between words in the same language and the correspondence between words in different languages. Thereby, even if it is the information containing the text described in the language different from a search request, the information which has a desired content can be obtained.
[0031]
According to the invention described in claim 6, each searched information is based on the relationship between the words extracted from the information, the relationship is clarified, or the hit word and the information search device have a search key. Based on the similarity to the word given as or both, the retrieved information is sorted and output. Thus, it is possible to refer to information that is semantically clear or from information that is semantically close to the word given as the search key, and the user can easily obtain a desired search result.
[0032]
According to the seventh aspect of the present invention, the degree of coincidence is determined when comparing the relationship extracted from the retrieved information with the relationship extracted from the retrieval request, and the retrieved pieces of information are arranged based on the degree of coincidence. Instead, for example, by outputting to a display device or the like, it is possible to present to the user from information that is considered as close to the search request as possible. Thus, the user can obtain a desired search result in a short time.
[0033]
In the multilingual search according to claims 4 to 7, the language type is designated as in the invention according to claim 8, and only information that matches the designated language type is taken as the search result. It can be configured to do so.
[0037]
【Example】
FIG. 1 is a block diagram showing a first embodiment of the information search apparatus of the present invention. In the figure, 1 is input means, 2 is information search means, 3 is information storage means, 4 is search result storage means, 5 is relationship extraction means, and 6 is display means.
[0038]
The input unit 1 is used to input a search key related to information to be searched, and can be composed of an input unit such as a keyboard. A plurality of words can be input as a search key.
[0039]
The information retrieval unit 2 retrieves information from the information storage unit 3 using the retrieval key input by the input unit 1 and stores the retrieval result in the retrieval result storage unit 4. As a search method, a keyword is assigned in advance corresponding to each information stored in the information storage means 3, and an index search method for searching for information having a keyword that matches the search key, or a keyword is assigned in advance. Any method may be used such as a full text search method in which information including a search key input by a user is directly searched by a method such as matching character strings. As described above, these can be easily realized by using a conventionally used search technique.
[0040]
The information storage means 3 is a storage device that stores a large amount of information. The information storage means 3 may have any configuration as long as it can store information, such as a magnetic disk. Further, the stored information may include a picture, a figure, or the like as long as it includes text written in at least a natural language (in this example, Japanese).
[0041]
The search result storage unit 4 is a storage device that stores information on the results retrieved from the information storage unit 3. The search result storage means may have any configuration as long as it can store information such as a magnetic disk. The search result storage unit 4 stores information on the search results, and stores the relationship information between words extracted from the relationship extraction unit 5 in association with each stored information. it can.
[0042]
The relationship extraction means 5 extracts the relationship between a plurality of words from the sentence. A morphological analysis is performed on the sentence to extract a word string, and the word string is compared with a relation extraction rule to extract a relation between a plurality of related words.
[0043]
The display means 6 displays both information and relationships held in the search result storage means 4, and can be constituted by a display device such as a CRT, for example.
[0044]
In the above-described configuration, the relationship extraction unit 5 performs the relationship extraction process after the search result is stored in the search result storage unit 4. However, the present invention is not limited to this, and receives the search result directly from the information search unit 2. The extraction process may be performed. At this time, information other than the relationship between words displayed by the display unit 6 may be received via the relationship extraction unit 5 or directly from the information search unit 2.
[0045]
It is a flowchart which shows an example of operation | movement in the 1st Example of the information search device of this invention. In this information search apparatus, when a search key is first input, the information search means 2 searches for information from the information storage means 3 and stores the search results in the search result storage means 4 in S11. Next, in S <b> 12, the relationship extraction unit 5 extracts a relationship between a plurality of words used as a search key from the search result stored in the search result storage unit 4 and stores it in the search result storage unit 4. . Finally, in S13, both the information stored in the search result storage unit 4 and the relationship are displayed on the display unit.
[0046]
FIG. 3 is a flowchart showing an example of the operation of the relationship extraction means in the first embodiment of the information search apparatus of the present invention. When the information retrieved by the information retrieval unit 2 is stored in the retrieval result storage unit 4, the relationship extraction unit 5 is activated. First, in S21, one of the search results stored in the search result storage means 4 is taken out. Here, the case where the search key used for the search is two words is shown. The search key used for the search is stored in variables W1 and W2, and one retrieved result is stored in variable TEXT.
[0047]
In S22, it is determined whether or not the search key word stored in the variables W1 and W2 exists in one sentence. If it exists in one sentence, in S23, the search result stored in the variable TEXT or one sentence including the search key in the search result is subjected to morphological analysis. Morphological analysis is widely known as a basic technology of natural language processing. For example, “Basic technology of natural language processing” (Hirogo Nomura, The Institute of Electronics, Information and Communication Engineers, 1988) and “Information Processing” (Vol. 30, No. 10, 1989) and can be easily realized by the method described in “3.1 Morphology”.
[0048]
In S24, the partial word string including the words stored in the variables W1 and W2 is collated with the relationship extraction rule. In the relationship extraction rule, at least information related to the arrangement of words in the word group and the relationship of the word group are paired and registered. In S25, it is determined whether or not there is a word group that matches the word group. If there is a match, the corresponding relation in the relationship extraction rule is stored in the search result storage means 4. In FIG. 3, the search result storage unit 4 is described as “REL”. If there is no match, information “0 sentence” is stored in the search result storage means 4 to indicate that there are at least two search keys in the same sentence.
[0049]
If the word stored in the variables W1 and W2 does not exist in one sentence in S22, the sentence including the word stored in the variable W1 and the sentence including the word stored in the variable W2 in S29. Is stored in the search result storage means 4. Here, the sentence distance is, for example, a sentence including the word stored in the variable W1, and when the word stored in the variable W2 is included in the next sentence, the distance is “1 sentence”. And In addition, when one sentence is sandwiched between a sentence including the word stored in the variable W1 and a sentence including the word stored in the variable W2, the distance is set to “two sentences”. Similarly, the distance is calculated.
[0050]
In S26, S28, and S29, when the relationship or distance between the variables W1 and W2 is stored in the search result storage unit 4, whether or not an unprocessed search result of the relationship extraction process exists in the search result storage unit 4 in S27. If an unprocessed search result remains, the process returns to S21 to process the next search result. When the relationship extraction process for all search results is completed, the operation of the relationship extraction means is terminated. At this point, information on the relationship or distance in which the search key appears in the search result is added to each search result stored in the search result storage unit 4. Such information is also displayed on the display means 6. Thereby, the user can know whether or not the search result is a desired one by referring to the displayed relationship or distance. Therefore, desired information or information close to the desired information can be obtained in a short time.
[0051]
Next, based on a specific example, the above operation will be described in detail with reference to the flowcharts shown in FIGS. 2 and 3 with reference to FIGS. Here, as a specific example, an example will be described in which “workstation & development” is input from the input means 1 in order to search for a newspaper article about the development of a workstation.
[0052]
Information storage means 3 stores a large amount of newspaper article information. Each article is composed of Japanese text and figures and pictures. The information search means 2 has an independent word included in a Japanese sentence part of each article as an index.
[0053]
In S11, the information search unit 2 outputs an article having a keyword that matches the search key input from the input unit 1 as a result. Here, the information search means 2 searches the information storage means 3 by the logical product of the two words “workstation” and “development”.
[0054]
FIG. 4 is an explanatory diagram showing an example of a search result by two keys in the first embodiment of the information search apparatus of the present invention. The information search means 2 obtains a search result as shown in FIG. 4 and stores it in the search result storage means 4. For example, as shown in FIG. 4, the search result storage means 4 includes an item called key 1 and key 2 for storing the search key used for the search, an item called search result for storing information of the searched result, It can be composed of an item for storing a relationship between words obtained later by the relationship extracting means. In FIG. 4, the information in the search result item is indicated by underlining the word portions of the keys 1 and 2. As can be seen from FIG. 4, the search result is information gathered merely by having the word “workstation” and the word “development”. In addition to information, various types of information are mixed.
[0055]
Next, in S12, the relationship extraction unit 5 extracts a relationship between a plurality of words used as a search key from the search result stored in the search result storage unit 4, and stores it in the search result storage unit. Relation extraction is performed by morphologically analyzing a sentence including a search result sentence or a search key, dividing the sentence into word strings, and collating with a relation extraction rule including an expression to be extracted as a relation.
[0056]
First, in S21 of FIG. 3, a search result and one search key are extracted from the search result storage unit 4. Here, first, from the search result in the search result storage means 4 shown in FIG. 4, the search result whose title is “release proprietary WS” is extracted and processed. Search keys 1 and 2 are “workstation” and “development”, respectively. The text of the search result is “Company A developed a workstation with a RISC type central processing element that has a high processing speed for 5 days. , The company announced that it will release a total of 7 models and 11 models. Search key 1 and search key 2 are stored in variables W1 and W2, and the text of the search result is stored in variable TEXT.
[0057]
In this search result, search key 1 and search key 2 "workstation" and "development" are in the same sentence, so the process proceeds from S22 to S23, the morphological analysis is performed on the sentence in variable TEXT, and the word string Divide into FIG. 5 is an explanatory diagram showing an example of a morphological analysis result in the first embodiment of the information search apparatus of the present invention. The result of morphological analysis is obtained by pairing the divided words and their parts of speech.
[0058]
Next, in S24, the word string divided by the morphological analysis in S23 is collated with a relationship extraction rule that defines an expression to be extracted as a relationship. FIG. 6 is an explanatory diagram showing an example of the relationship extraction rule in the first embodiment of the information search apparatus of the present invention. In the relationship extraction rule, as shown in FIG. 6, a rule for arranging word strings, a word extracted at that time, and a relationship are registered as a pair. Here, sa-variant nouns are shown as sa-variants, and adjective verb stems are shown as adjective verbs. For example, if a word continues like noun + “ga” + sa-variant noun, then the noun being the word 1 and the sa-noun being the word 3 are extracted as a relationship.
[0059]
From the result of the morphological analysis shown in FIG. 5, the word string “workstation” + “to” + “development” in the “workstation” and “development” portions and the relation extraction rule shown in FIG. , It conforms to the rule of “noun / sa modification +“ to ”+ sa modification”. For this reason, the process proceeds from S25 to S26, where the relation “wo” is extracted and stored in the search result storage means 4.
[0060]
In S27, the search result storage means still has a search result for which no relation has been extracted. Therefore, the process returns to S21, and the relation extraction means 5 performs processing for the next search result. This time, the processing is continued for the search result whose title is “A partnership between company C and company D is started”. Search keys 1 and 2 are “workstation” and “development”, respectively. The text of the search results is “Company C ... sells a new workstation supplied by Company D through an overseas sales network. In addition, company D ... also announced the sale of the developed large computer in the UK ... ". Search key 1 and key 2 are stored in variables W1 and W2, and the text of the search result is stored in variable TEXT.
[0061]
Since the search keys “workstation” and “development” are not in the same sentence, the process proceeds from S22 to S29. Since the search key “development” is included in the sentence next to the sentence including the search key “workstation”, the distance between the sentences including both words is one sentence. Therefore, the relationship “1 sentence” is extracted, and this relationship “1 sentence” is stored in the search result storage means 4.
[0062]
Thereafter, the same processing is performed for all search results. FIG. 7 is an explanatory diagram showing an example of a search result after the relation extraction process in the first embodiment of the information search apparatus of the present invention. As described above, the information stored in the search result storage unit 4 when the relationship extraction process by the relationship extraction unit 5 is completed is as shown in FIG. A relationship is assigned to each search result.
[0063]
Thus, the process of extracting the relationship from the search result is completed. In S13, this relationship is displayed on the display unit 6 together with the search result information. FIG. 8 is an explanatory diagram showing an example of a search result display after the relationship extraction process in the first embodiment of the information search apparatus of the present invention. In the display example shown in FIG. 8, three areas of an input area, a candidate display area, and a result display area are displayed. The input search key is displayed in the input area. In the candidate display area, the search result and the relationship extracted by the relationship extracting means 5 are paired and displayed in the searched order. In the case of the specific example described above, the user wants to search for a newspaper article about the development of a workstation, so in FIG. Good, no need to see other information. Therefore, the user can obtain desired information immediately.
[0064]
By selecting one of the pair of the search result and the relationship, the content of the search result is displayed in the result display area on the right side. In FIG. 8, the selected search results are shown hatched. In this case, the text “Develop an originally developed WS” is selected as the title, and the retrieved text is displayed in the result display area. At this time, as shown in FIG. 8, the search key may be underlined, highlighted, or highlighted so as to change color or brightness.
[0065]
FIG. 9 is an explanatory diagram of another example of display of search results after the relationship extraction process in the first embodiment of the information search apparatus of the present invention. FIG. 9 shows an example in which the same relations corresponding to the search results are displayed together in the candidate display area. By displaying in this way, the user can easily find the result he / she wants to search for. In the case of the specific example described above, the user wants to search for a newspaper article about the development of the workstation, so in FIG. 9, it is only necessary to look at the information where the relationship is lined up in the “wo case” There is no need to look at other information. Compared to the display example shown in FIG. 8, the time and effort for searching for a related portion is reduced, so that more desired information can be easily obtained.
[0066]
Here, an example is shown in which only the relationship is displayed together with the search result, but the display method is not limited to this. It may also be displayed together with it, or it may be displayed in ordinary Japanese, such as “develop workstation”, instead of a symbol representing a relationship such as “wo case”. Furthermore, for example, “a. Wo case b. De case c. 1 sentence d. 2 sentence e. 3 sentence” is displayed first, and only one of the relations is displayed and one of them is selected. Search results having the selected relationship may be displayed.
[0067]
FIG. 10 is a block diagram showing a second embodiment of the information search apparatus of the present invention. In the figure, the same parts as those in FIG. Reference numeral 7 denotes synonym storage means. The second invention further includes synonym storage means 7 which holds a list of words similar to a certain word in the first invention. The information retrieval unit 2 retrieves information from the information storage unit 3 using a synonym obtained from the synonym storage unit 7 as a search key for a certain word.
[0068]
The synonym storage means 7 holds a list of words that are similar to a certain word. For example, for the word “workstation”, a similar word such as “computer” or “office computer” is held. In contrast to the word “development”, it holds a word such as “creation”. When such a synonym is held in the synonym storage means 7, the information search means 2, for example, performs a search under the search condition “workstation & development”, “(workstation or personal computer or office computer)”. & (Development or creation) "will be searched.
[0069]
Thereafter, the relationship extraction means 5 extracts the relationship between the plurality of words used as the search key from the search results stored in the search result storage means as in the first invention, and stores them in the search result storage means. . The relationship extraction unit 5 may be configured to receive the search result directly from the information search unit 2 and output the extracted relationship information to the display unit 6.
[0070]
FIG. 11 is an explanatory diagram of an example of a result of a search using a synonym and a relationship extraction process in the second embodiment of the information search apparatus of the present invention. As shown in FIG. 11, the search key 1 has not only “workstation” but also “computer” and “office computer”, and the search key 2 has not only “development” but also “creation”. Information is also being searched. Thereby, when referring to an article developed for a workstation, for example, an article developed for a personal computer can also be searched, and a search omission can be prevented.
[0071]
Thus, the process of extracting the relationship from the search result is completed, and this relationship and the result information are displayed as in the first invention. FIG. 12 is an explanatory diagram showing an example of the display of the search result after the relation extraction process in the second embodiment of the information search apparatus of the present invention. In FIG. 12, like the display example shown in FIG. 9, those having the same relationship are collectively displayed. By displaying in this way, the user can easily find the result he / she wants to search for.
[0072]
FIG. 13 is an explanatory diagram of another example of the display of the search result after the relationship extraction process in the second embodiment of the information search apparatus of the present invention. In the display example shown in FIG. 13, the search key is displayed together with the relationship. By displaying in this way, the user can determine what kind of search key is used to search for information, easily find the result he wants to search for, and what words exist as synonyms, In addition, it is possible to know that a search has been performed using synonyms.
[0073]
As described above, when the user wants to search for a newspaper article about the development of a workstation, in FIG. You can see the result information that is “workstation” and “development” respectively, and if necessary, also refer to the result information that key 1 and key 2 are “PC” and “development” respectively. It ’s fine.
[0074]
In the above-described example, the synonym storage unit 7 has been described with respect to an example in which synonyms such as “personal computer” and “office computer” are held for the word “workstation”. In addition to this, for example, a high-level concept such as “computer” or a low-level concept such as a product name for “workstation” is stored in the synonym storage means 7 for the word “workstation”. The search may be performed using these superordinate concepts or subordinate concepts. For example, when searching using the superordinate concept, for example, when searching with the search condition “workstation & development”
(Workstation or computer) & (development)
Search with
(Workstation or computer or off-computer or computer) & (development or creation)
You can search with. The processing when a search is performed under these search conditions is the same as in the above example.
[0075]
Usually, when a search is performed including synonyms, a large amount of information is obtained as a search result, and it is not easy to select desired information from the information. However, as in the present invention, desired information can be easily obtained by picking up and referencing only information having a keyword intended relationship with the user.
[0076]
FIG. 14 is a block diagram showing a third embodiment of the information search apparatus of the present invention. In the figure, the same parts as those in FIG. 8 is an input information storage means, and 9 is a relation comparison means. In the first and second embodiments described above, when a search request is input from the input means 1, a plurality of words are input. In the third embodiment, a case where a search request is input in a natural language is shown.
[0077]
The input means 1 inputs a search request related to information to be searched in a natural language. For example, it can be configured by an input device such as a keyboard. In the present embodiment, the search request is shown for those entered in Japanese. In this case, when Kana-Kanji conversion or the like is performed as Japanese input, the input unit 1 may be configured to include software for Kana-Kanji conversion.
[0078]
The input information storage unit 8 stores a plurality of words extracted by the relationship extraction unit 5 from the search request and the relationship between the words together with the search request input from the input unit 1. The input information storage means 8 may have any configuration as long as it can store information such as a magnetic disk.
[0079]
The relationship extraction means 5 extracts a plurality of words and the relationship between the words from the sentence. A morphological analysis is performed on the sentence to extract a word string, and the word string is compared with a relation extraction rule to extract a relation between a plurality of related words. The extraction of the relationship is performed on the search result information stored in the search result storage means, and in the third embodiment, it is also performed on the search request stored in the input information storage means 8.
[0080]
The relationship comparison unit 9 compares the relationship stored in the input information storage unit 8 with the relationship stored in the search result storage unit 4 and sends matching information to the display unit 6.
[0081]
The display means 6 displays the search result information that matches the search request by the comparison processing by the relation comparison means 9. For example, a display device such as a CRT or LCD can be used.
[0082]
In the configuration of the third embodiment, the relationship extraction unit 5 can be configured to obtain the search result directly from the information search unit 2 without using the search result storage unit 4. Furthermore, the relationship comparison unit 9 can also be configured to obtain the relationship extracted directly from the relationship extraction unit 5. In this case, the searched information may be configured to be supplied to the display unit 6 via the relationship extraction unit 5 or the relationship comparison unit 9 or directly from the information search unit 2 or the information storage unit 3.
[0083]
FIG. 15 is a flowchart showing an example of operation in the third embodiment of the information search apparatus of the present invention. In the third embodiment, when a search key is first input, the input search request is stored in the input information storage means 8 in S31. In S 32, the relationship extraction unit 5 extracts a plurality of words and the relationship between the words from the search request and stores them in the input information storage unit 8. Next, in S33, the information search means 2 searches the information storage means 3 using a plurality of words stored in the input information storage means 8 as search keys, and stores the search results in the search result storage means 4. To do.
[0084]
Thereafter, in S 34, the relationship extraction unit 5 extracts a relationship between a plurality of words used as a search key from the search result stored in the search result storage unit 4 and stores it in the search result storage unit 4. Finally, in S35, the relationship comparison unit 9 compares the relationship stored in the input information storage unit 8 with the relationship stored in the search result storage unit 4, and displays matching information on the display unit 6. indicate.
[0085]
FIG. 16 is a flowchart showing an example of the operation of the relationship extraction means 5 in the third embodiment of the information search apparatus of the present invention. As described above, in the third embodiment, the relationship extraction unit 5 performs the relationship extraction process for the search request stored in the input information storage unit 8 and the search result stored in the search result storage unit 4. Perform relationship extraction processing. First, in the case of a search request, the process proceeds from S41 to S42, and the search request stored in the input information storage unit 8 is read into the variable TEXT. The read search request is decomposed into words by morphological analysis in S43. In S44, the word string decomposed by the morphological analysis is compared with the relationship extraction rule, and it is determined in S45 whether there is a match in the relationship extraction rule. If there is a matching relationship extraction rule, a word group and a relationship are extracted based on the matched relationship extraction rule and stored in the input information storage unit 8 in S46. If there is no matching relationship extraction rule, an error occurs in S47.
[0086]
In the case of the extraction process for the search result, the process proceeds from S41 to S48. Since the processing from S48 to S56 is the same as the processing from S21 to S29 in FIG. 3 of the first embodiment, the description thereof is omitted here.
[0087]
Next, based on a specific example, the above operation will be described in detail with reference to the flowcharts shown in FIGS. 15 and 16 with reference to FIGS. Here, as a specific example, an example in which “develop workstation” is input from the input means 1 to search for a newspaper article about the development of a workstation will be described, as in the above-described embodiment.
[0088]
In the third embodiment, when “develop workstation” is input as a search request from the input means 1, the input search request is stored in the input information storage means 8 in S31. FIG. 17 is an explanatory diagram showing an example of a search request stored in the input information storage means in the third embodiment of the information search apparatus of the present invention. At this stage, as shown in FIG. 17, the input search request is stored in the input information storage unit 8 as it is.
[0089]
Next, in S 32, the relationship extraction unit 5 extracts a plurality of words that are related to each other from the search request stored in the input information storage unit 8 and the relationship between the words, and the input information storage unit 8. To store.
[0090]
In the relationship extraction processing in the relationship extraction means 5, first, since it is the relationship extraction processing for the search request, the process proceeds from S41 to S42 in FIG. In S42, the search request “Develop workstation” is extracted, and the extracted search request is morphologically analyzed in S43 and divided into word strings. FIG. 18 is an explanatory diagram of the morphological analysis result of the search request in the third embodiment of the information search apparatus of the present invention. The input search request “develop workstation” is subjected to morphological analysis as shown in FIG. 18 and divided into words.
[0091]
Next, in S44, the word string obtained by the morphological analysis is collated with a relation extraction rule that defines an expression to be extracted as a relation. As the relation extraction rule, the same one as used when extracting the relation from the search result can be used. For example, the relationship extraction rule shown in FIG. 6 of the first embodiment can be used.
[0092]
When the word string shown in FIG. 18 and the relation extraction rule shown in FIG. 6 are collated, “workstation” is a noun and “development” is a sub-variable noun. The process proceeds from S45 to S46, and the two words “workstation” and “development” and the relationship “wo case” are extracted. A plurality of extracted words and the relationship between the words are stored in the input information storage unit 8. FIG. 19 is an explanatory diagram showing an example of a search request after the relationship extraction process in the third embodiment of the information search apparatus of the present invention. Along with the input search request, the words and relationships extracted by the relationship extracting means 5 are stored. Similarly, when a plurality of words and relationships are extracted, they are also stored in the input information storage unit 8. This completes the process of extracting the relationship from the search request.
[0093]
Next, in S33 of FIG. 15, the information search means 2 uses the two words “workstation” and “development” extracted from the search request by the relationship extraction means 5 as search keys, and uses the logical product as a search condition. Information is retrieved from the information storage means 3. That is, the same search is performed as when “workstation & development” is input in the specific example of the first embodiment. As a search result, for example, the search result shown in FIG. 4 is obtained as in the first embodiment. The obtained search result is stored in the search result storage means 4.
[0094]
Next, in S <b> 34, the relationship extraction unit 5 extracts a relationship from the search result stored in the search result storage unit 4 and stores it in the search result storage unit 4. The process proceeds from S41 to S42 in FIG. 16, and the relationship is extracted for each search result stored in the search result storage unit 4 and stored in the search result storage unit 4 as in the first embodiment. Regarding the search results shown in FIG. 4, the contents of the search result storage means 4 after the relation extraction means 5 extracts the relation are as shown in FIG. 7 in the first embodiment.
[0095]
Finally, in S35, the relationship comparison unit 9 displays the relationship stored in the input information storage unit 8, that is, the information shown in FIG. 19 and the relationship stored in the search result storage unit 4, such as FIG. The information shown is compared and the matching information is displayed on the display means 6. In this example, the relationship stored in the input information storage means 8 is a relationship of “wo” as shown in FIG. Comparing this relationship with the relationship stored in the search result storage means 4 shown in FIG. 7, two searches with the title “release proprietary WS” and “P release WS” Since the result has “Wo”, it matches. The matched search result is displayed on the display means 6.
[0096]
20 and 21 are explanatory diagrams of display examples of search results in the third embodiment of the information search apparatus of the present invention. Only search results that satisfy the relationship of the search request input by the user are extracted. That is, the displayed search results are narrowed down to only those intended by the user. Therefore, there is no need to display the relationship information as in the first and second embodiments described above. For example, as shown in FIG. 20, the search results may be displayed in order in the display area. Alternatively, as shown in FIG. 21, a title or the like may be temporarily displayed in the candidate display area, and a sentence may be displayed by selecting the candidate. Alternatively, it may be displayed by other methods. In addition, as shown in FIGS. 20 and 21, related words may be underlined, inverted, or displayed with different colors and brightness.
[0097]
Thus, according to the third embodiment of the present invention, the user can input a search request in a natural language, and the search operability can be improved. In addition, since a plurality of words related to each other and a relationship between the words are extracted from a search request input in a natural language and a search result having the relationship is presented to the user, the information desired by the user can be accurately obtained. It is possible to search well.
[0098]
Next, a fourth embodiment will be described. The configuration of the fourth embodiment is the same as that of the above-described third embodiment. In the fourth embodiment, the relationship comparison unit 9 calculates the degree of coincidence when comparing the relationship stored in the input information storage unit 8 and the relationship stored in the search result storage unit 4. The information held in the search result storage unit 4 is rearranged and presented based on the degree of coincidence. Also in the case of this embodiment, the relationship extracted from the search result may be obtained directly from the relationship extraction means 5.
[0099]
FIG. 22 is an explanatory diagram of an example of a relationship comparison table in the fourth embodiment of the information search apparatus of the present invention. When calculating the degree of coincidence, the relationship comparison unit 9 can use a relationship comparison table as shown in FIG. 22, for example. The relationship comparison table uniquely determines the relationship extracted from the search request stored in the input information storage unit 8 and the relationship extracted from the search result stored in the search result storage unit 4. The degree of coincidence can be obtained. In FIG. 22, the case where the relationship is coincident is 100, and the degree of coincidence in each case is indicated by a numerical value.
[0100]
The operation in the fourth embodiment is the same as that in the third embodiment described above except for the operation in the relationship comparison means 9. Only the operation of the relationship comparison means 9 will be specifically described below.
[0101]
As in the third embodiment, as a result of extracting the relationship from the search request, the storage content of the input information storage means 8 is as shown in FIG. 19, and as a result of extracting the relationship from the search result, the search result A case where the storage contents of the storage means 4 are as shown in FIG. 7 will be described. In other words, the relationship stored in the input information storage unit 8 is a relationship of “wo”.
[0102]
The relationship comparison means 9 determines the degree of coincidence using the relationship comparison table shown in FIG. First, the relationship “wo case” stored in the input information storage unit 8 is compared with the relationship “wo case” of the first search result stored in the search result storage unit. According to the relationship comparison table shown in FIG. 22, the degree of coincidence has a value of “100”, so this value is stored in the coincidence degree column of the search result storage unit 4. Similarly for the next search result, the relation “wo case” stored in the input information storage unit 8 is compared with the relation “one sentence” of the search result, and the degree of coincidence “50” is obtained from the relation comparison table. . In this way, the degree of coincidence with all the search results stored in the search result storage means 4 is determined. FIG. 23 is an explanatory diagram showing an example of a search result including a matching degree determination result in the fourth embodiment of the information search apparatus of the present invention. The degree of coincidence is determined for all the search results, and the obtained degree of coincidence value is stored in the search result storage unit 4. For example, the degree of coincidence is obtained as shown in FIG. 23 from the relationship between the search requests shown in FIG. 19 and the relationship between the search results shown in FIG.
[0103]
FIG. 24 is an explanatory diagram of a display example of search results in the fourth embodiment of the information search apparatus of the present invention. As shown in FIG. 23, when the degree of coincidence is obtained, the search results are rearranged in the descending order of the obtained degree of coincidence and displayed on the display means 6. When the search results shown in FIG. 23 are rearranged in descending order of the degree of coincidence, the title order becomes the order shown in the candidate display area of FIG. This display order is close to the search request input by the user. Therefore, the user can easily reach a desired search result by referring to the search result in order from the top.
[0104]
In the display example shown in FIG. 24, only the titles are displayed in the descending order of the matching degree in the candidate display area. However, the matching degrees are displayed together, and each search is performed as in the first embodiment. It is also possible to display the relationship of results.
[0105]
FIG. 25 is a block diagram showing a fifth embodiment of the information search apparatus of the present invention. The reference numerals in the figure are the same as those in FIGS. In the fifth embodiment, the synonym storage means 7 described in the second embodiment is provided in the fourth embodiment described above. In this case, in addition to the word extracted from the search request by the relationship extraction means 5, a search is performed for a word similar to that word, so the relationship comparison means 9 not only compares the relationship but also compares the word similarity. judge.
[0106]
The synonym storage means 7 holds a list of words similar to a certain word. The information retrieval unit 2 obtains a synonym obtained from the synonym storage unit 7 for the word extracted from the search request stored in the input information storage unit 8 by the relationship extraction unit 5. Then, information is searched from the information storage unit 3 using the word extracted from the search request and the synonym obtained from the synonym storage unit 7 as a search key. The searched information is stored in the search result storage means 4. The relationship comparison unit 9 includes a plurality of words stored in the input information storage unit 8 and their relationship, and a plurality of words extracted from each search result stored in the search result storage unit 4 and their relationship. Are compared to determine the degree of coincidence. Then, based on the degree of coincidence, the information held in the search result storage unit 4 is rearranged and displayed on the display unit 6. Of course, the information retrieved by the information retrieval unit 2 may be directly transferred to the relationship comparison unit 9, and the relationship extraction unit 9 may directly transfer the extracted relationship to the relationship comparison unit 9.
[0107]
The operation of the fifth embodiment of the present invention will be described based on a specific example. The synonym storage means 7 holds similar words such as “computer” or “computer” for the word “workstation” and “create” for the word “development”. Suppose you hold the correct word. For example, when a search request “develop workstation” is input from the input means 1, it is stored in the input information storage means 8, and the words and relations are extracted by the relation extraction means 5, for example, as shown in FIG. Information is stored in the input information storage means 8. The information search means 2 refers to the synonym storage means 7 based on the words “workstation” and “development” extracted from the search request. As a result, the words “computer” and “computer” similar to the word “workstation” are obtained. In addition, the word “creation” similar to the word “development” is obtained. Then, using the word obtained from the input information storage means 8 and the word obtained from the synonym storage means 7, for example,
(Workstation or computer or computer) & (development or creation)
The information search means 2 performs a search according to the search condition.
[0108]
FIG. 26 is an explanatory diagram showing an example of a search result in the fifth embodiment of the information search apparatus of the present invention. By searching using similar words as described above, not only information with “workstation” and “development” as keywords, but also information with “computer”, “computer”, and “creation” are searched. Has been.
[0109]
FIG. 27 is an explanatory diagram showing an example of a search result after the relation extraction process in the fifth embodiment of the information search apparatus of the present invention. When the search result shown in FIG. 26 is obtained and stored in the search result storage unit 4, the relationship extraction unit 5 extracts the relationship between the search keys. This process is the same as in the above-described embodiments. The contents of the search result storage means 4 after the relation is extracted by the relation extraction means 5 are shown in FIG.
[0110]
When the relationship between each search result is extracted as shown in FIG. 27, the relationship comparison means 9 determines the degree of coincidence of the relationship in the same manner as in the fourth embodiment, and also determines the degree of coincidence of words. Perform a comprehensive decision on the degree of match. Here, it is assumed that the relationship obtained from the search request is “wo” as shown in FIG. 19, for example.
[0111]
Next, as in the fourth embodiment, the degree of coincidence between the relationship “wo case” obtained from the search request and each relationship shown in FIG. 27 is calculated using, for example, the relationship comparison table shown in FIG. The determination is made and stored in the search result storage means 4. For example, since the first search result has the relationship “wo”, the degree of coincidence of the relationship is 100. Also, since the second search result has the relationship “one sentence”, the degree of coincidence of the relationship is 50.
[0112]
Further, the degree of matching of the words is determined by comparing the words “workstation” and “development” shown in FIG. 19 with the search keys of the respective search results shown in FIG. In this example, the calculation is performed assuming that all the synonyms obtained from the synonym storage means 7 have a degree of coincidence of “90”. However, the present invention is not limited to this, and the synonym storage means 7 may store a synonym and similarity (matching degree) as a pair, or may provide a relationship comparison table for calculating the matching degree of words. Good.
[0113]
Finally, the overall matching degree is determined from the relation matching degree and the word matching degree, and stored in the search result storage means 4. Here, the average of the three matching degrees of the matching degree of the relationship, the matching degree of the search key 1 and the matching degree of the search key 2 is calculated as the overall matching degree. Of course, the overall matching degree may be calculated by other methods. For example, various methods such as weighting each matching degree and calculating a weighted average can be used.
[0114]
FIG. 28 is an explanatory diagram showing an example of the search result after the matching degree determination in the fifth embodiment of the information search apparatus of the present invention. For each search result, the degree of coincidence between the relation extracted by the relation extracting means 5 and the relation extracted from the search request and the degree of coincidence between the word extracted from the search request and the word used as the search key are obtained. The total matching degree is retrieved from these matching degrees and stored.
[0115]
As shown in FIG. 28, when the degree of coincidence is obtained, these are rearranged according to, for example, the whole degree of coincidence and displayed on the display means 6. As a display method, for example, as shown in FIG. 24 in the above-described fourth embodiment, the titles of the search results are displayed in the candidate display area in descending order of the overall matching degree, and the sentence having the selected title Can be configured to be displayed in the result display area. In addition, it is possible to change the display by rearranging on the basis of the degree of coincidence of relationships and the degree of coincidence of words according to instructions from the user. Also in the case of the fifth embodiment, for example, it is possible to display the overall matching degree together with the title, or to display the relation matching degree and the word matching degree.
[0116]
The information with the higher degree of coincidence is the information desired by the user or the information closer to the desired information. The user desires by referring to the search results displayed in descending order of the degree of coincidence. Information can be easily obtained. As described above, in the fifth embodiment described above, searching using synonyms can prevent omission of search and can be displayed in descending order of coincidence, so that information desired by the user can be easily obtained. I have to.
[0117]
In the above third to fifth embodiments, the search request input from the input means 1 has been described as being in natural language, but of course, it is configured to input a plurality of words and their relationships. Also good. At this time, for example, as in the above-described specific example, when searching for information on the contents for developing a workstation, the words “workstation” and “relation” and the relationship “wo” may be input.
[0118]
In the second and fifth embodiments described above, whether to search by acquiring synonyms from the synonym storage means 7 or to search only by the input search request is switched according to the user's instruction. It is possible to configure.
[0119]
Further, in each of the first to fifth embodiments described above, the relationship is extracted from the search result by comparing the word string and the extraction rule, and “wo” and “ga” are used as the relationship. In addition to this, case relationships obtained by syntactic / semantic analysis of sentences may be used. The case relationship obtained when syntax / semantic analysis is performed is, for example,
Development: <Target> Workstation
Development: <Subject> Company A
It ’s like that. <Subject>, <Subject>, etc. are case-related, indicating that the subject of "Development" is "Workstation" and the subject of "Development" is "Company A". For example, from the sentence "Company A develops a workstation with a RISC-type central processing unit"
Development: <Target> Workstation
Development: <Subject> Company A
Place: <Target> RISC type central processing unit
Place: <Location> Workstation
Can be extracted and searched by the search key "workstation &development"
Target: <Development> Workstation
The relationship may be displayed as follows.
[0120]
In addition to these, various relationships between words obtained by analyzing sentences may be used. Further, in each of the embodiments described above, the number of words serving as search keys has been described as two, but three or more words and a relationship between them may be used.
[0121]
In each of the above-described embodiments, the case where information including Japanese text is stored in the information storage means 3 and a search is performed by a search request in Japanese has been described as an example. In these embodiments, even when a language other than Japanese is used, for example, information including English text is searched by an English search request, the same configuration can be made. However, in each of the above-described embodiments, for example, information including English text cannot be searched by a search request in Japanese. In the following, some embodiments that can perform such multilingual search are shown.
[0122]
FIG. 29 is a block diagram showing a sixth embodiment of the information search apparatus of the present invention. In the figure, the same parts as those in FIG. Reference numeral 61 is a search request analysis unit, 62 is a bilingual dictionary, 63 is a relationship analysis unit, 64 is a language analysis switching unit, 65 is a language-specific language analysis unit, 66 is a language-specific relationship extraction unit, and 67 is a search result output unit.
[0123]
The information storage means 3 stores information including at least text written in a natural language. The description of the text is not limited to Japanese, and various languages such as English, German and French are used.
[0124]
As in the above-described embodiments, the input unit 1 can input a plurality of words, sentences, and the like as search requests from an input device such as a keyboard. The search request analysis unit 61 divides and holds the search request input from the input unit 1 into a plurality of words serving as search keys.
[0125]
The bilingual dictionary 62 is an extension of the synonym dictionary to multiple languages. As described above, the synonym dictionary is a list of words having similar meanings with respect to a certain word. For example, in the synonym dictionary, “manufacturing”, “production”, “production” and the like are obtained for the word “production”. The bilingual dictionary 62 includes words having similar meanings recorded not only in the same language but also in a plurality of languages. For example, for the same “production”, Japanese synonyms and “production”, “manufacture”, “create”, etc. in English, and “schaffen”, “produzieren”, etc., are obtained in German. In the bilingual dictionary 62, these words can be stored as a list, for example, in pairs with the respective languages. It is good to make it so that this list can be obtained from any word in any language.
[0126]
The information search unit 2 expands each of the plurality of words divided by the search request analysis unit 61 using the bilingual dictionary 62. Assume that the words forming the original input search request are two words a and b. The expanded words are a1, a2,. . . , Am and b1, b2,. . . , Bn. These expanded words include synonyms of multiple languages. The information retrieval unit 2 retrieves information from the information storage unit 3 using each expanded word as a key. The search key here is (a1 ORa2 OR... OR am) AND (b1 OR b2 OR... OR bn). The search method can use the existing technology as in the above-described embodiment. Text in information obtained by searching is described in various languages.
[0127]
The relationship analysis unit 63 requests the language analysis switching unit 64 to analyze the text in the retrieved information, and obtains the relationship in the text of the word used as the search key. Then, the information obtained by the search, the word obtained from the language analysis switching means 64, and the relationship thereof are sent to the search result output means 67. Here, it is determined whether or not the search key used for the search exists in one sentence. If the search key exists in one sentence, the analysis is requested. If the search key does not exist in one sentence, the analysis is performed. Is not requested, and the subsequent processing is not performed. However, when there is no search key in one sentence, a relationship such as “NO-REL” representing irrelevance is given as a relationship, and the output is directly output from the relationship analysis unit 63 and passed to the search result output unit 67. May be. As a result, at least a function equivalent to a Boolean search can be achieved.
[0128]
The language analysis switching means 64 selects a corresponding one of the plurality of language-specific language analysis means 65 according to the type of language describing the text in the information obtained as a search result, and selects the selected language-specific language analysis means. Pass information to. As a method for discriminating the type of language in which the text is described, information indicating the language is stored together with the text, and a method of referring to the information indicating the language or a method of analyzing the characteristics of the text is used. be able to. In the case of the former method, it is necessary to add information indicating the language used for the text description to each piece of information stored in the information storage means 3. In the case of the latter method, the determination can be made by obtaining the language of each word constituting the text using the bilingual dictionary 62 or a similar dictionary. The dictionary used at this time can be sufficiently used even with about 10 words. Here, it is assumed that the former method is adopted and the language type is described in the information in advance. As the description method, there are a method of describing separately from the text of the main body and a method of embedding in the main body using a tag or the like, like the title and the creation date. The latter method can cope with a case where a part of the text is described in another language. In this embodiment, for simplicity, a method described separately from the former body text is adopted.
[0129]
The language-specific language analyzing means 65 is provided for each language, and has a grammar rule, a dictionary, etc. dedicated to the language, and has a configuration suitable for analyzing each language. However, a common analysis engine may be used. On the other hand, it is also possible to change the level of analysis for each language. For example, a common language analyzer can be used for English and German, and a configuration in which Japanese is analyzed by morphological analysis and pattern matching can be used. The analysis result is sent to the language-specific relationship extraction unit 66.
[0130]
A language-specific relationship extraction unit 66 is also provided for each language, and extracts relationships between a plurality of words from the analysis result sent from the corresponding language-specific language analysis unit 65. In the case of Japanese relation extraction means, as in each of the embodiments described above, a plurality of words related to each other are extracted by extracting a string of words from the morphological analysis result of the sentence and collating the word string with the relation extraction rules. Extract relationships between them. In English, syntax (tree structure) matching can be used.
[0131]
The search result output means 67 is composed of an output device such as a display, for example, and displays the word groups extracted by the language-specific relationship extraction means 66 and their relationships, searched information, and the like.
[0132]
FIG. 30 is a flowchart showing an example of the operation in the sixth embodiment of the information search apparatus of the present invention. First, in S71, a search request is accepted from the input means 1. The accepted search request is divided into a plurality of words serving as search keys by the search request analysis means 61 in S72. Next, in S73, the information search means 2 expands the plurality of words obtained in S72 using the bilingual dictionary 62. In S74, a search expression is created using the expanded result, and the search is performed.
[0133]
In S75, the type of language used for describing the text in the information obtained as a result of the search is determined, and processing corresponding to the type of language is selected. The processes of S76 and S77 are provided for each language. Only one is shown here. In S76, the language-specific language analysis unit 65 analyzes the text in the information obtained as a result of the search. In S77, the language-specific relationship extraction unit 66 uses the relationship extraction rule to extract a plurality of words used as search keys and the relationship between these words.
[0134]
In S <b> 78, the relationship analysis unit 63 passes the word groups extracted by the language-specific relationship extraction unit 66, their relationship, and information obtained as a result of the search to the search result output unit 67 and outputs from the search result output unit 67. Do
[0135]
Hereinafter, an example of the operation in the sixth embodiment of the information search apparatus of the present invention will be described using a specific example. FIG. 31 is an explanatory diagram showing an example of information stored in the information storage means 3 in the sixth embodiment of the information search apparatus of the present invention. In this specific example, the information stored in the information storage means 3 includes at least text, and a language mark indicating the type of language used for describing the text is given. “J” indicates Japanese, “D” indicates German, and “E” indicates English. Of course, information having text written in another language may be stored. The information may include pictures and figures. As a keyword for searching for each piece of information, a word used in the text in each piece of information is used. In the example shown in FIG. 31, the keywords are collected as an index, and each keyword and a pointer to each piece of information having the keyword are stored in association with each other. This pointer is indicated by an arrow line in FIG. Of course, other information management methods may be adopted.
[0136]
FIG. 32 is an explanatory diagram of an example of a bilingual dictionary in the sixth embodiment of the information search apparatus of the present invention. Here, a list of words having similar meanings in each language is registered as a pair with symbols for identifying language types, here “J”, “E”, and “D”. In the example shown in FIG. 32, for the word “production”, the words “production”, “manufacturing”, “production”, “production”, together with the symbol “J” indicating Japanese, “Production”, “manufacture” and “create” words are registered together with the symbol “E” indicating English, and “schaffen” and “produzieren” words are registered together with the symbol “D” indicating German. ing. Similarly, the word “alcohol” includes “alcohol”, “alcohol”, “alcohol”, “beer”, “wine”, “brandy” in Japanese, and “alcohol”, “ “wine”, “beer”, “brandy”, “champagne” are registered, and German “Alkohol”, “Champagne”, “Branttwein”, “Wain”, “Bier” are registered.
[0137]
A search request is input from the input means 1. In this specific example, the search request is input by separating a plurality of words with a delimiter such as a comma. Of course, other delimiters such as a blank may be used. Here, “wine, production” is given as an input. The search request analysis means 61 divides the input search request into a plurality of words that serve as search keys. In this specific example, the search request analysis means 61 obtains “wine” and “production”.
[0138]
The information search means 4 develops each word divided by the search request analysis means 61 using, for example, the bilingual dictionary 62 shown in FIG. “Wine” means “alcohol”, “alcohol”, “alcohol”, “beer”, “wine”, “brandy”, “alcohol”, “wine”, “beer”, “brandy”, “champagne”, “Alkohol” "," Champagne "," Branttwein "," Wain "," Bier ". “Production” is expanded into “production”, “manufacturing”, “production”, “production”, “production”, “manufacture”, “create”, “schaffen”, and “produzieren”. Then, a conditional expression in which the expanded words are connected by OR is created, and a search expression is created by connecting two conditional expressions by AND. For example, (alcohol OR liquor OR liquor OR beer OR wine OR brandy OR alcohol by wine OR beer OR brandy OR champagne OR Alcohol OR Champ OR OR produced by OR OR OR search OR OR produzieren) is created. The information storage means 3 is searched using this search formula.
[0139]
The information shown up to a part of the text content in FIG. 31 shows an example of information searched by this search formula. For example, information (1) has “manufacturing” and “alcohol”, information (2) has “Wein” and “produzieren” variants “geproduzieret”, and information (3) has “ Information (4) has the noun “production” for “wine” and “production”, and information (5) has “alcohol” and “production”. ing. Therefore, the above retrieval formula is satisfied and such information is retrieved.
[0140]
The search results obtained in this way are information including text and the type of language describing the text. These are transferred to the next relation analysis means 63. The information including the text obtained by the search may pass an address to the information storage unit 3 instead of passing the information itself to the relationship analysis unit 63. This address can be composed of the position indicating the head of the information and the relative position of the sentence to be analyzed in the information. The absolute address of the target sentence may be passed as the address, or only the head position of the information may be used.
[0141]
The relation analysis unit 63 receives the search result from the information search unit 2, refers to the text in the searched information, and determines whether or not the search key used for the search exists in one sentence. In the five pieces of information shown in FIG. 31, since the information (1) includes the search keys “manufacturing” and “alcohol” in different sentences, the subsequent processing is not performed on this information. In this case, the information (1) is removed from the search result, or the relationship “NO-REL” representing irrelevance is given as the relationship, and the output is directly output from the relationship analyzing means. The information (2) to (5) is sequentially sent to the language analysis switching means 64 because the search key exists in one sentence.
[0142]
The language analysis switching means 64 discriminates the language type in which the text in the information is described from the search result sent from the relation analysis means 63, and a plurality of language analysis means for each language according to the language type. A corresponding one is selected from 65 and a search result is passed. For example, in the information (2) shown in FIG. 31, since the text is described in German, the language analysis switching means 64 selects the German language analysis means. Similarly, in the case of the information (3) and (5), for example, the Japanese language analysis means as used in the above first to fifth embodiments is selected. In the case of information (4), an English language analysis means is selected.
[0143]
Each language-specific language analyzing means 64 analyzes a sentence including a search key for each language. For example, the Japanese language analyzing means performs morphological analysis of the sentence as shown in the first to fifth embodiments, and generates, for example, a pair of words and parts of speech as shown in FIG. FIG. 33 is an explanatory diagram of an example of an English syntax analysis result in the sixth embodiment of the information search apparatus of the present invention. The English analysis means can be configured to analyze one sentence including the search key and generate, for example, a syntax tree as shown in FIG. FIG. 33 shows an example of an analysis result of a sentence “The ABC Ltd. will produce wine.”. Here, for the sake of understanding, the analysis result is shown as a tree diagram and shown with attributes. The data structure is arbitrary. The analysis result is sent to the language-specific relationship extraction unit 66 corresponding to each language.
[0144]
Each language-specific relationship extraction unit 66 extracts a plurality of words and a relationship between these words based on the analysis result for each language. At this time, a relationship extraction rule is used. FIG. 34 is an explanatory diagram of an example of a Japanese relationship extraction rule in the sixth embodiment of the information search apparatus of the present invention, and FIG. 35 is an explanatory diagram of an example of an English relationship extraction rule. In the case of the Japanese relationship extraction means, the relationship extraction rule as shown in part in FIG. 34 is used to extract two words and the relationship between the words. This relationship extraction process in Japanese is substantially the same as the operation of the relationship extraction means 5 used in the first to fifth aspects described above. The relationship extraction rule shown in FIG. 34 is almost the same as the relationship extraction rule shown in FIG. 6, and only the notation of the extracted relationship is different. Here, in order to match the notation of the relationship in other languages, the notation of the relationship as shown in FIG. 34 is used.
[0145]
In the English relationship extraction means, for example, a relationship extraction rule as shown in part in FIG. 35 can be used. In FIG. 35, the left side of “−>”, that is, the left side, is a word pattern. Among them, (# 1 # 2 # 3 # 4) represents a tree structure having # 1 as a parent and three children # 2, # 3, and # 4 below it. The portion enclosed by “<” and “>” is the designation of the attribute of the node. “Cat” is a category and indicates a syntactic role. “Form” indicates the word itself, and “infl” indicates a change in word form. “Pres-p” indicates the present participle (ing form), and “past-p” indicates the past participle. The right side is a relational expression to be extracted, where the first term is a relation name and the second and third terms are related words. “# 2.head” indicates a modified noun phrase, and “# 3.main” indicates a main verb of the verb phrase (excluding auxiliary verbs). Although not shown in the rules, verbs are basically kept in their original form and nouns in their singular form in relational expressions. In the example of the analysis result shown in FIG. 33, when collating with the English relationship extraction rule shown in FIG. 35, the first rule and the second rule are applied, and the following relationship between words is extracted. .
(NOM “ABC Ltd.” “production”)
(ACC “WINE” “production”)
[0146]
The words and the relations between the words thus obtained are returned to the relation analysis means 63 via the language analysis switching means 64 or directly, and include information including the language type and the text obtained by the search. At the same time, it is sent to the search result output means 67.
[0147]
The search result output unit 67 displays the analysis result sent from the relationship analysis unit 63 on the display device, and displays information in accordance with an instruction from the user. At this time, if the address indicating the information is sent, the information is extracted from the information storage means 3 based on the address and displayed. FIG. 36 is an explanatory diagram showing an example of a display screen in the sixth embodiment of the information search apparatus of the present invention. In this example, a plurality of words extracted by the language-specific relationship extraction means 66 and their relationships are displayed in the candidate display area. The display of the relationship displays the relationship name defined in the relationship extraction rule as it is. However, the present invention is not limited to this. For example, the case (such as the case used in the first to fifth embodiments described above) You may make it display a case, a de-grade, ...). Further, if there is a margin in the display area, for example, as shown in FIG. 13, a part of the title and information can be displayed. Of course, for example, as shown in FIG.
[0148]
Next, a description will be given of a seventh embodiment of the information search apparatus of the present invention. The configuration in this embodiment is the same as that shown in FIG. 29 described in the sixth embodiment. Also, the operation is the same as that of the sixth embodiment except for the processing when outputting the search result. In the seventh embodiment, when the search result is output, the clarity of the relationship extracted from the searched information and / or the similarity of the words are obtained, and these are used to arrange the search results. Change the output. Although these processes are described here as being performed by the relationship analysis unit 63, they may be configured to be performed by the search result output unit 67.
[0149]
FIG. 37 is an explanatory diagram of rules that give clarity of relations in the seventh embodiment of the information search apparatus of the present invention. The relationship analysis means 63 has a rule that pairs the relationship between words with the value of clarity between the words. This rule is prepared for each language and is defined so that the value becomes larger as the relationship between words becomes clearer. For example, in both English and Japanese, when nouns are continuous (relation “UNSPEC”), the semantic connection between the nouns is ambiguous, so it is determined that the clarity is low. In the case of Japanese, embedding is the same. Using such rules, a value of clarity is obtained from the relationship extracted by the language-specific relationship extraction means 66. For example, when the relationship “GA-NOMINATIVE” is extracted by the Japanese relationship extracting means, the value of clarity is “5”.
[0150]
Next, the similarity will be described. As described in the sixth embodiment, the word given as the search key is expanded using the bilingual dictionary 62 in the information search means 2, and a search formula is set using the expanded word as the actual search key. Generate and search. Therefore, the word that matches the search key included in the searched information may be a word given as a search key or a word developed in the bilingual dictionary 62. At this time, it is considered that it is possible to indicate how close the searched information is to the search request by obtaining the similarity between the word given as the search key and the word included in the searched information. It is done.
[0151]
Here, as a method for obtaining the similarity, a method is used in which words registered in the bilingual dictionary 62 are hierarchically constructed, and the similarity is obtained by a difference in the hierarchy. FIG. 38 is an explanatory diagram of an example of a bilingual dictionary in the seventh embodiment of the information search apparatus of the present invention. In the example shown in FIG. 38, a hierarchy is configured for each language, and when a plurality of words are described in each node, they are regarded as the same group. In addition, words connected by broken lines between languages are also regarded as the same group. Here, symbols indicating language types are omitted. As for the similarity of words, for example, the similarity between words belonging to the same group is set to 10, and the degree of similarity is reduced by 3 when going up the hierarchy and by 1 when going down the hierarchy. In the example shown in FIG. 38, “beer” and “wine wine” have a similarity of 10. In the case of “beer” and “alcohol”, if the input is “alcohol”, the similarity of “beer” is 9, and if the input is “beer”, the similarity of “alcohol” is 7. Similarly, in the case of “beer” and “Japanese sake”, the level is increased by one and the similarity is decreased by six. According to the similarity obtained in this way, for example, the search results may be rearranged and output in descending order of similarity. In addition, it is possible to limit the search results to be output, such as outputting information with a similarity greater than 0. If there is too much output even with this restriction, you can reduce the initial value of similarity, increase the threshold, or increase the difference in similarity when moving up and down the hierarchy. . Further, these values are not fixed in advance, but can be configured to be adjustable by the user. In the case of adopting such a method for obtaining the similarity, it is configured so that the relationship analysis means 63 can refer to the bilingual dictionary 62, although not shown in FIG.
[0152]
Search results can also be rearranged using both the above-described relationship clarification and word similarity. For example, when “word 1” and “word 2” and their relationship are extracted by the language-specific relationship extraction unit,
Using the similarity of “word 1” × the similarity of “word 2” × the clarity of the relationship, the search results can be rearranged.
[0153]
FIG. 39 is an explanatory diagram showing an example of a display screen in the seventh embodiment of the information search apparatus of the present invention. 37 shows a display screen when a search similar to that in the sixth embodiment shown in FIG. 36 is performed. The word “sake” given as the search key, the similarity between the words “Wein”, “Wine”, “alcohol” and “sake” extracted from the searched information, and the word “manufacturing” given as the search key , Similarities to the words “produzieren”, “product”, “production”, “production” extracted from the searched information, and the relationships “ACC”, “ACC”, “ From the values of clarity obtained from “DE-INST” and “NO-UNSPEC”, the scores are calculated by the above formula and rearranged. Compared to FIG. 36, the display order of the candidate display areas is different.
[0154]
The above-mentioned score is an example, and other calculation methods such as using three addition values can be used. Furthermore, these values may be obtained by a method other than the above-described method for obtaining the degree of similarity or clarity, or a measure other than the degree of similarity or clarity may be introduced.
[0155]
FIG. 40 is a block diagram showing an eighth embodiment of the information search apparatus of the present invention. In the figure, the same parts as those in FIG. Reference numeral 68 denotes a relationship correspondence storage means. In the eighth embodiment, an example is shown in which a plurality of words and their relationships are given as search requests instead of word sequences. As a search request, a plurality of words and their relations are given as a case where a plurality of words and relations are given separately from the user, and given in a natural language as shown in the above third to fifth embodiments. It may be possible. Here, a case where a search request is given in a natural language will be described.
[0156]
The search request analysis unit 61 analyzes the search request input from the input unit 1 and divides the search request into a plurality of words serving as search keys and their relationships. For this analysis, language-specific language analysis means can be used. That is, the search request is sent to the relation analysis unit 63 or the language analysis switching unit 64 together with the language type in which the search request is described, and the analysis is requested. The language type is specified by, for example, a method in which a user mainly specifies a language to be used in advance, a method in which a language type is specified from the input unit 1 at the same time as a search request, or a simple analysis is performed on the search request. The method of making decisions can be considered.
[0157]
The information search unit 2 searches only by using the word obtained from the search request analysis unit 61 and the word, and expands the word by the bilingual dictionary 62 as in the sixth embodiment. The search result is sent to the relation analysis means 63, and also between the words from the information searched using the language analysis switching means 64, the language-specific language analysis means 65, and the language-specific relation extraction means 66, as in the sixth embodiment. Extract relationships.
[0158]
The relationship analysis means 63 compares the relationship in the search request extracted by performing the analysis previously with the relationship extracted from the searched information. In this comparison, the relationship correspondence storage means 68 is used.
[0159]
The relationship correspondence storage means 68 has two types of relationship correspondence information between languages and relationship correspondence information within a language. Inter-language relationship correspondence information is information that associates relationships extracted in each language. The intra-language relationship correspondence information is information that associates a certain relationship with a relationship that does not contradict the relationship. This intra-language relationship correspondence information is prepared for each language.
[0160]
Now, the search request is described in the first language, and the relationship extracted from the search request is REL-S. Assume that the relationship REL-S in the first language and the relationship REL-T in the second language are associated as the inter-language relationship correspondence information. In addition, as the intra-language relationship correspondence information in the first language, the relationship REL-S and the relationships REL-S2, REL-S3,. . . Are associated with each other, and the relationship REL-T and the relationships REL-T2, REL-T3,. . . Are associated with each other. When the text in the retrieved information is described in the second language, the second in which the text in the information retrieved by the inter-language relationship correspondence information is described with respect to the retrieval request relationship REL-S. Get the relationship REL-T in the language. Further, the relationship REL-T is expanded using the in-language relationship correspondence information in the second language, and the relationships REL-T2, REL-T3,. . . Get. Further, the relationship REL-S extracted from the search request is expanded using the in-language relationship correspondence information of the first language, and the relationships REL-S2, REL-S3,. . . Get. When the text in the retrieved information is described in the third language and the relationship in the third language corresponding to the relationship REL-S of the retrieval request is not associated in the inter-language relationship correspondence information, , The relationship REL-S extracted from the search request using the in-language relationship correspondence information of the first language is expanded, and the relationships REL-S2, REL-S3,. . . Get. Then, for each of the obtained relationships, the inter-language relationship correspondence information is referred to obtain a relationship in the third language. In this case, the relationship in the obtained third language is not expanded further.
[0161]
FIG. 41 is an explanatory diagram of an example of inter-language relationship correspondence information in the eighth embodiment of the information search apparatus of the present invention, and FIG. 42 is an explanatory diagram of an example of the intra-language relationship correspondence information in Japanese. In the inter-language relationship correspondence information shown in FIG. 41, the English relationship “ACC”, the German relationship “ACC”, and the French relationship “ACC” are associated with the Japanese relationship “WO-ACCUSIVE”. Yes. For example, if the input search request is “A to B”, “WO-ACCUSIVE” is obtained as the Japanese relationship. When the language in which the text in the searched information is described is English, the English relation “ACC” is obtained from the Japanese relation “WO-ACCUSIVE” obtained from the search request. Further, based on the English relation “ACC”, the English correspondence relation information in the language is referred to. Here, the relationship correspondence information in English is not shown, but for example, the relationship “UNSPEC” or the like is obtained as a relationship consistent with the English relationship “ACC”. On the other hand, using the Japanese intra-language relationship correspondence information shown in FIG. 42, all relationships that are consistent with the Japanese relationship “WO-ACCUSIVE” are obtained. In FIG. 42, the relationships “HA-THEME”, “MO-ADDITION”, “CASE-UNSPEC”, and “UNSPEC” are associated with each other. All these relationships are treated as being consistent with the original relationship "WO-ACCUSIVE".
[0162]
The relationship analysis means 63 compares the relationship obtained in this way with the relationship extracted from the searched information, and outputs the search result that does not match. Alternatively, information indicating that they do not match is added and output. The search result output means 67 displays the search result output in this manner, for example, on a display device, so that the user can use information having text described using a language different from the language in which the search request is described. Even if it exists, information with similar contents can be obtained. In this way, information similar in content can be obtained regardless of the type of language in which the search request and the text in the information are described.
[0163]
Further, as described in the fourth or fifth embodiment, the relation analysis unit 63 calculates the degree of coincidence using the relation comparison, and rearranges and displays the search results based on the calculated degree of coincidence. It can also be configured to. For the calculation of the degree of coincidence, for example, a relationship comparison table as shown in FIG. 22 is prepared for each language, and this can be used. Further, as shown in parentheses in the intra-language relationship correspondence information of FIG. 42, a value for reducing the similarity is added to the non-conflicting relationship, for example, the initial value of the similarity is set to 5, and only the respective values are similar. It can also be configured to reduce the degree. In this case, with respect to correspondence between languages, for example, it is conceivable that the degree of similarity is uniformly reduced by 1, or the degree of similarity is used as it is. The search results may be rearranged using the similarity value calculated in this way as the degree of coincidence. Furthermore, using the clarity of relations and word similarities described in the seventh embodiment, the search results are rearranged, and combined with the matching degrees, the total matching degree is calculated as in the fifth embodiment. And sort the search results. The calculation of the degree of coincidence and the rearrangement processing can be realized by sharing the processing between the relationship analysis means 63 and the search result output means 67 or both.
[0164]
In the sixth to eighth embodiments described above, it is also possible to output a search result limited to a specific language. For example, by designating a specific language from the input unit 1, when the information retrieval unit 2 expands a word using the bilingual dictionary 62, the search is performed by expanding only the word in the specified language. The analysis unit 63 selects only the information in which the text is described in the specified language and performs the subsequent processing, or the search result output unit 67 relates to the information in which the text is described in the specified language. Can only be configured to display.
[0165]
In the sixth to eighth embodiments, the search result by the information search means 2 is shown to be directly passed to the relation analysis means 63, but it is shown in the above first to fifth embodiments. As described above, the search result storage unit may be used to temporarily store the search result and perform a relationship extraction process or the like.
[0166]
Since the information search device of the present invention can be realized by using it as the information search means 2 without changing the configuration of the conventional information search device, it can be easily configured from the information search device already introduced. .
[0167]
【The invention's effect】
As is clear from the above description, according to the present invention, since the relationship between words used as search keys is extracted from the search results searched by a plurality of search keys, for example, the search results and By displaying the relationship together, it is possible to make it easier to find the information desired by the user compared to the conventional display of only the result. Further, even when a search including synonyms is performed in order to prevent a search omission, it is possible to easily obtain desired information from a large amount of search results by referring to related information.
[0168]
Furthermore, from a search request input in a natural language, a plurality of words related to each other and the relationship between the words are extracted, and the relationship between the words is also determined from a search result obtained by performing a search using the plurality of words as a key. Since they are extracted and compared with each other, the information desired by the user can be searched with higher accuracy than the conventional search using only words. At this time, by calculating the degree of coincidence between the relationship obtained from the search request and the relationship obtained from the search result, and rearranging the display order based on this, the user can refer to the related information in the order of importance. it can. Also in this case, in order to prevent omission of search, it is possible to perform a search including synonyms and present the search information to the user in the order considered to be important based on the matching degree of the relationship and the matching degree of the word. Desired information can be easily obtained from the retrieved results
[0169]
Furthermore, there is an effect that information having a desired content can be easily obtained regardless of the description language of the text in the information stored in the information storage means and regardless of the language used for the search request. is there.
[Brief description of the drawings]
FIG. 1 is a block diagram showing a first embodiment of an information search apparatus of the present invention.
FIG. 2 is a flowchart showing an example of operation in the first exemplary embodiment of the information search apparatus of the present invention.
FIG. 3 is a flowchart showing an example of the operation of the relationship extraction means in the first embodiment of the information search apparatus of the present invention.
FIG. 4 is an explanatory diagram showing an example of a search result using two keys in the first embodiment of the information search apparatus of the present invention;
FIG. 5 is an explanatory diagram showing an example of a morphological analysis result in the first embodiment of the information search apparatus of the present invention.
FIG. 6 is an explanatory diagram of an example of a relationship extraction rule in the first embodiment of the information search apparatus of the present invention.
FIG. 7 is an explanatory diagram showing an example of a search result after a relation extraction process in the first embodiment of the information search apparatus of the present invention.
FIG. 8 is an explanatory diagram showing an example of a search result display after the relationship extraction process in the first embodiment of the information search apparatus of the present invention;
FIG. 9 is an explanatory diagram of another example of search result display after the relationship extraction process in the first embodiment of the information search apparatus of the present invention;
FIG. 10 is a block diagram showing a second embodiment of the information search apparatus of the present invention.
FIG. 11 is an explanatory diagram showing an example of a result of a search using a synonym and a relation extraction process in the second embodiment of the information search apparatus of the present invention;
FIG. 12 is an explanatory diagram showing an example of a search result display after the relationship extraction process in the second embodiment of the information search apparatus of the present invention;
FIG. 13 is an explanatory diagram of another example of search result display after the relationship extraction process in the second embodiment of the information search apparatus of the present invention;
FIG. 14 is a block diagram showing a third embodiment of the information search apparatus of the present invention.
FIG. 15 is a flowchart showing an example of operation in the third embodiment of the information search apparatus of the present invention;
FIG. 16 is a flowchart showing an example of the operation of the relationship extraction means in the third embodiment of the information search apparatus of the present invention.
FIG. 17 is an explanatory diagram showing an example of a search request stored in the input information storage means in the third embodiment of the information search apparatus of the present invention.
FIG. 18 is an explanatory diagram of a morphological analysis result of a search request in the third embodiment of the information search apparatus of the present invention.
FIG. 19 is an explanatory diagram showing an example of a search request after the relation extraction process in the third embodiment of the information search apparatus of the present invention.
FIG. 20 is an explanatory diagram of a display example of search results in the third embodiment of the information search apparatus of the present invention.
FIG. 21 is an explanatory diagram of another display example of search results after the relationship extraction process in the third embodiment of the information search apparatus of the present invention;
FIG. 22 is an explanatory diagram showing an example of a relationship comparison table in the fourth embodiment of the information search apparatus of the present invention;
FIG. 23 is an explanatory diagram showing an example of a search result including a matching score determination result in the fourth embodiment of the information search apparatus of the present invention.
FIG. 24 is an explanatory diagram of a display example of search results in the fourth embodiment of the information search apparatus of the present invention;
FIG. 25 is a block diagram showing a fifth embodiment of the information search apparatus of the present invention.
FIG. 26 is an explanatory diagram showing an example of a search result in the fifth embodiment of the information search apparatus of the present invention;
FIG. 27 is an explanatory diagram showing an example of a search result after the relation extraction process in the fifth embodiment of the information search apparatus of the present invention;
FIG. 28 is an explanatory diagram showing an example of a search result after the matching degree determination in the fifth embodiment of the information search apparatus of the present invention;
FIG. 29 is a block diagram showing a sixth embodiment of the information search apparatus of the present invention.
FIG. 30 is a flowchart showing an example of operation in the sixth embodiment of the information search apparatus of the present invention;
FIG. 31 is an explanatory diagram showing an example of information stored in the information storage means 3 in the sixth embodiment of the information search apparatus of the present invention;
FIG. 32 is an explanatory diagram showing an example of a bilingual dictionary in the sixth embodiment of the information search apparatus of the present invention;
FIG. 33 is an explanatory diagram of an example of an English syntax analysis result in the sixth embodiment of the information search apparatus of the present invention;
FIG. 34 is an explanatory diagram showing an example of a Japanese relation extraction rule in the sixth embodiment of the information search apparatus of the present invention;
FIG. 35 is an explanatory diagram showing an example of an English relation extraction rule in the sixth embodiment of the information search apparatus of the present invention;
FIG. 36 is an explanatory diagram showing an example of a display screen in the sixth embodiment of the information search apparatus of the present invention.
FIG. 37 is an explanatory diagram of rules for giving clarity of relations in the seventh embodiment of the information search apparatus of the present invention;
FIG. 38 is an explanatory diagram showing an example of a bilingual dictionary in the seventh embodiment of the information search apparatus of the present invention;
FIG. 39 is an explanatory diagram showing an example of a display screen in the seventh embodiment of the information search apparatus of the present invention.
FIG. 40 is a block diagram showing an eighth embodiment of the information search apparatus of the present invention.
FIG. 41 is an explanatory diagram showing an example of correspondence relation information between languages in the eighth embodiment of the information search device of the present invention;
FIG. 42 is an explanatory diagram showing an example of correspondence relation information in Japanese language in the eighth embodiment of the information search device of the present invention.
[Explanation of symbols]
DESCRIPTION OF SYMBOLS 1 ... Input means, 2 ... Information search means, 3 ... Information storage means, 4 ... Search result storage means, 5 ... Relation extraction means, 6 ... Display means, 7 ... Synonym storage means, 8 ... Input information storage means, 9 ... Relation comparison means 61 ... Search request analysis means 62 ... Bilingual dictionary 63 ... Relation analysis means 64 ... Language analysis switching means 65 ... Language analysis means for each language 66 ... Relation extraction means for each language 67 ... Search result output Means 68... Relationship correspondence storage means.

Claims (10)

In an information search device for searching for information that matches a search request, information storage means for storing information including at least text, information search means for searching for information from the information storage means using a plurality of words as search keys, and the information A relationship extraction unit that extracts a relationship between a plurality of words used as a search key from information searched by the search unit, and a pair of a search result by the information search unit and a relationship extracted by the relationship extraction unit Display means for displaying, and when the plurality of words are present in one sentence, the relation extraction means extracts a relation between the words by analyzing the one sentence, and the plurality of words are included in one sentence. An information search apparatus characterized by calculating a distance between a sentence including each word when it does not exist. In an information search apparatus for searching for information that matches a search request, information storage means for storing information including at least text, information search means for searching for information from the information storage means using a plurality of words as search keys, and the information Search result storage means for holding information searched from the search means, and a relationship between a plurality of words used as a search key is extracted from the information stored in the search result storage means and stored in the search result storage means A relation extracting means for storing; and a display means for displaying the retrieved information held in the search result storing means and the extracted relation in pairs, wherein the relation extracting means includes the plurality of words. If one sentence is present, one sentence is morphologically analyzed to extract the relationship between words. If the plurality of words are not present in one sentence, the distance between the sentence including each word and the sentence. Information retrieval apparatus and calculates a. Furthermore, it has synonym storage means for holding a list of words similar to a certain word, and the information search means uses the synonym obtained from the synonym storage means for a certain word as a search key to receive information from the information storage means. The information search device according to claim 1, wherein the information search device searches. A bilingual dictionary that inputs a word in a language and outputs a word equivalent to or similar to the word in each language together with the type of the language, and the information search means stores the word given as a search key in the bilingual dictionary Given the word obtained from the bilingual dictionary as a search key to retrieve information from the information storage means, the relationship extraction means, according to the type of language in which the text in each searched information is described, The information search apparatus according to claim 1, wherein a plurality of words used as search keys for each language and relationships between the words are extracted. It further has relation correspondence storage means for holding relation correspondence information describing the relation between words in a certain language and the relation between words in another language, and the relation extraction means comprises a search request written using a certain language, According to the type of language in which the search request is described, a plurality of words given as search keys to the information search means and the relationship between the words are extracted, and each information searched by the information search means In accordance with the type of language in which the text is described, a relationship between a plurality of words used as a search key in the text is extracted for each language, and the relationship between the words extracted from the search request and the information The relation extracted from the information retrieved by the retrieval means is compared using the relation correspondence information held in the relation correspondence storage means, and only information that does not match or contradict is extracted. Information retrieval apparatus according to claim 4, characterized in that. The relation extracting means is based on the relation between the plurality of words extracted from each searched information and the degree of similarity between the plurality of words and the plurality of words given as the search key. 6. The information search apparatus according to claim 4, wherein the searched information is rearranged and output based on both of them. The relationship extraction unit compares the relationship extracted from the search request with the relationship extracted from the information searched by the information search unit to determine a matching level, and determines the information searched based on the matching level. 6. The information search apparatus according to claim 5, wherein the information search apparatus is arranged. It comprises language comparison means for retrieving only information in which the language type in which the text of the information is described matches the specified language type among the information searched by the information search means. The information search device according to any one of claims 4 to 7. In an information search method for searching for information that matches a search request, information including at least text is stored in the information storage means, and information is searched from the information storage means by the information search means using a plurality of words as search keys, A relationship extraction unit extracts a relationship between a plurality of words used as a search key from the retrieved information, and displays the search result and the extracted relationship on a display unit as a pair. When a word is present in one sentence, morphological analysis is performed on the one sentence to extract a relationship between the words. When the plurality of words are not present in one sentence, a sentence between the sentence including each word and the sentence is extracted. An information search method characterized by calculating a distance. In an information search method for searching for information that matches a search request, information including at least text is stored in the information storage means, and information is searched from the information storage means using a plurality of words as search keys by the information search means, The retrieved information is held in the retrieval result storage means, and the relation between the plurality of words used as the retrieval key is extracted from the information stored in the retrieval result storage means by the relationship extraction means, and the retrieval result storage means When the plurality of words are present in one sentence, the retrieved information stored in the retrieval result storage unit and the extracted relationship are displayed on the display unit as a pair. A morphological analysis is performed on the one sentence to extract a relationship between words, and when the plurality of words do not exist in one sentence, a distance between the sentence including each word and the sentence is calculated. Information retrieval how to.

Images