JPH0750487B2 - Information extraction device - Google Patents

Description

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an information extraction device used for text information processing.

2. Description of the Related Art In recent years, research has been underway to obtain a gist from keywords in Japanese sentences. To get an accurate summary,
Extracting the most suitable keyword is an important point. Unlike Japanese languages, such as Western languages, in which words are separated by spaces and the like, characters are written in solid. In order to extract the optimal keyword, it is necessary to divide the input Japanese sentence into words. By the way, there are many character types in Japanese, and the character types contain certain information. Usually, in Japanese sentences, particles, auxiliary verbs, etc. are written in hiragana, and the head of an independent word is often written in a character other than hiragana. Therefore, in order to find a word that is a keyword, it is often used as an index to change from hiragana to characters of a character type other than hiragana. Hereinafter, characters of a character type other than Hiragana will be expressed as "Kanji".

An example of a conventional information extraction device will be described below with reference to the drawings.

FIG. 5 shows a block diagram of a conventional information extraction device. In FIG. 5, 4 is a word extraction unit for extracting a word from an input character string, 5 is a word dictionary having information corresponding to the word, and 6 is a search for the word dictionary 5 using the word extracted by the word extraction unit 4 as a keyword. It is a word dictionary search unit. Sixth
The figure shows a part of the inside of the word dictionary 5.

The operation of the information extraction device configured as described below will be described below with reference to examples.

It is assumed that a Japanese character string, for example, a sentence "I am not impressed with this floor plan" is input. In order to find a word from this character string, in the word extraction unit 4, as described above, mark the positions of "ma" and "feeling" where the character type changes from hiragana to "kanji", and from the mark to the next Take out the character strings up to the mark, "this", "too much floor space", "not impressed". Next, the word dictionary search unit 6 searches the word dictionary 5 with the character string extracted by the word extraction unit 4. If the searched character string does not exist in the word dictionary 5, the word dictionary 5 is searched again with a new character string in which the last character of the character string is deleted. In this way, the keyword "floor plan" and information on the keyword can be obtained. (For example, “Nikkei Electronics” No. 324, pages 199 to 215) Problems to be Solved by the Invention However, in the above configuration, a character including a word having a “Hiragana kanji” part in one word Columns,
For example, if you enter a string that says "has a listening test" or a string that says "Aya to my sister", you will not be able to get the desired keyword "hear" or "Aya to". Become. Then, in order to obtain the desired keyword, a dictionary of words containing "Kanji Hiragana" is provided as an exception dictionary, and the above-mentioned exception is made before matching with a dictionary of words not containing "Kanji Hiragana". A method of matching with a dictionary, or omitting the process of the word extraction unit 4, the word dictionary search unit 6 from the beginning to the end of the input character string.
Must be used to simply match with the word dictionary by the longest match from the beginning of the input string. The former method requires a matching process with two dictionaries at all times, and the latter method has a problem that the number of matching times becomes huge.

In view of the above problems, the present invention is capable of inputting a character string including a word having a part of "Hiragana Kanji" from a desired keyword without providing an exception dictionary and increasing the number of matching times. An information extraction device capable of obtaining information is provided.

Means for Solving the Problems In order to solve the above problems, the information extraction device of the present invention, a keyword including a character string before and after the keyword and information corresponding to the keyword consisting of the same character type character string It is composed of a search key extraction unit that extracts a search key of the same character type character string from the dictionary and the input character string, and a keyword dictionary search unit that searches for a keyword in the keyword dictionary using the search key extracted by the search key extraction unit. , When the search key and the keyword match, the keyword dictionary search unit recognizes the character strings before and after the keyword in the keyword dictionary corresponding to the keyword and the character strings before and after the search key in the input character string. Are compared, and if they match, the information in the keyword dictionary is extracted.

Effect According to the present invention, even if a character string including a word having a "Hiragana kanji" portion is input according to the above-described configuration, a keyword from a desired keyword can be input without providing an exception dictionary and increasing the number of matching times. You will be able to obtain information.

Embodiment Hereinafter, an information extracting apparatus according to an embodiment of the present invention will be described with reference to the drawings.

FIG. 1 is a block diagram of an information extraction device in one embodiment of the present invention. In FIG. 1, 1 is a search key extraction unit for extracting a search key of the same character type character string from the input character string, and 2 is a keyword corresponding to the character string of the same character type. The included keyword dictionary 3 is a keyword dictionary search unit that searches for a keyword in the keyword dictionary 2 using the search key extracted by the search key extraction unit 1, and the keyword dictionary search unit 3 is
When the search key and the keyword match, the character strings before and after the keyword in the keyword dictionary corresponding to the keyword are compared with the character strings before and after the search key in the input character string, and if they match, Extract the information in the keyword dictionary.

The operation of the information extraction device configured as described above will be described below with reference to FIGS. 1, 2, 3, and 4.

First, FIG. 2 shows a storage state of keywords of the same character type character string in the keyword dictionary 2, character strings before and after the keyword, and word information. For keywords in the keyword dictionary, if the word consists of multiple characters, from the beginning to the characters of characters other than hiragana, the same character type character string is used as the keyword, and if the word consists of hiragana only, the hiragana character example Use as a keyword. FIG. 3 shows that the keyword and the word information storage section in FIG. 2 are separated. FIG. 4 shows a process of extracting the optimum information from the input character string.

As a first example, it is assumed that a character string including a word having a "Hiragana kanji" portion in one word, for example, a character string "Ayatori". The search key extraction unit 1 searches for the "Hiragana Kanji" portion in the input character string. And when you find the corresponding "Tori",
Mark the position. Next, I'll search for the first change to a character type that is different from the character that I just found. In the case of this example, "ri" is found and its position is marked. Then, the character string from the previous mark to the next mark is taken out. In this example, it is "Tori". This is,
It becomes a search key. The keyword dictionary search unit 3 searches for the keyword dictionary 2 using this search key "Tori" as a key. Then, "take" is searched for in the keywords, and all words having "take" as a keyword are taken out from the keyword dictionary. Then, among all the extracted words, the character strings before and after the keyword are compared with the character strings existing before and after the search key "Tori" in the input character string, and the matching words are compared. Just take out. as a result,
The word "Ayari" will be extracted.

Next, as a second example, it is assumed that a character string "It is important to spread the knowledge" is input. Similar to the first embodiment, the search key extraction unit 1 searches for the "Hiragana Kanji" portion in the input character string. When it finds the corresponding "look", it marks that position. next,
Now, I'm going to search for the first change to a character type that is different from the character that I found. In this case,
Find the "" and mark its position. And then
Extract the character string from the previous mark to the next mark. In this example, it's a "view." This is the search key. The keyword dictionary search unit 3 uses this search key “Kami” as a key.
Go to search the keyword dictionary 2. Then "search" is searched for in the keywords, and all words having "look" as the keyword are extracted from the keyword dictionary. And
From all the extracted words, the character strings before and after the keyword are compared with the character strings existing before and after the search key "Kami" in the input character string, and only matching words are compared. Take it out. As a result, only the word “view” is extracted and the word “view” is removed.

As described above, according to the present embodiment, a keyword dictionary containing character strings before and after a keyword and information corresponding to a keyword composed of a character string of the same character type, and a search for the character string of the same character type from an input character string. A search key extraction unit for extracting a key and a keyword dictionary search unit for searching for a keyword in the keyword dictionary using the search key extracted by the search key extraction unit are provided, and the keyword dictionary search unit matches the search key with the keyword. Then, the character strings before and after the keyword in the keyword dictionary corresponding to the keyword are compared with the character strings before and after the search key in the input character string, and if they match, the information in the keyword dictionary is extracted. By doing
Even if a character string including a word having a “Hiragana kanji” part is input, information from a desired keyword can be obtained without providing an exception dictionary and increasing the number of matching times.

In the first and second embodiments, the keyword dictionary 2 has been described by using the one in which the keyword and the word information storage unit are not separated as shown in FIG. 2, but the keyword dictionary 2 is shown in FIG. Alternatively, the keyword and the word information storage unit may be separated as shown in FIG. In this case, the keyword acts like an index.

Also, in the first and second embodiments, the method of searching the keyword dictionary for words mixed with kanji and kana as the input character string has been described. However, for words consisting only of hiragana, the keyword dictionary is defined by the hiragana character string. You will be searching.

EFFECTS OF THE INVENTION As described above, according to the present invention, a keyword dictionary containing character strings before and after a keyword and information corresponding to a keyword consisting of a character string of the same character type, and a search for the same character type character string from an input character string A search key extraction unit for extracting a key and a keyword dictionary search unit for searching for a keyword in the keyword dictionary using the search key extracted by the search key extraction unit are provided.
When the search key and the keyword match, the keyword dictionary search unit searches for the character string before and after the keyword in the keyword dictionary corresponding to the keyword and the character string before and after the search key in the input character string. By comparing and extracting the information in the keyword dictionary if they match, even if you enter a character string containing a word that has the part of "Hiragana kanji", the number of matching times is increased without providing an exception dictionary. It is possible to provide an information extraction device that can obtain information from a desired keyword without any need.

[Brief description of drawings]

FIG. 1 is a block diagram of an information extraction device in one embodiment of the present invention, FIG. 2 is a diagram showing a storage state of keywords and word information in the keyword dictionary 2 of FIG. 1, and FIG. 3 is a diagram of FIG. FIG. 4 is a diagram showing a keyword and a word information storage unit separated from each other, FIG. 4 is a diagram showing a process until optimum information is extracted from an input character string, and FIG. 5 is a configuration of a conventional information extraction device. FIG. 6 and FIG. 6 are views showing a part of the inside of the word dictionary 5 of FIG. 1 ... Search key extractor, 2 ... Keyword dictionary, 3 ...
Keyword dictionary search unit, 4 ... Word extraction unit, 5 ... Word dictionary, 6 ... Word dictionary search unit.

Claims (4)

[Claims] 1. A keyword dictionary containing character strings before and after a keyword and information corresponding to a keyword having the same character type character string, and a search key extraction for extracting a search key of the same character type character string from an input character string. And a keyword dictionary search unit that searches for a keyword in the keyword dictionary using the search key extracted by the search key extraction unit, and the keyword dictionary search unit matches the search key with the keyword. At this time, the character strings before and after the keyword included in the keyword dictionary corresponding to the keyword are compared with the character strings before and after the search key in the input character string, and if they match, the character string in the keyword dictionary is compared. An information extraction device, which extracts information of. 2. A keyword of a keyword dictionary searches for a character type other than hiragana from the beginning of the character string, and if there is a character of a character type other than hiragana, the same character type character string after that character is used as the keyword, and other than hiragana 2. The information extraction device according to claim 1, wherein the hiragana character string is used as a keyword if there is no character type. 3. The search key of the search key extraction unit searches for a character type other than hiragana from the beginning of the sentence, and if there is a character of a character type other than hiragana, the same character type character string after that character is used as the search key. The information extraction device according to claim 1, wherein if there is no character type other than hiragana, the hiragana sentence is used as a search key. 4. The information extraction device according to claim 1, wherein the character strings before and after the keyword in the keyword dictionary are a character string that comes before the keyword according to claim 2 and a character string that comes after the keyword.