JPH0765010A - Sentence converting device - Google Patents

Abstract

PURPOSE:To improve the comprehensive faculty of a receiver of information and to transmit the contents of information with extremely high precision by transforming the data into the contents proper to an output form when the contents of information are transmitted through various media. CONSTITUTION:A type detecting part 1 inputs a structured medium document and detects its type. A conversion rule is retrieved out of those document conversion rules which are previously stored in a document conversion rule storing part 2. This retrieved rule is sent to a rule activating part 3 which activates a document conversion rule 2. Then the part 3 sends the activated rule 2 to a document conversion interpreting part 4 in an interpretable state. An activated conversion rule 8 is temporarily stored in a sentence converting device until a read-aloud document 7 is completely generated. The part 4 uses a structured medium document 6 as the data to successively interrupt the rule 8. The activated rule 8 is expanded into a storage area of a computer and then interpreted by a general method that is used for collation of patterns.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a text conversion device for converting a description format so that an existing information content can be appropriately adapted to the purpose of use of the information content without impairing the information content described therein.

[0002]

2. Description of the Related Art Conventionally, when transmitting some information from one side to the other side, information transmission has been performed which is bound to the properties of a medium or a device which is a communication means.

For example, even when transmitting the same event, different description formats are used in accordance with respective transmission modes in media such as televisions that handle audio information and media such as newspapers that handle character information. Also, regarding media that handle audio information, the description format differs between television broadcasting and radio broadcasting. That is, in the radio broadcast, the portion lacking the image information is complemented, and the text and the details are provided to provide news and topics having the same contents as the television broadcast.

In recent years, multimedia information equipment has been developed which transmits information in multiple channels by integrating various media by making full use of video and audio. If this multimedia information device is used, for example, a document such as a newspaper article is converted into a description format of the information content, for example, by using voice, the information content can be efficiently added to the recipient and in a convenient form. Can be told.

Therefore, attempts have conventionally been made to take advantage of technically changing the device. It is a voice synthesis technology for converting document information into voice information. The focus of the conventional speech synthesis technology has been to refine the phonetic signal level so that the descriptive expressions such as sentences are faithfully reproduced, and phonetic sounds are simulated with a voice quality closer to human reading. Then, we have aimed to realize high-quality speech synthesis as an acoustic physical signal.

[0006] Considering the development of the multimedia information equipment utilizing various kinds of media mentioned above, the intelligibility of the information content is considered not only in the acoustic quality of the synthesized speech but also in the environment where the listener is placed. , And from the viewpoint of improving the understanding of the information content, the voice must be transmitted. That is, the sentence to be uttered is not only read literally but also composed of sentence patterns and phrases suitable for reading aloud, and it is necessary to realize communication in a sentence that is easy to hear.

As a familiar example, even if a newspaper article is digitized by some means and the digitized sentence is read as a target of speech synthesis with a sound quality almost equal to that of a human being, the intelligibility and sentence of the sentence are still kept. The understanding level of the contents is not perfect. The reason for this is that the sentence style premised on reading and the sentence style premised on reading aloud are different.

[0008] The sentence given as an example in FIG. 1 is a newspaper article whose content is "to convey the resignation of the president of Fiji". When actually reading aloud as it is, it turns out that it is difficult to make phonetics by the sentence style of the fact explanation on the premise of reading.

For example, there are the following problems.

Reading “[” as “parentheses” breaks the rhythm of the sentence.

The meaning cannot be transmitted even if “=” is read as “Ikoru” or “Ishito”.

[0012] When aloud, a sentence with a blatant phrase such as "-being inaugurated as president at the age of fifty-four" has the impression that it is inaccurate information transmission because there is no narrative part.

[0013] If only the end of the sentence is mentioned, it will be "expressed", "transmitted", "prospect", "was", "rejected", etc. If you continue speaking, you will get a bad impression.

As shown in the above-mentioned example, (1) the form of the sentence structure is different depending on whether the recipient is the reader or the listener, and (2) the characteristic of the rhetoric in the sentence expression. different.

Generally speaking, a communication means such as a sentence presented at a lecture or a voice media such as a television or a radio is mediated by voice so that it is easy for a listener to hear, and has a meaning and meaning in order to promote clear understanding and accurate interpretation. Words and sentence patterns are used that are closely related to phonetics. From this, the quality that considers the intelligibility and comprehension of the information content is limited only by the level that can be reached by creating the phonetic sound that is transmitted as speech by blindly reading the surface form of the sentence.

Conventionally, the sentence conversion for changing the normal body into the polite body has been already realized, but it is premised that the sentence is expressed in the same medium even after the conversion. Further, there is a technique for converting a sentence form suitable for a reader into a sentence form suitable for machine processing, but it is intended for operation on a single medium, and does not enable appropriate information transmission to different media.

[0017]

Therefore, in the transmission of information from now on, not only the acoustic quality of the synthesized speech but also the intelligibility of the information content in consideration of the environment in which the listener is placed and the information content Speech must be transmitted from a comprehensive perspective that improves comprehension.
That is, the sentence to be uttered is not only read literally but also composed of sentence patterns and phrases suitable for reading aloud, and it is necessary to realize communication in a sentence that is easy to hear. Further, not only when transmitting by voice but also when transmitting information by image or paper, it is necessary to consider the position of the person to be transmitted in the same manner.

In view of such circumstances, the present invention aims to appropriately convey a certain transmission content through different media without impairing the information content. For example, when transmitting information content to a recipient through voice media such as voice synthesis, by converting the symbol example into a description format suitable for phonetic representation, the information content can be transmitted smoothly and at the same time. The aim is to improve the intelligibility and understanding of the listener who is the recipient.

[0019]

The sentence conversion device of the first invention is created in accordance with the purpose of transmission, and is classified data indicating the classification of the purpose or purpose of the communication and the classified data. A document storage means capable of storing or specifying a document consisting of content data indicating the content of the purpose or use of
Detecting means for detecting the type of the document retrieved from the document storing means from the type data, and content data for each type data,
A conversion rule storage unit that stores a document conversion rule for converting into a predetermined output form, and content data for each type data detected by the detection unit is interpreted by the document conversion rule and converted into a predetermined output form The rule interpreting means and the output means for outputting the document converted into a predetermined output form by the rule interpreting means by voice, image or printout.

Note that the document is a document responsible for transmitting information content, and its actual form does not matter here. For example, there are forms such as an electronic document, a document on paper, and a document encoded as a signal.

The sentence conversion device of the second invention has an interpreting means for interpreting the content data in the first invention so as to correspond to the document conversion rule stored in the conversion rule storage means.

[0022]

[Operation] The operation of the document conversion apparatus of the first invention will be described.

The detection means retrieves the type data indicating the type of the document from the document storage means. Further, the conversion rule storage means stores a document conversion rule for converting the content data for each type data into a predetermined output form. Then, the rule interpretation unit interprets the content data for each type data detected by the detection unit according to the document conversion rule, and converts the content data into a predetermined output form. The output means outputs the converted output document by voice, image or printout.

In the document converting apparatus of the second invention, the interpreting means first interprets the content data retrieved from the document storing means so as to correspond to the document converting rule stored in the converting rule storing means, and then the rule. Send the data to the interpreter.

[0025]

EXAMPLE A first example of the present invention and a processing example 1 thereof will be described below.
Will be described with reference to FIGS. 2 to 6 and FIG.

FIG. 2 shows the system of the first embodiment.

This system includes a type detection unit 1 for reading the type of a structured media document, a document conversion rule storage unit 2 for storing the document conversion rules, and a document conversion rule according to the document type read by the detection unit 1. A rule activation unit 3 that retrieves the document conversion rule storage unit 2 and activates the rule, a document conversion interpretation unit 4 that converts a structured media document using the document conversion rule, and a document converted by the document conversion interpretation unit 4 The transmission device includes a document generation unit 5 that generates a read-aloud document that is a voice, and a voice synthesis unit 9 that converts the document into voice.

Reference numeral 6 is a structured media document that describes the information content to be converted, reference numeral 7 is a read-out document of the conversion result, and reference numeral 8 is the activated conversion rule. The solid line indicates the processing flow, and the dotted line indicates the data reference path and flow.

A method of converting information contents in the system having the above configuration will be described in detail.

(1) Detection of Document Type of Structured Media Document The type detection unit 1 inputs a structured media document and detects its type. Structured media document 6 shown in FIG.
In the conversion example targeted at, the type from the document file
The e attribute is searched to determine that it is a newspacer. Based on this result, the document conversion rules stored in the document conversion rule storage unit 2 in advance are searched for a conversion rule for newspacer, and the resulting conversion rule is transmitted to the rule activation unit 3.

(2) Activation of type-specific document conversion rule The rule activation unit 3 activates the document conversion rule 2, puts it in an interpretable state, and informs the document conversion interpretation unit 4 of the conversion rule. The activated conversion rule 8 is temporarily stored in the device and is stored until the reading document 7 is completely generated.

(3) Interpretation of Document Conversion Rule The document conversion interpretation unit 4 sequentially interprets and executes the document conversion rule 8 using the structured media document 6 as data. The document conversion / interpretation unit 4 applies the condition part of the rule to the content part indicating the information content of the structured media document 6 to obtain a symbol example in which the search condition items match. Source the condition part
Then, "Nandy 24th = ..." is obtained. By adapting the action part to this symbol example, the actual document is rewritten. The action section describes a procedure for inputting a document and converting the document directly by analyzing the keywords included in the document. Document conversion interpretation unit 4
At, this condition judgment and the symbol example suitable for the judgment are acquired. Further, the symbol example data is set as an operation target of the procedure of the action section. Interpret and execute the above series of rules. This operation is performed for all the given document conversion rules.

The activated conversion rule 8 is developed in, for example, a storage area of a computer and is interpreted by using a general method in pattern matching. In the document conversion / interpretation unit 4, the keyword of the structured media document (in this example, the keyword consisting of English characters and the trailing “:” symbol) and the document paired with it (the structured media document 6 of FIG. 2).
In the example shown in, the part from the end of the keyword to the next keyword) is extracted. In the document conversion and interpretation unit 4, first,
Judge the conditional section. To judge, (1) match with the specific keyword, (2) do not match with the specific keyword,
(3) The condition part is judged by a general matching operation such as matching a part of the keyword. If it is determined that the condition is satisfied by the matching operation of the condition part,
Next, the action part of the document conversion rule is evaluated.

(4) Document Generation The action section is composed of a document rewriting condition and a document transformation rule, and inputs a document associated with a keyword (extracted together with the keyword) and constructs the document. Rewrite the document by changing the word or phrase. First, the rewriting condition is checked by the above-described general collation operation, and if the condition is satisfied, the transformation of the document is executed. The document to be transformed is connected to the non-transformed part. In this way, the generated document is created.

In the action section, the document is subjected to, for example, morphological analysis or syntactic analysis as necessary, and the result is used again for the transformation operation of the action section. For example,
When there is an operation of changing the style of the tail verb in the action part, the document is once subjected to morphological analysis to check the type of the tail verb. According to the type, the document of the part is converted from the normal body "do" to the polite body "shimasu".

The result of interpretation and execution of the document conversion rule is sequentially transmitted to the document generation unit 5. In the document generation unit, the sequentially given document is converted into a read-aloud document 7 into text.

(5) Speech Synthesis The phoneticized document generated by the document generation section 5 is converted into speech in the speech synthesis section 9 using a speech synthesis device having a function of generating a speech waveform from the document.

Processing Example 1 According to the first embodiment, the newspaper document shown in FIG. 1 will be taken up and an example of rewriting the document will be described.

(1) Detection of Document Type of Structured Media Document The original document shown in FIG. 1 is changed into a structured media document by an apparatus for analyzing the structure of a sentence. This structured media document becomes the input of the device of the present invention. FIG. 3 is a structured media document that is an input to the device shown in FIG. The type detection unit 1 finds out the sentence type in FIG. Based on the result of this determination, the type detection unit 1 obtains the document conversion rule relating to newspacer from the document conversion rule storage unit 2 and sends it to the rule activation unit 3.

(2) Activation of type-specific document conversion rules FIG. 4 shows a list of activated conversion rules.

The above-mentioned rule is interpreted by the document conversion / interpretation unit 4 of the conversion rule in FIG. 2 to convert the document.

FIG. 6 is an expanded list of action parts. The document conversion / interpretation unit 4 inputs the text portion of the document based on the rewriting condition, and when the rewriting condition is satisfied, transforms the document according to the rewriting rule.

FIG. 5 shows a document portion corresponding to the keyword text, and the document is rewritten according to the rule of FIG. The rewritten result is sent to the document generation unit 5.
It should be noted that symbols (A) to (Y) on the shoulders of FIG. 5 are added for convenience in order to show in which part the rewriting conditions match with respect to the document of FIG. The same symbol (P) ~
In the case of (Y), the description of the rules is omitted due to space limitations.

(3) Example of Conversion Result of Processing Example 1 FIG. 18 shows a result of rewriting the original document shown in FIG. 1 based on the method of the present invention. When this document is output by voice, the contents of the original document shown in FIG. 1 can be understood very well.

Next, a second embodiment of the present invention and a processing example 2 thereof will be described with reference to FIGS. 7 to 12 and 19.

FIG. 7 shows the system of the second embodiment.

In this system, a type input unit 1 for manually inputting the type of a document bearing information content, a document conversion rule storage unit 2 for storing document conversion rules, and a type input unit 1
The document conversion rule corresponding to the document type specified by is searched from the document conversion rule storage unit 2, and the document whose structure is instructed through the rule activation unit 3 for activating the rule and the document input unit 6 is used as the document conversion rule. The document conversion / interpretation unit 4 converts the converted document into a voice, and the document generation unit 5 generates the document converted by the document conversion / interpretation unit 4 as a reading document whose voice is transmitted by the transmission device. To be done.

Reference numeral 6 is a document input unit for manually instructing the system to describe a document in which information content to be converted is described, reference numeral 7 is a read-out document of the conversion result, and reference numeral 8 is activated. It is a conversion rule. Then, the solid line shows the flow of processing, and the reference and flow of broken line data. A broken line 10 indicates a path for instructing document input through the document input unit 6.

A method of converting information content in the system having the above configuration will be described in detail.

(1) Input of Document Type of Document Carrying Information Content The type input unit 1 is used to input the document type of the document carrying information content. In the conversion example for the document 11 in FIG. 7, the report is instructed to the document file. Based on this input instruction, the system searches the document conversion rule storage unit 2 for a conversion rule stored in advance for the report, and transmits the conversion rule obtained as a result to the rule activation unit 3.

(2) Activation of type-specific document conversion rule The rule activation unit 3 develops the conversion rule, makes rule interpretation possible, and informs the document conversion interpretation unit 4 of the conversion rule. The expanded rule 8 is temporarily held in the device and is stored until the reading document 7 is completely generated.

(3) Interpretation of Document Conversion Rule In the document conversion interpreting unit 4, the document inputting unit 6 is used to obtain the attribute indicating the division of roles of the document data (broken line 1).
0), the document conversion rule 8 is sequentially interpreted and executed. The document conversion / interpretation unit 4 applies the condition part of the rule to the attribute information to obtain a symbol example in which the search condition items match. Attribute information is manually provided from the document input unit 6 and the condition part is addressed
Let s. Then, “Distribution destination: (Production), (Engineering)” is designated. The system adapts the action part to this symbol example to rewrite the actual document. The action section describes a procedure for inputting the instructed document and directly converting the document by analyzing a specific phrase included in the instructed document. The document conversion / interpretation unit 4 obtains the condition judgment and a symbol example suitable for the judgment, and makes the symbol example data subject to the procedure of the action unit. This series of operations is executed for all the given document conversion rules.

The activated conversion rule 8 is, for example, expanded in the storage area of the computer and interpreted by using a general method in pattern matching. The document conversion / interpretation unit 4 obtains the instructed attribute and the subsequent document portion from the document input unit 6. The document conversion / interpretation unit 4 first determines the condition section based on the attribute. To judge, (1) match with the specific keyword, (2) do not match with the specific keyword,
(3) The condition part is judged by a general matching operation such as matching a part of the keyword. The rule judged that the condition is satisfied by the matching operation of the condition part is
The action part is then evaluated.

(4) Document Generation The action section is composed of document rewriting conditions and document transformation rules. The document given along with the attribute is input, and the document is rewritten by modifying the words and phrases that compose the document. First, the rewriting condition is checked for the document by the above general matching operation. When the condition is satisfied, the transformation for the document is performed. The transformation target document is connected to the non-transformation part to create a generated document.

In the action section, the document is subjected to, for example, morphological analysis or syntactic analysis as needed, and the result is used again for the transformation operation of the action section.

The results of interpretation and execution of the document conversion rules are sequentially transmitted to the document generation section 5. The document generation unit 5 converts the sequentially given document into a read-aloud document 7.

(5) Speech Synthesis The phoneticized document generated by the document generation section 5 is converted into speech in the speech synthesis section 9 by using a speech synthesis device having a function of producing a speech waveform from the document.

Processing Example 2 Based on the system shown in FIG. 7, an example of processing for rewriting a business trip report will be described.

(1) Document Type Indication of Document FIG. 8 shows an original document held in the document input unit 6.

FIG. 9 shows an attribute (written and shown in English letters) sent to the document conversion and interpretation unit 4 and a document attached to the attribute. It is assumed that the original document is instructed and sent manually through the document input unit 6. Then, the type input unit 1 instructs the sentence type of FIG. Based on the instruction result, the system obtains the document conversion rule regarding the report from the document conversion rule storage unit 2 and sends it to the rule activation unit 3.

(2) Activation of type-specific document conversion rules FIG. 10 conceptually shows the activated conversion rules.

The above-mentioned rules are interpreted by the document conversion / interpretation unit 4 of the conversion rules of FIG. 7 to convert the documents.

FIG. 11 shows the action part expanded and conceptualized. The document conversion / interpretation unit 4 inputs the text portion of the document based on the rewriting condition, and when the rewriting condition is satisfied, transforms the document according to the rewriting rule.

FIG. 12 shows a document portion corresponding to the keyword text, and the document is rewritten according to the rule of FIG. The rewritten result is sent to the document generation unit 5. It should be noted that the symbols (A) to (G) on the shoulders in FIG.
It is added for the sake of convenience in order to show in which part the rewriting conditions match with respect to the document.

(3) Example of Conversion Result of Processing Example 2 FIG. 19 shows the conversion result shown in the processing example 2 described above.
As a result, the contents of the business trip report can be understood very well by voice.

Processing Example 3 Next, a second processing example using the system shown in FIG. 2 will be shown with a cooking instruction as a text example.

(1) Detection of Document Type of Structured Media Document FIG. 13 is an original document, and FIG. 14 is a structured media document 6. It is assumed that the original document shown in FIG. 13 becomes the structured media document shown in FIG. 14 via the sentence structure analysis device. The type detection unit 1 finds out the sentence type in FIG. Based on this determination result, the type detection unit 1
The document conversion rule regarding pe is obtained from the document conversion rule storage unit 2 and is transmitted to the rule activation unit 3.

(2) Activation of Type Document Conversion Rule FIG. 15 conceptually shows the conversion rule for the activated recipe. The document conversion / interpretation unit 4 interprets the above rules and converts the document.

FIG. 16 shows the action part developed and conceptualized. The document conversion / interpretation unit 4 determines the sequence part and the how-to based on the rewriting condition.
When a rewriting condition is satisfied by inputting a partial document, the document is transformed according to the rewriting rule.

FIG. 17 shows a document portion corresponding to the keywords sequence and how-to, and the document is rewritten according to the rules of FIG. The rewritten result is sent to the document generation unit 5. It should be noted that the characters (A) to (B) at the shoulders of FIG. 17 are added for the sake of convenience in order to show in which part the rewriting conditions match with respect to the document of FIG.

(3) Example of Conversion Result of Processing Example 3 FIG. 20 shows the conversion result of Processing Example 3. As a result, the contents of the cooking instruction can be understood very well by voice.

As described above, when the information content is transmitted through various media, even if the transmission medium is not limited to voice utterance, by converting it into a symbol example suitable for voice synthesis, it is possible to hear the uttered content. It is possible to improve the intelligibility and understanding of the sentence content when the information is played, and to transmit the information content extremely accurately.

According to this apparatus, the sentence portion described with the intention of communication using the visual information is appropriately omitted and the document to be uttered is converted, so that a redundant and unnatural sentence is not read aloud. Therefore, there is an advantage that the viewing load of the listener can be reduced.

In the above two embodiments, the structured media document is converted so that the information can be easily understood by voice to the person to be transmitted. However, instead of this, the structured media document is converted.
When transmitting information to the recipient by image or paper,
A document conversion rule may be set so that information can be easily viewed on an image or paper, and the document generated by this conversion may be output by a display or a printer.

[0075]

As described above, according to the present invention, when the information content is transmitted through various media, the data is converted into the content suitable for the output form and is output. And improve the understanding of sentence content,
Information content can be transmitted extremely accurately.

[Brief description of drawings]

FIG. 1 is a computerized newspaper article document.

FIG. 2 is a schematic diagram of a system showing a first embodiment of the present invention.

FIG. 3 is a structured media document in FIG.

FIG. 4 is a list of activated conversion rules.

FIG. 5 is a document portion corresponding to a keyword text.

FIG. 6 is a rule of a text action part.

FIG. 7 is a schematic diagram of a system of a second embodiment.

FIG. 8 is an original document stored in a document input unit 6.

FIG. 9 is an attribute sent to the document conversion and interpretation unit 4 and a document attached to the attribute.

FIG. 10 is a conceptual diagram of activated conversion rules.

FIG. 11 is a conceptual illustration of an expanded action section.

FIG. 12 is a document portion corresponding to a keyword text.

FIG. 13 is an original document in a third processing example.

FIG. 14 is a structured media document of the original document in FIG.

FIG. 15 conceptually shows a conversion rule for activated recipe.

FIG. 16 is a diagram in which an action part is expanded and conceptualized.

FIG. 17: Keywords sequence and how-to
Is a document part corresponding to.

FIG. 18 is a result of conversion in processing example 1.

FIG. 19 is a result of conversion in processing example 2.

FIG. 20 is a result of conversion in processing example 3.

[Description of Codes] 1 type detection unit 2 document conversion rule storage unit 3 rule activation unit 4 document conversion interpretation unit 5 document generation unit 6 structured media document 7 read-aloud document 8 activated conversion rule 9 speech synthesis unit

Claims (2)

[Claims] 1. A document storage unit capable of storing or designating a document composed of type data indicating a type of transmission and a type of use and content data divided for each type data and showing contents of a purpose and a use of transmission. Detecting means for detecting the type of the document retrieved from the document storing means from the type data, and a conversion rule storage storing a document conversion rule for converting the content data of each type data into a predetermined output form Means, a rule interpretation means for interpreting the content data for each type data detected by the detection means by the document conversion rule, and converting the content data into a predetermined output form, and a document converted into the predetermined output form by the rule interpretation means. Is a voice, an image, or a printout, and a text conversion device. 2. The sentence conversion device according to claim 1, further comprising an interpreting unit that interprets the content data so as to correspond to the document conversion rule stored in the conversion rule storage unit.