JPH08305695A - Document processing device - Google Patents

Abstract

PURPOSE: To improve the efficiency of dictionary retrieval by retrieving a fundamental dictionary by a dictionary retrieval part with respect to a word candidate obtained by a word discrimination part and referring to a priority table to retrieve technical term dictionaries in the case of the absence of this word candidate in the fundamental dictionary and retrieving technical term dictionaries in the priority order. CONSTITUTION: A document or an indication input 1 is inputted by an input processing part 2. A word discrimination part 3 divides the inputted document into words in the unit of sentences in accordance with a prescribed reference and transfers the result to a dictionary retrieval part 6. With respect to the word candidate transferred from the word discrimination part 3, the dictionary retrieval part 6 retrieves a dictionary group, namely, a fundamental dictionary 8a and technical term dictionaries 1 to N (8b to 8n) to obtain attribute data or the like. The order of dictionaries to be retrieved is determined by referring to a priority table 5a, a related dictionary table 5b, and a retrieval record 5c. Retrieval result data such as the attribute of the word detected as the result of retrieval is processed in a syntax/meaning analysis part 4 to discriminate whether it is valid or not. If it is valid, it is transferred to a processing part for the succeeding processing.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a document processing apparatus having a plurality of technical term dictionaries in addition to a basic dictionary and processing documents relating to a wide range of fields.

[0002]

2. Description of the Related Art In a device for processing a document such as a word processor, machine translation, index generation, etc., when a dictionary is searched to obtain word information, a basic dictionary is searched first, and if it is not there, a technical term dictionary is found. To search. FIG. 4 shows an example of dictionary search. (1) is an example in which a sentence is divided into words. (2)
Is an example of the contents of the dictionary. The basic dictionary holds words (particles, adverbs, verbs, inflectional endings, etc.) that make up a general sentence. Multiple specialized term dictionaries (words such as nouns, special verbs, etc.) that have meanings that differ from one specialized field to another or are often different depending on the specialized field (general language dictionary contents should be treated as one of the specialized term dictionaries). It may be). The sentence is divided to obtain word candidates, a dictionary is searched for each word candidate, and the obtained attributes, meanings, etc. are passed to the syntax / semantic analysis unit etc. as original information for processing such as translation.

Conventionally, the technical term dictionary to be used is determined by searching the technical term dictionary in order until a word is found, or by specifying the technical field of the document to be processed in advance by the user. It was

[0004]

Therefore, if the user does not specify the field, it is not possible to select an appropriate dictionary for the document. Therefore, even a technical term dictionary unrelated to the target document can be selected. You will be searching. Also,
When processing documents that have content that spans different fields, or when processing multiple documents that belong to different fields in succession, you need to specify an extra dictionary as the search target.
The search efficiency was low.

It is an object of the present invention to improve the efficiency of search by automatically and preferentially selecting and searching a technical term dictionary suitable for a technical term in a document.

[0006]

[Means for Solving the Problems]

Claim 1: In a document processing apparatus comprising a basic dictionary for basic words and a technical term dictionary group for words in a specialized field, a word determination unit for dividing a sentence of a document to be processed into word candidates and separating the words. A priority table that lists the search priority of the dictionary, a dictionary search unit that sequentially searches each dictionary according to the priority order shown in the priority table for the separated word candidates, and each dictionary of the priority table And a table updating unit that updates the priority of the search according to the detection status from the dictionary.

According to a second aspect of the present invention, in the document processing apparatus according to the first aspect, the priority table has an item for recording the number of word detections for each dictionary, and the table updating unit detects the words in the dictionary that can be searched. The priority table is updated based on the number of times, and the priority is updated in descending order of the number of times.

A third aspect of the present invention is the document processing apparatus according to the first or second aspect, which has a related dictionary table in which the degree of association between the dictionaries is recorded, and the table updating unit updates the priority table. Sometimes, referring to the related dictionary table, the priority of the related dictionary is updated.

A fourth aspect of the present invention is the document processing apparatus according to any one of the first to third aspects, in which information indicating a word name as a search target and a dictionary in which the word name is detected as a result of the search and information at that time point are displayed. It has a search record holding information indicating the priority of each dictionary, the table update unit updates the search record when updating the priority table, and the dictionary search unit re-searches from the syntax / semantic analysis unit etc. When the instruction is given, the search record is referred to, and if the name of the word candidate to be searched exists in the search record, the dictionary lower than the priority of the dictionary in which the word candidate was detected last time is searched.

[0010]

The word determining unit cuts out word candidates from the input document and sends them to the dictionary searching unit. The dictionary search unit first searches the basic dictionary for the word candidate obtained by the word determination unit, and if found, returns the dictionary contents (search result) to the word determination unit.

If it is not in the basic dictionary, the technical term dictionary is searched. At this time, referring to the priority table, the technical term dictionary with the highest priority is searched in order. When the word is detected, the dictionary name is sent to the table updating unit.

The table updating unit updates the priority table by a predetermined algorithm so that the searched dictionary has a higher priority. With this configuration, when searching for the next word candidate, the dictionary suitable for the field to which the document that contains it is prioritized to be searched, and the efficiency of the search can be improved. .

Claim 2: When the priority table has an item for recording the number of times of word detection for each dictionary, the table updating unit updates the number of times of word detection of the retrieved dictionary in the priority table and By updating so that the priority increases in descending order of the number of times, the priority of the frequently used dictionary becomes higher. Therefore, the search efficiency is increased.

Claim 3: When updating the priority of a dictionary, the ranks of related dictionaries are processed at the same time.
The search efficiency at the time of the next search is improved. Claim 4: In the case of re-searching, by referring to the search record, the dictionary that was unsuitable for the previous search is removed and the search is performed, so the search efficiency is increased.

[0015]

Embodiments of the present invention will be described below with reference to the drawings. 1 is a block diagram of an embodiment of the present invention.
It is assumed that a machine translation device to which the inventions 4 to 4 are applied.

The document or instruction input 1 is input by the input processing unit 2. The word determination unit 3 divides the input document into words on a sentence-by-sentence basis according to a predetermined criterion, and the dictionary search unit 6
Pass to. The dictionary search unit 6 uses the word candidates passed from the word determination unit 3 as a dictionary group, that is, the basic dictionary 8a and the technical term dictionaries 1 to 1.
The technical term dictionary N (8b to 8n) is searched to obtain attribute data and the like. The order of dictionaries to be searched is the priority table 5a, related dictionary table
Determined by referring to 5b and search record 5c.

As a result of the search, the search result data such as the attributes of the detected words are processed by the syntax / semantic analysis unit 4,
It is judged whether it is appropriate. If appropriate, it is passed to a processing unit (not shown) that performs subsequent translation processing and the like.
If it is determined to be inappropriate, the dictionary search unit 6 is instructed to search again.

The operation of the dictionary search unit 6 will be described below. Example 1: When only the invention of claim 1 is applied = only the priority table 5a is provided, and the priority table 5a holds only the priority order of each dictionary: word candidates are searched from the technical term dictionaries 8b to 8n In order to do so, the priority table 5a is referred to and the dictionary ranked first is searched. If the first-ranked dictionary is not found, the second-ranked dictionary is searched. Hereinafter, the dictionary is searched in the order of the priority table until the word is found.

When it is possible to detect it from the first-ranked dictionary, the priority of the priority table 5a is not updated. When the search is performed from the second rank or lower, the detected dictionary is set to the first rank, and the rank of the dictionary positioned higher than the rank of the detected dictionary is lowered one by one.

FIG. 2A shows an example of updating the priority table. 5
It is assumed that there are two technical term dictionaries a, b, c, d, and e, and the priorities are a, b, c, d, and e in this order. Search is
The processing is executed in order of priority from the dictionary a having the first priority. If a word can be detected in the dictionary c, the dictionary c is ranked first and the ranks below the dictionary a are lowered. When searching for the next word candidate, it is executed from the dictionary c. If it is not found in the dictionary c and can be detected in the dictionary b, then the dictionary b is ranked first and the ranks below the dictionary c are lowered.

Example 2: Application of the invention of claim 2 = Priority table 5a holds the rank and the number of times a word is detected (the number of times the dictionary is used): As in Example 1, the priority table 5a has the same priority. Then, search the dictionary in order. The priority table 5a records the dictionary name and the number of times the word is found by searching the dictionary.

FIG. 2B shows an example of updating the priority table 5a. In the initial state, the number of times is all 0. If it can be detected from the dictionary a, the number of times of the dictionary a is set to 1. If it can be detected from the dictionary c in the search of the next word, the number of times of the dictionary c becomes 1, and it ranks in the upper rank of the dictionary b of 0 times and the second rank of the dictionary a of 1 times. Further, in the next word candidate search, when it is detected again from the dictionary c, the number of times of the dictionary c becomes 2, and since it exceeds the number of times of the dictionary a, it is moved to the first place. That is, it is placed higher in the number of times found less and lower in the number of times found the same.

Example 3: Application of the invention of claim 3 = Related dictionary table 5b
When: The related dictionary table 5b holds two related dictionary names as a set. Although the content may be fixedly decided, for example, at a predetermined time (eg, after processing one document,
After retrieving a predetermined number of words), the table updating unit 7 determines that the priority table 5a
If the ratio of the number of searches of the top two dictionaries to the total number of word searches reaches a predetermined level (eg, 30%), it is determined that the two dictionaries are used in connection with a certain field. Then, by performing a process of registering the two dictionary names as a pair in the related dictionary table 5b, a better relationship is set.

As in Example 2, 1 is added to the number of dictionaries that can be detected by searching for word candidates, the number of dictionaries with a higher priority is checked, and the order is changed. At this time, referring to the related dictionary table 5b, it is checked whether a set including the dictionary name is registered. If so, the other dictionary name in the set is the priority table.
Find out which rank of 5a. If it is higher than the dictionary whose order has been changed, leave it as it is. If it is in the lower rank, the rank is changed according to the number of times as in Example 2.

FIG. 2C shows an example of the priority table 5a and the related dictionary table 5b. The related dictionary table 5b includes a set of dictionaries (a, b).
It is assumed that (a, c) and (c, d) are recorded. Here, when a certain word is detected from the dictionary c, the number of times of the dictionary c becomes 1, and it is ranked first. Since it can be seen from the related dictionary table 5b that the dictionary d is related to the dictionary c, the dictionary d is moved up to the second position, and the dictionaries a and b are moved down respectively.

Next, when another word candidate is detected from the dictionary a, the dictionary a is moved up to the second position. It can be seen from the related dictionary table 5b that the dictionary b and the dictionary c are related, but the dictionary c
Is higher than the dictionary a, the carry-up process is performed only for the dictionary b.

Next, when it is further detected from the dictionary a, the dictionary a becomes the number of times 2, so it is moved up to the first place. Dictionary b and dictionary c
Is related, but since it is already in 3rd and 2nd place, the result remains the same.

Example 4: Corresponding to the invention of claim 4: Search record 5c
For each certain processing unit (one sentence or a certain number of words),
Keep a record of your searches. The recorded contents are the word name searched, the rank of the dictionary that could be detected, and the priority table when searching 5a
It is a copy of the ranking (before updating).

In a normal search, similar to Examples 1 to 3, the search is performed according to the priority table 5a, and a copy of the rank of the priority table 5a before updating is detected for each detection and the rank of the dictionary that can be searched. It is also saved in the search record 5c.

For example, in the case of machine translation, the search result is syntax /
When passed to the semantic analysis unit, the syntax / semantic analysis unit and other processing units perform processing and output a translated sentence as a final result. If the syntax / semantic analysis unit judges or the operator who saw the display of the final result gives an inappropriate result, a re-search is required. When the re-search is instructed, the search record 5c is checked to see if the result of searching the same word candidate remains.
When the name of the word candidate does not exist in the search record 5c (when the word is cut out differently, etc.), the same processing as the normal search is performed based on the current priority table. If the name of the word candidate exists in the search record 5c, the priority of the previous dictionary remaining in the record is referred to, and the dictionary lower than the previously searched order is searched.

When it is detected, the rank of the recorded dictionary is rewritten to the rank of the dictionary that can be searched again. Also, for the priority table 5a, the number of times the dictionary searched last time is reduced,
Increase the number of re-searched dictionaries and change the ranking if necessary.

When the processing of the predetermined unit is completed, the recording is discarded and the recording of the next processing unit is started. FIG. 3 is an example of updating the priority table 5a and the search record 5c. (A) shows the priority table 5a and the search record 5c after the word W1 was detected as a result of the search of the dictionary (c in this case) ranked second in the word W1 at a certain time. Search record 5c is the word name of the search target, the rank of the dictionary in which it was detected, and the priority table before update
Keep a copy of the ranks of the dictionary in 5a. (B) shows that the word W2 is detected next from the third dictionary (b in this case) in the priority table and added to the search record 5c. At the same time, the number of times the dictionary b in the priority table 5a is updated to 1.
In (C), it is added to the search record 5c that the word W3 has been detected from the second dictionary (c). Since the dictionary c has been used three times, the rank of the priority table 5a is updated to the first rank.

When a re-search instruction is given, the search is performed according to the priority order recorded in the search record 5c instead of the priority table. If the word W1 is instructed to be re-searched, it was detected from the second-ranked dictionary c (but was inappropriate) last time, so the dictionaries b, d, and e that are third-ranked or lower are searched in order. . If it can be detected from the dictionary d, the rank 2 left in the search record 5c is rewritten to the rank 4 of the dictionary d. Also, the priority table 5a reduces the number of times of the dictionary c by 1,
The number of times of the dictionary d is increased by 1, and the rank is changed if necessary.
As a result of the above re-search, (D) shows the priority table 5a and the search record 5c.
Shows the updated state.

In FIG. 1, the word determination unit cuts out words from the input document, but in the case of re-search, the operator (user) directly specifies each word,
It may be passed to the dictionary search unit 6.

[0035]

As described above, according to the present invention,
Even if you do not specify the technical term dictionary used by the user,
The technical term dictionary suitable for the field to which the document belongs is preferentially searched, and the efficiency of the dictionary search can be improved. It should be noted that even the word that was searched for for the first time is likely to belong to the same field as the previously searched word, so it will be searched from the dictionary of the appropriate field (dictionary with high priority at that time). , High search efficiency.

Further, by estimating the relevance between the specialized dictionaries from the related dictionary table and updating the priority of the dictionaries, the priority of the dictionaries which are considered to be related is also increased at the same time.
It is possible to improve the search efficiency even when processing a document that spans a plurality of fields, or when processing documents in different fields in succession.

When a search is performed again because the search result is unsuitable, by using the search record, it is possible to automatically search from another dictionary and improve the search efficiency.

[Brief description of drawings]

FIG. 1 is a configuration diagram of an embodiment.

[Fig. 2] Example of updating the priority table

[Fig. 3] Example of updating the priority table and search record

[Figure 4] Example of dictionary search

[Explanation of symbols]

 1 input document / instruction input 2 input processing unit 3 word determination unit 4 syntax / semantic analysis unit 5a priority table, 5b related dictionary table, 5c search record 6 dictionary search unit 7 table update unit 8 basic dictionary 8a, 8b-8n specialty Term dictionary

Claims (4)

[Claims] 1. A document processing apparatus comprising a basic dictionary for basic words and a technical term dictionary group for words in a specialized field, wherein a word determination unit divides a sentence of a document to be processed into word candidates and separates them. , A priority table that describes the search priority of each dictionary, a dictionary search unit that sequentially searches each dictionary according to the priority order shown in the priority table for the separated word candidates, and a priority table And a table updating unit that updates the search priority of each dictionary according to the detection status from the dictionary. 2. The document processing device according to claim 1, wherein the priority table has an item for recording the number of times of word detection for each dictionary, and the table updating unit sets the number of times of word detection of the dictionaries that can be searched. A document processing apparatus, characterized in that the priority table is updated based on the above, and the priority is updated in descending order of the number of times. 3. The document processing apparatus according to claim 1 or 2, further comprising a related dictionary table that records a degree of association between the dictionaries, and the table updating unit updates the priority table when updating the priority table. A document processing apparatus characterized by updating a priority of a related dictionary by referring to a related dictionary table. 4. The document processing apparatus according to any one of claims 1 to 3, wherein a word name as a search target, information indicating a dictionary in which the word name is detected as a result of the search, and each dictionary at that time are included. The table update unit updates the search record when updating the priority table, and the dictionary search unit instructs the syntax / semantic analysis unit to perform a re-search. When the name of the word candidate to be searched is present in the search record, the search is performed from a dictionary lower than the priority of the dictionary that detected the word candidate last time. Document processing device.