JPS6395569A - Language analyzing device - Google Patents

Abstract

PURPOSE:To realize a proper morpheme analysis in terms of the expression including the numerical value by including the discrimination display showing that the dictionary consulting unit shows numbers in the form of the dictionary data for the dictionary consulting unit showing numbers. CONSTITUTION:A dictionary retrieving part 104 retrieves a word dictionary 18 based on the retrieving key character string inputted from a unit segmenting part 102 and extracts the dictionary information to send this information to processing parts 110, 1120, 114 and 118 respectively. The part 110 arranges the numerals into numerical value and the part 114 processes the numerals connected by hyphens. Then a processing part 116 processes the continuous numeric characters and the part 118 arranges the currency symbols and the numerical value into a single noun. A processing part 120 arranges the numerical value and the units into a single noun. Then a processing part 122 arranges the immediately preceding numerical values after conversion of the numerals into the numerical value, the processing of numerals containing hyphens and the continuous processing of numerals.

Description

DETAILED DESCRIPTION OF THE INVENTION TECHNICAL FIELD The present invention relates to a language analysis device, and more particularly to a language analysis device for analyzing natural language, which is useful, for example, in automatic translation devices.

BACKGROUND ART For example, when creating a corresponding Japanese sentence from a sentence in a foreign language such as English, the morphemes of the input English sentence are analyzed, its syntax is analyzed, and the sentence structure is copied. Generate translations of words.

When analyzing the morphemes of a sentence, the number expressions in Arubo5Δ may not necessarily correspond one-to-one with the number expressions in other 100 words. For example, in European languages such as English Jh, the basic idea of counting, that is, place value, is different from Japanese. Therefore, if words representing numbers are made to correspond one-to-one between English and Japanese, appropriate translation may not be achieved. For example, "ten" in English is "ten" in Japanese j/l, and "thousand" in English is "thousand" in Japanese. If such a simple correspondence method is adopted, for example, "ten thousand" will be translated as "ff". In other words, a shift in scale occurs. In order to convert this ``Jucho'' into the word ``10,000,'' which indicates the scale, the conventional method includes a table that shows the correspondence between the two.
I referred to this point by point. If such scale conversion data were provided as a correspondence table for all scales, the overall amount of data held by the system would become too large.

For example, the English numerical expression ``a hundred and
tw.

thousand and two hundred and
In a simple system that simply decomposes ``four'' into its component parts and replaces them with their corresponding numerical expressions in Japanese, this would be parsed as [Hyaku to 2 cho - 200 and 4]. It's just 0. Originally, this was finally translated into Japanese as well.
It should be interpreted as 102,204 J, or "102,204 F204."

For example, in English, “$1.5 m1llio”
The expression "n" is also often used, but in order to properly translate the J!
ion" with the unit symbol "S" added to it. However, in conventional systems that perform simple analysis such as using spaces or blank characters as padding breaks, this is not possible. ”1.5 dolla
If it consists of 2il(7) elements of "rs" and "+5illion'", it will be incorrectly analyzed.

OBJECTS In view of these demands, an object of the present invention is to provide a language analysis device that can perform appropriate morphological analysis on expressions that include numerical values.

In order to achieve the above object, the composition tree invention includes a dictionary means storing dictionary data for each dictionary lookup unit, divides an input sentence into dictionary lookup units, and refers to the dictionary means for the dictionary lookup unit. In a language analysis device having an analysis means for performing morphological analysis, the dictionary means includes, as dictionary data, an identification display indicating that the dictionary lookup unit represents a number for a dictionary lookup unit representing a number; The dictionary means is referred to for each dictionary lookup unit included in the sentence, and when the dictionary data retrieved includes an identification display, the dictionary lookup unit for which this identification display is retrieved is searched for in that dictionary! ' Combine with the dictionary lookup unit that is near the nth position and for which other identification indications have been found, calculate the meanings of both dictionary lookup units together to obtain the value of ton, and combine both dictionary lookup units with It is characterized by having the unit of analysis as the unit of analysis.

Hereinafter, a detailed explanation will be given based on one embodiment of the wooden invention.

Referring to FIG. 2, there is shown the overall configuration of an embodiment in which the language analysis device according to the present invention is applied to an English-Japanese visual translation device. The present invention is effective not only for an English-Japanese automatic translation device that translates English into Japanese, but also for any language analysis device that analyzes sentences in an input language when translating one language into another language. Needless to say, it is applicable.

This embodiment has an input section 1O, through which an English text 12 to be translated into Japanese is input. The input unit IO includes, for example, a keypad F having character keys such as alphanumeric keys and function keys, an optical character reader (OCR) that reads English text recorded on paper, and/or
It may include a file storage device that reads English text recorded on a storage medium such as a magnetic disk.

The English text input by the input unit 1O is read into the pre-editing unit 14 and pre-processed for translation. Here, as the king, he recognizes sentences and processes unknown words. This is used as part of morphological analysis.

The pre-edited English text data is transferred to the morphological analysis unit 16 together with the information obtained in the previous four collections. Morphological analysis section 16
Now, we will index the trL language dictionary, divide it into sentences, analyze the morphemes of the English sentence, process unknown words, express proper nouns at 1 o'clock,
We provide various summaries such as number expressions, additional questions,
Morphological analysis rules for processing the entire sentence, such as recognition of apposition, are stored in the analysis rule file 38.

The English sentence data subjected to the morphological analysis is transferred to the parser 1 section 20 together with the dictionary information obtained through the morphological analysis. m
The sentence analysis section I 20 is a functional section that applies grammar rules to English data to analyze the surface structure of a sentence, and finds all syntactic possibilities.

The English text data parsed by the parser 1 section 20 is sent to the parser II section 22 together with the analysis information. Here, a solution is selected by applying a structural description from the superficial syntactic analysis result of syntactic analysis (■). This creates a reliable parse tree for the English sentence and creates its structure. These parsing rules are also stored in the parsing rule file 3B.

The parsed English data is transferred to the structure conversion unit 24 as parse tree data. The structure conversion unit 24 creates a syntax tree for a corresponding Japanese sentence from a syntax tree that is an intermediate structure of an English sentence, and converts the Japanese sentence into a basic Japanese structure that is easy to translate.

The syntax tree data indicating the basic structure of the Japanese language whose structure has been converted in this way is sent to the translation generation unit 2B, and the latter generates a translation. This is an a function section that generates Japanese sentences from the tree structure of Japanese syntax trees.

The translated Japanese text data, that is, the translated text data, is sent to the post-editing section 30. The post-editing unit 30 uses the information used in the translation process to index the dictionary 18 and correct the translated data to complete a more natural Japanese sentence. This Japanese sentence data is transferred to the output unit 32 and outputted from the output unit 32 as a translated Japanese sentence 34. Output section 32
includes, for example, a printer, a display, and/or a file storage device such as a magnetic disk.

The flow of a series of these translation processes is controlled by the control unit 3B that controls the entire water device.

In this embodiment, the 41-word dictionary 18 stores dictionary data for English, 7M, and Japanese words, and includes not only narration, but also dependency relationships, that is, co-occurrence relationships, meanings, singularity, plurality, parts of speech, etc. Information is written. The analysis rule file 36 also stores rule data for morphological analysis and syntactic analysis.

An operation display section 40 is connected to the control section 38 . The operation display section 40 is used to input various instructions from the operator to the water device, such as operation keys such as translation instruction keys and cursor keys, intermediate data such as input English text, translated Japanese text, dictionary information, etc. It has a display and indicators that visually display various instructions to the operator.

Note that many of these operation display functions may be configured to be included in the keyboard if the single-person power unit 10 is equipped with a keyboard, or in the display if the output unit 32 is equipped with a display.

Referring to FIG. 1, a detailed configuration regarding number processing of the morphological analysis unit 16 is illustrated. Shape y5 elementary analysis section 16
Of course, it also has other analysis functional units, but only those that are directly relevant to understanding the present invention are shown here. In the morphological analysis, the dictionary search is instructed in accordance with the character string of the search key in order from the beginning of the human-powered character string, and the dictionary search unit 10
This is done by executing processing on the dictionary information obtained from step 4 according to numerical flags, which will be described later.

Morphological analysis department! Reference numeral 6 denotes a human power processing unit 1 for receiving input character string data input from the preprocessing unit 14 and performing input processing.
00. The input processing unit 100 includes, for example, an ASC
English character string data is input in the form of code data such as II, and an input character string buffer is provided for temporarily storing the character string data.

A unit extraction unit 1 that extracts input character string data temporarily stored in the input processing unit 100 into dictionary lookup units such as pm, ;h, etc.
Sent to 02. The unit cutting section 102 is.

When the dictionary search section 104 later searches the dictionary for one day, this section is used to identify the dictionary lookup unit that constitutes the search character string. Dictionary lookup delimiters used in the dictionary lookup unit extraction process are alphanumeric characters.

Characters other than apostrophes, hyphens and periods,
and at the apostrophe following a whitespace character. This is stored in the territory table 10B, and is referred to by the unit extraction unit 102 when extracting the R search unit.

! T! ,t/! The dictionary 18 stores information particularly for searching for cut-out units. For example, as shown in Figure 8, an example of the entry information shows that for each dictionary lookup unit, for example, a word entry, in addition to grammatical information such as the part of speech, for words that express numbers, there is an identification that indicates that the word expresses a number. A display, that is, a numerical flag, and numerical information indicating the numerical value are stored.

As illustrated in the figure, each entry in the word dictionary 18 is written in both singular and plural forms, and each constitutes one entry. The numeric flag is a flag that, when "1" is displayed, indicates that the word is +iA, which means a number. Other information registered includes, for example, whether a noun is countable or uncountable, whether it is an intransitive verb or a transitive verb, and its translation. For example, "thousand" is a noun that indicates a number, so its numeric flag is "1" and the numerical value is rloooO'-c. Also “thread”
” is a noun, but it is not a noun that indicates a number, that is, it is not a number word, so the number flag is registered as “0”.

In the case of a word registered in the dictionary 18, such as one "thousand", the number is recognized using its numerical flag. Even if it is an unregistered word, for example r123
Number series such as J, r 2 like decimals such as 10.24
those with a period between the numbers in the set, and rl,
A number that includes a comma between numbers, such as 000,000 J, is also recognized as a number. In addition, in this specification, the term "number" usually refers not only to simply an arithmetic number but also to "thi
It also includes number expressions spelled out, such as "rteen".

As shown in FIG. 12, the dictionary 18 includes a currency symbol table 18a in which various currency symbols are registered.
, "," (space), etc., and a decimal point table 18c, in which decimal points ",", ", etc. are registered. The reason why we have tables for place value symbols and decimal points is that, as is well known, Japanese and English use "," for place value symbols and "," for decimal points, whereas French and German languages use "," for place value symbols and "," for decimal points. Other European languages, such as , mainly use a space or "," for the place mark and "," for the decimal point, etc.
This is because the usage of symbols differs depending on the target language.

The dictionary search unit 104 searches the word dictionary 18 to extract dictionary information based on the sample key character number 1 inputted from the unit extraction unit 102, and sends the dictionary information to the processing units 110, 112, 1.
14 and 11B.

Combining numbers is performed by the following two processes. First, when it is recognized as a number as described above, the next jido-hiki unit is looked at and if it is also recognized as a number, these are combined into one number. Repeat this operation as long as the number continues, for example "30 thousand" is r 300
00 J, "1.5 m1llion" is r150
When the 0th order, which is 0000 J, is followed by further number expressions with "and" in between, in terms of the meaning of those number expressions, "a"
When the digits on the left side of “and” corresponding to each digit of the numerical value pointed to by the pointer on the right side of “and” are all 0”,
Combine into one number. For example, “one hundred”
d and thirty” is r130J, and “
30 thousand and two hundred
ed' becomes r30200 J.

After determining such a number, further necessary local analysis is performed. This combines the form TrF: of each analysis unit based on local analysis rules, and the consecutive analysis units activated from elementary activation information into one analysis unit. For example, the currency symbol and number ``¥1,000'' are combined into ``r 1000 yen,'' and the number and unit ``165 layers'' are combined into ``1.5 kilometers.''

These grouping processes are performed by processing units 110 to 122. The processing unit 110 is a functional unit that performs a process of grouping numerals into currency symbols or units. Processing section 11
2 is a functional unit that performs a process of converting number words into numerical values. Further, the processing unit 114 is a functional unit that processes number words connected with a hyphen. Further, the processing section 1113 is an a-plane section that processes consecutive numbers.

When a numerical word is combined with a currency symbol or a unit, the processing unit 11B combines the currency symbol and numerical value into a single noun.

In addition, in the case of combining the numerical value and the unit, the processing unit 120 combines the numerical value and the unit into a single noun. Furthermore, the processing unit 122 performs a process of combining numerical words that have been subjected to numerical processing, hyphenated number processing, and continuous number word processing with the numerical value immediately before them. The dictionary information of the input character string that has undergone these processes is stored in the verified dictionary information buffer, that is, the dictionary information storage table 124.

The morphological analysis results are stored in the dictionary information storage table 124.
The data is then transferred to the syntax analysis section 1 20.

Processing using numerical flags is performed in the sequence shown in FIGS. 3A and 3B. The input processing unit 100 receives input character string data and performs input processing (200)
. Therefore, the unit extraction unit 102 extracts the input character string into units of extraction in order to index the dictionary 18 (201).
Then, the dictionary search unit 104 searches the dictionary 18i according to this (203), and if there is a dictionary entry (204).

The numeric flag is checked (205). If the numeric flag is not set, this is other than the number JI4, so the dictionary information is stored in the dictionary information storage table 124. If rlJ is set in the number flag, the processing unit 112 converts the number words into a-values (206), and the processing unit 122 performs processing 207 to combine them with the immediately preceding numerical value. When these processes are performed up to the final position of the sentence indicated by the input character string data (20
2) The processing units 11B and 120 perform a process 209 to combine the currency symbols or units, and output the morphological analysis results to the syntactic analysis unit 1 20 (210)
.

As a result of the dictionary lookup, if the entry does not exist in step 204, and if the element has a hyphen (212), the processing unit 114 performs hyphenated number processing 213.

If the first currency symbol is not a hyphen (214)
, save only the currency symbol in the dictionary information storage table 124 (218), delete the currency symbol from the dictionary lookup unit
(217). If the first symbol is not a currency symbol (214), the processing unit 11B performs numeric sequence processing 215, which is executed up to the final position (202).

The process 209 of combining currency symbols and units is performed in the fourth
The processing is performed by the processing unit 110 according to the processing flow shown in the figure. First, in initial processing 220, a processing start pointer is initially set to the beginning of the buffer. If the element pointed to by the pointer is not the number 11 (221), the pointer is incremented (226).Even if it is a number, if the immediately before it is not a currency symbol and the immediately after it is not a unit, the pointer is still is incremented (222, 224), and this is done until the final position of the dictionary lookup unit (227).

If it is a numerical value (222), the currency symbol and the numerical value are combined into one noun (223). For example, the currency symbol and the number "¥1,000" are treated as one noun. Also,
When the immediately preceding symbol is not a common IY symbol and the immediately following is an honor, the number and unit are combined into one noun (225). For example, the number and the unit "1.5 part m" are one = Use it as a noun. Do this until the final position of the 3-pull unit (2
27).

The processing 213 of hyphenated number words is shown in Figure 755A and in Figure 5.
The processing is performed by the processing unit 114 according to the processing flow shown in Figure B. First, in initial processing 230, a dictionary lookup unit with a hyphen is stored in a buffer. Also, save the numerical value rOJ and replace the hyphen in the original dictionary lookup unit with a space.

So let's take a look at the dictionary 231), Dictionary search 2
Do 35. As a result of the dictionary search, there is no entry, that is, the word is not registered in the dictionary (23B)
, the entire dictionary lookup unit with the hyphen is stored in the dictionary information storage table 124 as an unregistered word in the dictionary (23
7).

When an entry is obtained as a result of dictionary lookup (23B).

Check whether the numerical flag is "l". If the number flag is not rlJ, this means it is not a number,
That hyphenated word;! Save the entire 2-lookup unit in the dictionary information storage table 124 as an unregistered word in the dictionary (2
37).

When rlJ is set in the number flag of the dictionary entry, the processing unit 12 converts the number J4 into the number 1 based on the entry data.
1 (239). Next, convert this numerical value into the currently saved value /Il'l(240),
Save the addition result (241), so that, for example, "20" of twenty-two' is added to "20" of "twenty" immediately before it, and becomes "22".
This is done until the final position of the dictionary lookup unit (
232).

When the step reaches the final position, the process 23 is performed in step 232.
3, the stored numerical value is used as the numerical value of the entire dictionary lookup unit with a hyphen. Next, a process 207 is performed in which this numerical value is combined with the immediately preceding numerical value.

Referring to FIGS. 6A and 6B, the numeric sequence process 215 executed by the processing unit 116 will be described. Note that in these flowcharts, the symbol “<=” indicates substitution. First, initialization 250 is performed in which the saved numerical value map a l-5ave is set to rOJ, the parameter "i" is set to rlJ, and the pointer p is set to the beginning of the character string of the dictionary lookup unit.

Next, check whether the character type pointed to by pointer p is a number (251), a scale character (252), or a decimal point (253), and if it is neither of these, the character string The entire word is stored in the dictionary information storage table 124 as a word not registered in the dictionary (255), and if it is a decimal point (253), the parameter "i" is multiplied by 10 (2
54), execute step 258 for one line. Step 258
Then, add the number of character books p to the number of saves (11).
ffnu 厘 (this p) is added to obtain a new saved numerical value.

The numerical value nun (zero p) is a value when the character (p) is viewed as a numerical value.

In step 251 or 252, it is a number,
If it is a scale character, step 257 is executed. In step 257, the saved numerical value maa1-save is multiplied by 10 and the character-like numerical value nu■ (book p) is added thereto to form a new saved numerical value.

After these processes, the pointer is incremented (259)
, repeat this process until the final position of the dictionary lookup unit (2
60), if it is the last position of the character string, the numerical value of the entire character string is saved as the numerical value 281), and the processing unit 122 executes the process 207 of combining it with the immediately preceding numerical value. With this, for example, the consecutive numbers "1,000.5" are number 4
parsed into rlooO,5J.

The process 207 of combining the values with the immediately preceding numerical value is performed by the processing unit 122 as follows. First, set the dictionary table pointer to the position immediately before the dictionary lookup unit (
270), if there is nothing in this position, it means that the first position of the storage table is the numerical value, and the numerical value of the current dictionary lookup unit is recorded in the dictionary storage table 124 (28
4), the recording position is the next position to the position indicated by the current pointer P.

In step 271, if a word exists immediately before, the entry pointed to by pointer P is "and" ((272
), and if pointer p does not point to a numerical value (273
), records the value of the current dictionary lookup unit in the position next to the position indicated by the current pointer p in the dictionary storage table 124.
(284), for example, "To him two..."
In the example, "two" is newly recorded as the numerical value "2".

In step 273, if pointer P points to a numerical value, the number of entries pointed to by pointer p is 4 (r
The number of entries pointed to by the new pointer p is 4e by adding the numerical value of the current dictionary lookup unit to p→
Ma (274). For example, “tw.

In the example “thousand”, r 2!1000=20
00 J and make the whole ``two thousand'' into one. Thereafter, the end position r of the current dictionary lookup unit is set as the end position of the entry of pointer p, that is, p→end position (282).

In step 272, if the entry pointed to by the pointer p is "and", the pointer p is moved to the previous dictionary lookup unit (275), and it is not the final position (27B).
, If L is also number 11 (277), round up the numerical value nov of the current dictionary lookup unit to the most significant digit,
Let this be the value Mar. Numerical value of current dictionary lookup unit no%
For example, F is r8J 1111 J r9aJ
rll”.

Each square is rlOJ rlOJ rlo
o J rloo J.

Therefore, the numerical value p of the entry pointed to by the pointer p →
The remainder when m is divided by m, that is, raod(p→v,
vl) is "O".

If not rQJ, increment pointer p by 2.
83), records the numerical value of the current dictionary lookup unit in the position next to the position indicated by the current pointer p in the dictionary storage table 124 (284). For example, ``I and two''
c7) In the example, "two" is newly recorded as "2" as a number (4r2J).

If the remainder is "0" in step 279, the numerical value p of the entry pointed to by the pointer p→ma is added to the numerical value m now of the current dictionary lookup unit, and a new numerical value of the entry pointed to by the pointer p is created. p→! (280). For example, in the example "two thousand and two", at this stage it is already 'two thousand'.
and'' are grouped together as r 2000J. Then, by adding M2O0, this is
"2" and 7Ia are calculated, r 2002J, and the whole is one. Thereafter, the "and" information pointed to by the pointer p÷1 is deleted from the information storage table 124 (281).

The process moves to step 282.

Let me explain with an example. For example, as shown in Figure 9, the input character string "To him two thousand
When a dictionary lookup is performed for "andtwenty-two,...", dictionary entry information as shown in FIG. 4'', the final r position is ``6'', and the part of speech is a pronoun.

In number processing, the number flag for "mass two" is "1".
” (205), and it is identified that the numerical value is [2”. In this character string, since the part immediately before "t%#0" is not a numerical value, it is stored as is in the same table 124 (206, 207, 284).

Then increment the pointer and
”.The numerical flag is “1” and the numerical value is r100OJ (205.208). Moreover, since the value immediately before that is "2" (207, 273), multiplication 2z1000 t-execution 274), Table 1
24 (FIG. 10B). Regarding the next "and", the dictionary information is stored in the table 124 as is (FIG. 100).

The pointer is further advanced and "twent7-two" is processed. As it is, the hyphen a that is not in the dictionary entry
h (212), and r20+2-224 is executed by the hyphenated number processing 213 (237.239
~241). Immediately before that is “and” (272)
, since the previous value r is 2000J (277)
, round the northernmost digit of the number "22" to rlooJ278), and perform division 273 in one line, the remainder is rO
J, so add r 2000J and "22" to 280
282), and saves the addition result r 2022J in the table 124 as a numerical value.

By this, “two thousand and
twenty-two" is recognized as r 2022J,
Processing 207 for combining the previous numerical value was performed.

Another example, as shown in FIG.
ou 5aid $1,000.5 thousand
Proceed with the analysis of was...” $1,00
0.5'' is not registered in the dictionary 18. Initially, it is a currency symbol ``-)''S'', and from the cursor and re-entry, it is recognized as a currency symbol. This is stored in the save table 124.
(214°216, Figure 13A).

Next, "1.QOo, 5" is converted into the number 44 rloooo, 5J by the number sequence processing 215. The symbol "S" immediately before it is not a numerical value, so this numerical value is recorded as it is (270 to 273.284 , Figure 13B).

The next middle sentence "thausand" is a number word, and its numerical value is r 100OJ. Since the immediately preceding value is a numerical value (
272,273), rlooo, 5x1000=100
Execute operation 274 of 0500 J (Figure 13C)
.

After the dictionary lookup is completed in this way, the contents stored in the dictionary information storage table 174 are sequentially checked. Number rlo0050
Since the currency symbol "S" exists immediately before 0J, both are combined to form a single noun entry r$1000500J (209..221-223.Figure 130).

Effects According to the present invention, when analyzing the form ygZ, regarding the number,
Replaces a number expression with a numerical value, and if there is a hyphenated number word, a series of numbers, or a numerical value immediately before it, it is combined with that and performs addition or east arithmetic to combine it into a single numerical value and unit of analysis. In addition, currency symbols and units associated with numbers are analyzed together with numerical values as a unit of analysis. This allows morpheme analysis to be performed using appropriate scale and units for expressions that include numerical values.

[BRIEF DESCRIPTION OF THE DRAWINGS] FIG. 7jS1 is a functional block diagram showing a detailed configuration example of the morphological analysis section of the embodiment shown in FIG. 2. FIG. 2 is an a-side block diagram showing the overall configuration of an embodiment in which the language analysis device according to the present invention is applied to an English-Japanese visual motion translation device, and 7JSGA diagram and FIG. 3B are morphological analysis in the embodiment shown in FIG. 1. FIG. 3 is a flow diagram showing an example of processing. FIG. 4 is a flowchart showing an example of the processing of summarizing currency symbols and units in the morphological analysis process, and FIGS. 5A and 5B are flowcharts showing an example of the processing of hyphenated numbers in the morphological analysis process. FIG. 6B is a flowchart showing an example of processing of consecutive numbers in morphological analysis processing, and FIG. 7JS7A and FIG. 7B are flowcharts showing examples of processing of combining numerical values immediately before in morphological analysis processing. FIG. 8 is an explanatory diagram showing an example of the configuration of a dictionary file with numerical flags in the same embodiment. FIG. 9 is an explanatory diagram showing an example of an input character string in the same embodiment. Fig. 12 is an explanatory diagram showing the contents of the dictionary information storage table obtained by lexicographically looking up the input character string exemplified in Fig. 12 according to the processing stage; is the dictionary currency symbol table in the same example,
Figures 13A to 13D are explanatory diagrams showing examples of the contents of a scale symbol table and a decimal point table. This is an explanatory diagram showing the symbols of the main parts 18, , morphological analysis unit 18, , dictionary 104, , dictionary search unit 110, , currency symbol/unit grouping processing unit 112, . Number word digitization processing unit 1 14,, Hyphenated number word processing unit 1 1Ei, , Number J! I! The following processing unit 118, processing unit 120 which combines currency symbols and numerals into one noun, processing unit 122 which combines numerical values and units into one noun,... Summary (de processing unit 124
10. Dictionary J) Information storage table book 3A5 5th Δ figure Tsuta 581!1 Gather 4 figure Nero B concave 7A I￥] Milk 7812I 阜δ concave/a Vermilion 9 figure/QC figure tail 700 figure, rod, HvJ... ..."raq to /H2-...You sa
t d % I r OO', ra+, hOuScLn
cl curvature αS...hair 12 figure groin J3A figure tail/38 concave Qin 13C figure

Claims (1)

[Claims] 1. Dictionary means storing dictionary data for each dictionary lookup unit, dividing an input sentence into dictionary lookup units, and performing morphological analysis for the dictionary lookup unit by referring to the dictionary means. and an analysis means, wherein the dictionary means includes, as the dictionary data, an identification display indicating that the dictionary lookup unit represents a number for a dictionary lookup unit representing a number; The dictionary means is referred to for each dictionary lookup unit included in a sentence, and when the dictionary data retrieved includes the identification display, the dictionary lookup unit from which the identification display is retrieved is used as the dictionary lookup unit of the dictionary lookup unit. It is combined with other nearby dictionary lookup units for which the above-mentioned identification indications have been retrieved, and the meanings of both dictionary lookup units are calculated together to form a single numerical value, and both dictionary lookup units are treated as a single analysis unit. A language analysis device characterized by: 2. In the apparatus according to claim 1, when the analysis unit is accompanied by a currency symbol or a dictionary look-up unit representing a unit, the analysis means collects this together with the numerical value to form a single analysis unit. A language analysis device characterized by: