JPH04213164A - Dictionary consulting system - Google Patents

Abstract

PURPOSE:To appropriately translate a word by translating the word from the context of the designated word, and checking the part of a text. CONSTITUTION:When there are an unknown word during a text input (edition), and the pertinent unknown word is desired to be translated, the word is indicated by an inputting device 10, and a dictionary consultation processor 8 is activated. Then, an analysis is executed for the range of the text including the word, and the appropriate meaning of the pertinent word is displayed. And also, a translated word is displayed after or under the designated word.

Description

[Detailed description of the invention]

0001

TECHNICAL FIELD The present invention relates to a dictionary lookup method, and more particularly to a dictionary lookup method in an editing device for language processing. For example, it is applied to word processors, translation machines, editor editing devices, user interfaces, and the like.

0002

[Prior Art] The meaning of a word in text that is being edited or input is searched in a dictionary without considering how the word is used in the text (sentence) or its context. For example, the items listed in JP-A-61-70 can be listed simply or in order of frequency of use.
No. 659 has been proposed. however,
In this case, if a word has many translations, the user must find the appropriate translation by himself from among the many translations, which is a burdensome method for the user.

[0003] Furthermore, the method described in Japanese Patent Application Laid-open No. 1-217565 is a method in which a sentence containing a specified word is translated as is, and the corresponding part is shown in Japanese, taking into account the context. When trying to produce a translation, it is difficult to analyze a sentence if there is a word in the sentence that has a specified word that is not registered in advance in the dictionary used for analysis. Or, it takes time to prepare a dictionary for analysis,
There was a problem in that it took a lot of processing time to analyze the text. Therefore, when translating an entire sentence,
All the words contained in a given sentence must be registered in advance, but since there are many words that are difficult to register in advance, such as proper nouns, there is a problem that the translation may not be successful. Was.

[0004] In addition, methods that perform translation involve not only syntactic analysis but also the processing of generating sentences corresponding to the target language, which requires extra processing time and the generation process. There is a problem that translation may fail due to this process because there are cases where the text does not match the generation method used for translation. On the other hand, since sentences that are actually required to be analyzed include many words that are difficult to register in advance, such as proper nouns, not all words included in the sentences are necessarily registered in advance. Therefore, a method that takes into account the influence of unknown words has become necessary.

[0005]

[Purpose] The present invention was made in view of the above-mentioned circumstances, and it is an object of the present invention to translate a specified word based on its context, and to properly translate the word by examining parts of the sentence. , was created with the purpose of providing a dictionary lookup method that lists translations that are considered more appropriate from parts other than the unregistered words, even when there are unregistered words before and after the specified word. .

[0006]

[Structure] In order to achieve the above objects, the present invention provides (1)
When entering or editing natural language text (sentence), the text that has already been entered is displayed on the screen, and the word (
a display/specifying means for specifying a range), a search means for searching a dictionary for words in a given natural language in the display/specifying means, and a syntactic relationship before and after the specified word in the search means. and (2) the instruction means lists the meanings of words suitable for the syntax even if there are unregistered words in other parts of the sentence. (3) When entering or editing natural language text, the already entered text is displayed on the screen and the word is specified. A display/designation means; a search means for searching a given natural language word in a dictionary; The present invention is characterized by having an instruction means for listing the meanings of specified words in such a way that they are best connected even if there are unregistered words in other parts of the text. Hereinafter, the present invention will be explained based on examples.

FIG. 1 is a block diagram for explaining an embodiment of the dictionary lookup method according to the present invention. In the figure, 1 is a word dictionary;
2 is an idiom dictionary, 3 is a parsing grammar, 4 is a work area, 5 is a parsing result registration data table, 6 is a display data storage area,
7 is an input data storage area, 8 is a dictionary lookup processing device, 9 is a display device, and 10 is an input device.

The word dictionary 1 lists the spellings, parts of speech, and translations of words, and the phrase dictionary 2 lists the spellings, parts of speech, and translations of words. The registered phrase string is shown in FIG. 4(a). Syntactic analysis grammar 3 lists rules constituting the language to be analyzed used for syntactic analysis. The work area 4 is a storage area used for dictionary lookup processing. Syntax analysis result registration data table 5 is stored in a storage area for storing syntax analysis results. An example of stored data is shown in FIG. 3(c). The display data storage area 6 is a storage area for storing data displayed on the display device, and in the present invention, it stores text data of sentences to be analyzed, data indicating specified word ranges, and results. Stores data for display. The input data storage area 7 is a storage area that temporarily stores data input from the input device 10. In the present invention, when a word to be analyzed in a sentence is specified using the input device, the specified word is is written to the input data storage area. Alternatively, the text to be analyzed is saved. Dictionary lookup processing device 8
has a function of executing the dictionary lookup process of the present invention. The display device 9 is a device for displaying text, and an example of the display on the display device is shown in FIG. 2(a) or FIG. 3(c). The input device 10 is a device for inputting text or specifying an unknown word (range).

When there is an unknown word during text input (editing) and you want to translate it, point at the word with the input device (pointing device) 10 and start the processing device of the present invention, and the text containing the word will be translated. performs analysis on a range of words and displays the appropriate meaning of the word. Or list the meanings of words that seem appropriate. In addition to displaying the word, it also has the function of displaying the translated word after or below (near) the specified word. During this analysis, even if there is an unknown registered word that is not registered in the dictionary in a sentence that includes the specified word, the syntax will be analyzed to the extent that it can be understood,
It has a function to list the meanings of words that have a high possibility of morphological connection based on the degree of connection before and after the words. It also has a means to efficiently check whether the specified words are constituent elements of an idiom.
When comparing registered phrases, it has a function of listing more appropriate meanings.

The dictionary lookup method of the present invention will be explained based on an example in which "like" is specified in a sentence as shown in FIG. 2(a). In this example, like has two possible parts of speech, the verb V and the preposition P, and we calculate which one is more appropriate, and select the parts of speech shown in Figure 3(c) in order of the most appropriate. List the meanings. here,
This article deals with the case where Neil is a word that is not registered in the dictionary, as is often the case with proper nouns.

The procedure of the dictionary lookup method according to the present invention will be explained below based on FIGS. 2(a) to 3(c). Step 1. Search the dictionary for the specified word. Data pairing a word's part-of-speech symbol and meaning is prepared (Figure 2(b)). If not, exit without doing anything. If the word in the registered dictionary has only one meaning, it is displayed and the process ends. The parts of speech used in the following explanations correspond to V: verb, P: preposition, N: noun, O: article, PP: prepositional phrase, VP: verb phrase, NP: noun phrase, and S: sentence. There is.

Step 2. A sentence containing the word is extracted from the sentence delimiter, and a table of word and number pairs is prepared by associating words with numbers (FIG. 2(c)). Step 3. Perform a registered dictionary search and prepare a table of part of speech and word number pairs for each word. At this time, if there is an unregistered word, mark it as such (Figure 3(a)). Step 4. A detailed explanation will be given later based on FIG. If it does not contain any unregistered words, the sentence is parsed as is. If this is not the case, syntax analysis is performed by extracting the largest continuous word in the range that includes the specified word and does not include unregistered words. In this example, the word Neil with word number 1 is an unregistered word, so the syntactic analysis proceeds with the remaining part including the designated word number 3 (FIG. 3(b)).

Step 5. The analysis results (FIG. 7) are examined, and the meanings of words are listed in order of appropriate meanings starting from the meaning of the part of speech that includes the longest number of words (FIG. 3(c)). This is based on the general tendency that the more words are matched, the more likely the context is (syntactically) grammatically correct and properly connected. Here, Figure 2
When checking the words with each part of speech symbol of the specified word in (b), and checking the application result item, mark the number with "?" among the attributes (part of speech symbol) and number of the application result. It extracts the length of the longest number string among the non-containing number strings, and displays the meaning of the part of speech with the larger value as a more appropriate meaning. In the example of the analysis results in Figure 7, “VP23
P becomes 4 from "45", V becomes 3 from "VP345", and P
The result is that the meaning of (preposition) is more appropriate. Also, instead of the above step 5, the following step 5'
(corresponding to claim 2). In other words, when examining the analysis results in step 5, the registered idioms (Figure 4(a))
is searched, and using the syntactic analysis results for phrases that include the specified word (Figure 4 (b)), if the part of speech matches, it is matched with the word, and if there is a match, it is , list its meaning at the top. Otherwise, execute step 5 above.

Next, look like in FIG. 4(a)
The dictionary lookup method of the present invention will be explained using the case as an example. From Figure 4(a), the part-of-speech patterns of the phrase are V, P, and P
is the third specified word, so it forms a pattern of V2, P3. On the other hand, from the analysis result example shown in FIG. 7, the application rule number is traced backwards to obtain the analysis development result (combination pattern of parts of speech) for possible parts of speech of the specified word. Among them, the part of speech pattern of the phrase extracted earlier V2, P
3 is found, and if they match, the words with the same word number of the idiom are compared with each other, and if they are the same, the meaning of the idiom is determined to be a more appropriate meaning. As a result of the syntactic analysis, only the part of the registered phrases that match the combination of parts of speech is compared, so the time required for matching can be reduced.

FIG. 5 is a flowchart of a parsing method for implementing the dictionary lookup method according to the present invention. Below, each step will be explained in order. Step 1: In the part of sequentially extracting the attribute (part of speech) symbol and word number of the word to be analyzed, if the analysis range is as shown in Figure 3(b), for example, the attributes of the word are extracted in the order as shown in Figure 6(a). (Part of speech) Pairs of symbols and word numbers are extracted. Here, word attributes (parts of speech) and word pairs in the order range shown in FIG. 3(b) are sequentially extracted from FIG. 3(a) and inputted in sequence. At this time, the table of word and number pairs is shown in Figure 2 (c).
The word number is determined based on , and the part-of-speech number and word number pair is determined by searching the table in FIG. 3(a). (Order does not affect results).

An example of the parsing dictionary used here is shown in Figure 6(b).
) applies to Verification of this rule (step 3 below)
), for example, as shown in FIG. 6(c), when the attribute (part of speech) of the word to be analyzed matches that shown in the item on the right side of the rule, that rule is applied. An example of the result of applying the rules is shown in FIG. 6(d). Here, the number listed to the right of the attribute symbol, or? The symbol indicates the number and position of the word in the sentence that constitutes the attribute (Figure 2(c)
) corresponds to the number in the table). In FIG. 6(d),
In the “V3” and “NP4?” parts of the application condition, word 3 is V.
, and when the fourth or subsequent words are NP, it indicates that 3 and the fourth and subsequent words constitute a VP.

The "?" symbol used here corresponds to whether a number with a difference of 1 from an adjacent number string in descending or ascending power order comes, or whether no number comes. for example,"
NP4? ” indicates that the word expressing the attribute “NP” is composed of the fourth word and the fifth or subsequent words, or only the fourth word. One attribute consists of a continuous word string, so one attribute (
When a symbol (part of speech) is composed of multiple consecutive attributes,
The number strings attached to the attributes must be consecutive in the order of their attributes. The number string corresponding to the number of the word attached to the attribute is attached so as to meet this condition. That is, when the attribute (part of speech) of the word to be analyzed matches all of the items on the right side of the rule, the corresponding word number is attached to each attribute as is.

On the other hand, when the attribute (part of speech) of the word to be analyzed matches some of the items on the right side of the rule, the corresponding word number is attached as is to the matching attribute symbol;
For the remaining parts, add the next number before or after the previous one to ensure continuity, and add a "?" symbol next to the previous one (on the left) or one after the previous one (on the right). The input example shown in FIG. 6(a) is analyzed and an example of the result registered in step 4, which will be described later, is shown in FIG.

Step 2: In the comparison with the previously registered analysis results to which the rule has been applied, a “?” mark is placed in the part after the application among the registered analysis results shown in FIG. 7. Among the uncertain analysis results, the attribute in the applicable conditions is compared with the number associated with it. "NP45" is "
NP4? ”, and the uncertain analysis result that was successfully matched is confirmed. For example, the condition part of the uncertainty analysis result is ``
V3” and “NP4?”, and the analysis result is “VP34?”
”, when the attribute of “NP4” is applied to this,
The condition parts are "V3" and "NP4", and the analysis result is "VP34", which is the final analysis result.

[0020] When a new definite result is obtained by comparing with the previously unconfirmed result in step 2 or as a result of applying the rules in step 3 (described later).
(inspect it in step 5 or step 7), input the pair of the attribute of the confirmed analysis result and its number, and step
Execute 2. For example, in the above example where "VP34" was obtained as the analysis result, "VP34" is newly input as s
tep2 is executed. When a new analysis result is registered, the part-of-speech information of the specified word is also copied as is. This is achieved by inputting the part-of-speech information of the specified word when the attributes of the word in step 1 in FIG. 5 are input in sequence.

[0021]Step 3; When comparing with the syntax analysis rules, if there is a result that is the same as a previously registered analysis result, the rule condition is considered to have failed in application. Further, when it is considered that as a result of applying the rule, a numerical value exceeding the maximum value of the number of words to be analyzed appears, the rule condition is considered to have failed in application. In addition, as a result of applying rules or as a result of comparing with previously registered analysis results, S(
sentence) may be extracted as an analysis result. At this time,
When it has the numerical strings of all the word numbers that make up the sentence, it is considered that the entire syntactic analysis of the sentence has been successful, but when it does not have even one word number that makes up the sentence, it is considered that the application has failed. For example, when a sentence consists of words 1 to 5 (consists of five words), even if an analysis result of "S2345" is obtained, the sentence does not include word number 1, so it is regarded as a failure.

Next, an example of a method for developing the rule result of step 5' described above will be described. Steps 4 to 7; To expand the rule results, use Figure 6 (
d) Alternatively, use the applicable rule number in the item of the table in Figure 7. The applied rule number is the number of the parsing rule used when the analysis result was registered in step 4, or the number of the registered analysis result, both of which are the number of the parsing rule used when the analysis result was registered in step 4.
It is written at the same time when the analysis results are registered. Among the applicable rule numbers, the numbers enclosed in parentheses ( ) represent the parsing rule numbers, and the numbers with nothing attached represent registered analysis results. For example, registration number 7 in FIG. 7 has applied rule numbers (1) and 6. This means that in order to obtain the analysis result "S?2345", you need to follow the rule (1) S->NP, VP of the parsing grammar in Figure 6(b) and the registered analysis result number 6 "VP345". ” is used.

Using this applied rule number, the rule result is developed as shown in FIG. 4(b). For example, “V
The analysis result "P345" has been obtained for registration number 6, which means that the applicable conditions are "V3" and "NP45".
It can be seen that it is. Furthermore, from the applied rule numbers, it can be seen that these results were obtained using the results of registration numbers 1 and 5, and the application of registration number 5, which has confirmed results that do not include "?" in the analysis results. By replacing the result (NP45) with the application conditions (D4, N5), the result shown in Fig. 4(b)
V3, D4, and N5 shown in the expanded example of the rule result are obtained.

FIG. 8 is a diagram showing another embodiment of the dictionary lookup method according to the present invention. In the figure, 11 is a word dictionary, 12 is a part-of-speech conjunction dictionary, 13 is a part-of-speech combination table, 14 is a work area, 1
5 is a dictionary lookup processing device, 16 is a display data storage area, 1
7 is an input data storage area, 18 is a display device, and 19 is an input device.

The word dictionary 11 lists the spellings, parts of speech, and translations of words, and the conjunction dictionary 12 shows the probability that the preceding part of speech and the following part of speech are combined in a general sentence.
An example of its contents is shown in FIG. 10(a). This is a statistical calculation of the possibility of combinations. The part-of-speech combination table 13 is a storage area that stores all combinations of parts of speech of a sentence. A specific example is shown in FIG. 9(e). The work area 14 is a storage area used for dictionary lookup processing. The dictionary lookup processing device 15 executes the dictionary lookup process of the present invention. The display data storage area 16 is a storage area that stores data displayed on the display device.

In the present invention, text data of a sentence to be analyzed, data indicating a specified word range, and data for displaying results are stored. The input data storage area 17 is a storage area that temporarily stores data input from an input device. In the present invention, when a word to be analyzed in a sentence is specified using an input device,
The specified word is written to the input data storage area. Alternatively, the text to be analyzed is saved. The display device 18 is a device that displays text. An example of display on the display device is shown in FIG. 9(a) or FIG. 10(d). Input device 1
9 is a device for inputting text or a device for specifying an unknown word (range).

When there is an unknown word during text input (editing) and you want to translate it, point at the word using the input device (pointing device) 19 and activate the processing device of the present invention, and the text containing the word will be translated. Perform parsing on a range and display the appropriate meaning of the word. Or list the meanings of words that seem appropriate. Alternatively, it has the function of not only displaying the word but also displaying the translated word after or below (near) the specified word. During this analysis, even if there is an unknown registered word that is not registered in the dictionary in a sentence that includes the specified word, the morphological connections are checked by checking the connections between the parts of speech before and after the word within the range that can be understood. It has a function to list the meanings of words with high probability as appropriate ones.

The dictionary lookup method of the present invention will be explained based on an example in which "like" is specified in a sentence as shown in FIG. 9(a). In this example, "like" has two possible parts of speech, the verb V and the preposition P. We calculate which is more appropriate, and select the parts of speech shown in Figure 10(d) in order of the most appropriate. List the meanings. Here, we will take up the case where Neil is a word that is not registered in the dictionary, as is often the case with proper nouns.

FIGS. 9(a)-(e) and FIGS. 10(a)-(
Based on d), the procedure of the dictionary lookup method according to the present invention will be explained below. Step 1. Search the dictionary for the specified word. Data pairing a word's part-of-speech symbol and meaning is prepared (Figure 2(b)). If not, exit without doing anything. If there is only one possible part of speech for a word in the registered dictionary, it is displayed and the process ends. Step 2. A sentence containing the word is extracted from the punctuation mark of the sentence, and a table of word and number pairs is prepared by associating the word with the number (FIG. 9(b)). Step 3. Perform a registered dictionary search and prepare a table of part of speech and word number pairs for each word. If there is an unregistered word at this time, mark it as such (FIG. 9(d)). Step 4. The analysis range is the longest word range that does not include unregistered words and that includes the specified word. In this example,
Since Neil is an unregistered word, the maximum length range that does not include Neil and includes the specified word like is targeted for analysis (FIG. 9(c)). Step 5. For the words in the analysis range obtained in step 4, possible combinations of parts of speech are saved in the part of speech combination table (Figure 9(e)).
).

Step 6. For each combination of parts of speech found in step 5,
The degree of possibility of the combination is calculated using the concatenated dictionary (FIG. 10(a)). The conjunction dictionary indicates the probability that the preceding part of speech and the following part of speech are combined in a general sentence, as shown in FIG. 10(a), for example. For example, the previous part of speech is V (verb)
The degree to which the following part of speech is also V (verb) is 9, and the degree to which the preceding part of speech is V (verb) and the following part of speech is P (preposition) is 694, and the latter is more important than the former. This shows that it is overwhelmingly more likely. For each combination in the part-of-speech combination table as shown in Figure 9(e), calculate the degree of possibility of conjunction for each two sets by shifting them one by one from the top, and by summing all of them, Find the degree of possibility. Figure 1 shows how the degree of possibility is calculated for the two possible combinations of parts of speech in this example.
0(b). For example, in the case of combination 1, V and V are 9, then V and N are 22, so the total is 31, while combination 2
In this case, V and P are 694, then P and N are 186, for a total of 880.

Step 7. Determines the display order of the parts of speech of the specified word. Compare the scores for the degree of possibility of the combinations obtained in step 6, examine the combinations of parts of speech with the highest scores, and list the meanings of the words by ordering the parts of speech of the specified word as more appropriate meanings. . In this example, combination 2 has a higher score, so P (preposition), which is the part of speech of "like" in specified word 3, is set first, and the next higher part of speech V (verb) is set second. At this time, if the part of speech has already appeared, ignore it and avoid duplication. Step 8. From the table of parts of speech and meanings of the specified word in FIG. 2(b), the corresponding meanings are listed in the order of the parts of speech determined in step 7, as shown in FIG. 10(d).

As described above, even if there is an unknown word in the sentence by examining the context of the parts of speech in the sentence, the combination of parts of speech that is most likely to form a sentence in a continuous range excluding unknown words is determined. It becomes possible to enumerate the meanings of the words. In this example, there are few combinations of verbs and verbs, but combinations of verbs and prepositions are possible, so the meanings of prepositions are listed as more likely meanings, as shown in FIG. 10(d). The meanings of the part-of-speech symbols used in this method are, for example, shown in Figure 10 (c
), each part of speech symbol corresponds to each part of speech.

[0033]

[Effects] As is clear from the above explanation, according to the present invention, a dictionary lookup candidate for a specified word can be obtained from the context of the word, and only a certain part of the sentence containing the specified word can be searched. Since words can be looked up in a dictionary, appropriate translation candidates can be obtained even if there are unknown words that are not registered in other parts of the dictionary. Furthermore, since the algorithm is simple, the above can be accomplished quickly (and efficiently). There is no need to prepare a special dictionary for analysis, and it can be easily realized.

[Brief explanation of the drawing]

FIG. 1 is a configuration diagram for explaining an embodiment of a dictionary lookup method according to the present invention.

FIG. 2 is a diagram for explaining an example when "like" is specified in an analyzed text.

FIG. 3 is a continuation of FIG. 2.

[Figure 4] Examples of idioms in analyzed sentences: look like
FIG. 3 is a diagram for explaining an example when specifying .

FIG. 5 is a diagram showing a flowchart of a parsing method for implementing a dictionary lookup method according to the present invention.

FIG. 6 is a diagram for explaining syntactic analysis based on FIG. 5;

FIG. 7 is a continuation of FIG. 6.

FIG. 8 is a diagram showing another embodiment of the dictionary lookup method according to the present invention.

FIG. 9 is a diagram for explaining another example when "like" is specified in an analyzed text.

FIG. 10 is a continuation of FIG. 9.

[Explanation of symbols]

1...word dictionary, 2...idiom dictionary, 3...parsing grammar, 4...
Work area, 5... Syntactic analysis result registration data table, 6... Display data storage area, 7... Input data storage area, 8... Dictionary lookup processing device, 9... Display device, 10... Input device.

Claims (3)

[Claims] Claim 1: Display/designation means for displaying text already input on a screen when inputting or editing natural language text, and for specifying words; A search method that can search for words in a dictionary, and a search method that examines the syntactic relationships before and after a specified word to find the meaning of a word that is syntactically appropriate even if there are unregistered words in other parts of the sentence. A dictionary lookup method characterized by having an instruction means for enumerating. 2. The dictionary lookup method according to claim 1, wherein the instruction means searches for a syntactically appropriate phrase from among the registered phrases and lists the meanings of the designated word. 3. Display/designation means for displaying text already input on the screen when inputting or editing natural language text, and specifying a word; A search means that can search for words in a dictionary, and a search means that checks the degree of connection between the parts of speech before and after the specified word, so that even if there are unregistered words in other parts of the sentence, the degree of connection is the best. A dictionary lookup method is characterized in that it has an instruction means for listing the meanings of specified words.