JPH0772888A - Information processor - Google Patents

Abstract

PURPOSE:To provide a reading device inputting a documemt and reproducing a sound with presence according to the characteristics of the characters and the surroundings in the document. CONSTITUTION:This processor is constituted of a key board 101 and a document input device 102 for inputting the document, a natural language processor 104 analyzing the document, a characters characteristics extraction device 106 extracting the characteristics of the characters, a voice synthesizer 109 synthesizing sounds using the characteristics of the characters, a surroundings extraction device 105 extracting the surroundings from the document, an effect sound generation device 107 generating an effect sound from the extracted surroundings and a sound output device 108 synthesizing the synthesized sound and the effect sound and applying an effector.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a reading machine for inputting a document and synthesizing and outputting sounds in accordance with the features and environment of the characters in the document.

[0002]

2. Description of the Related Art Conventionally, the technology of media conversion, which is one element of multimedia technology, has been actively researched. As a reader for the blind, a document is read by a scanner, characters are cut out, characters are recognized, and a voice is read. There is also a proposal of a synthesizing device (for example, Japanese Patent Laid-Open No. 26269/1990).
1 gazette).

[0003]

However, such a conventional reading machine merely converts the input document into speech, and the means for expressing the content of the document is changed from text to speech. Without the ability to express voice is not fully exerted.

Therefore, in the present invention, the features of the characters and the environment of the scene are extracted from the contents written in the document by natural language processing, and the voice quality of the voice synthesis is changed or the sound effect is added to output. It is an object of the present invention to add emotion and presence to the sound that sounds.

[0005]

[Means for Solving the Problems] In order to achieve this object, the reading machine has a document input device for inputting a document, a natural language processing device for analyzing a sentence, and an appearance for extracting the characteristics of characters. A person feature extracting device, a voice synthesizing device that synthesizes a voice using the character code and the features of the characters, an environment extracting device that extracts an environment from the character code, and a sound effect from the extracted environment. It comprises a sound effect generator for generating sound, and a sound output device for synthesizing the synthesized voice and the sound effect and applying an effector.

[0006]

The document is directly typed with the keyboard, or the document is input as image data with the document image input device, the character area of the image data is extracted with the character area extraction device, and the character recognition is performed with the character recognition device. Is output. The natural language processing device analyzes the sentence composed of the output character code, and the character feature extraction device uses the analysis result to extract the feature of the character. Then, by adding the characteristics of the characters to the character code corresponding to the part spoken by each character, the voice synthesizer synthesizes an emotional voice. On the other hand, the environment extraction device extracts the environment of the scene using the analysis result by the natural language processing device, and the sound effect generation device outputs the sound effect corresponding to the extracted environment to express the realism. The output of the voice synthesizer and the output of the sound effect generator are output from the sound output device. At that time, various effectors can be applied depending on the location of the scene obtained from the environment extraction device.

[0007]

An embodiment of the present invention will be described below with reference to the drawings. Here, the target document is a novel. This is because the novel contains a narration and a conversational sentence in two main parts, and the emotions of the characters and the scene scene are embedded in the narration, so that the effect of the present invention can be maximized. This is to think that. In addition,
The document is not limited to a novel, but may be a newspaper, magazine, dictionary or encyclopedia.

First, a document is typed from the keyboard 101 or converted into a character code string (hereinafter referred to as text) by the document input device 102. Here, the document input device 102 includes a document image input device 201, a character area extraction device 202, and a character recognition device 203, and a character area is extracted from the document image input from the document image input device 201 using the character area extraction device 202. (For example, Akiyama, Masuda: “Extraction Method of Character Areas in Newspaper Articles”, Sho 57 Shingaku Sogaku, 1297 (1983)), and character recognition device 203 is used to extract characters from the extracted character areas. Separation
It recognizes and outputs a character code (for example, Shimizu, Takenouchi, Waki, Fujiwara, Yokoe: "Document Seisho System (1) --System Outline-"), Sho 61 Shinbun University, 15
10 (1986)). The text thus obtained is stored in the memory 103. Here, the document image input device 201 is an image scanner or a camera. The text stored in the memory 103 as described above is subjected to morphological analysis and syntactic analysis in the natural language processing device 104 and analyzed.

Next, character feature extraction will be described. The character feature extraction device 106 receives the text analyzed by the natural language processing device 104 as an input. First here
Characters are extracted, but the subject / verb pairs are extracted, and if this verb is a verb in which a human being is usually the subject (for example, walking, running, eating, etc.), the subject is a candidate character. And the subject is not in the word dictionary,
It is determined that it is a person's name, and the character is used. For example, if the result of the analysis of the part "Koi is gripping the handle" is obtained as the subject: Kouki Verb: gripping, "grip" is an action that is generally performed by animals, but "Koki" is Because it is a proper noun, it does not exist in the word dictionary and is judged as a human being,
"Koki" can be one of the characters. On the other hand, prepare a character table 402 as shown in FIG.
"Koki" 602 is registered for the man with "Character name" 601. Similarly, each time a character is extracted, its name is extracted, and the characters are sorted into male and female and registered.
However, if the same name has already been registered, registration will not be performed. Furthermore, the emotion table 40 as shown in FIG.
3 is prepared, and when the verb portion of the subject verb pair matches the verb portion 501 of the emotion table 403, the corresponding parameter 502 for speech synthesis is output to the speech synthesis device 109. For example, if "Kouki laughed."
If the verb portion 501 of the emotion table includes "laugh" 503, the corresponding parameter "emotion1"
Get 504. Next, the character uttering the conversational sentence in the sentence enclosed by the brackets is interpreted from the contexts before and after, and the voice parameter 604 of the character and the emotional parameter (if there is a conversational sentence and emotion) ( Here, "emoti
on1 ”504) is output to the speech synthesizer 109.
Then, using the voice parameter 604 of the characters and the emotional parameter 504, the conversation sentence is changed, the voice is synthesized by the voice synthesizer 109, and the sound output device 108 is produced.
To output. Here, it is assumed that the speech synthesizer can be controlled by parameters (for example, Miyatake et al .: "Prosodic features and their control in various vocalization styles", Theory of theory, D-
2, vol J73-D-2, No12). In addition, here, although limited to men and women, ages are roughly distinguished by words such as "-professor""-boy" or explanatory texts in sentences, such as "middle-aged man", and the characters It can also be a feature.

Further, similarly to the character feature extraction, the environment is extracted from the analysis result of the text output from the natural language processing device 104. First, like the extraction of characters,
Extract subject / verb pairs. Then, the subject and the verb are compared with the environment name 701 of the environment table 302 shown in FIG. 7, and if they match, the sound effect index 703 corresponding to the environment name 701 is output. For example, if the result that the subject is "rain" is obtained from the part "rainfall", the verb "rain" and the subject "rain" are compared with the environment table 302 and they match. The index "nature1" 704 of the corresponding sound effect is output. It is also possible to extract a verb / object pair in a similar manner and add a sound effect that knocks the door from a sentence such as "knocked the door." Although the tense is not taken into consideration here, it is possible to determine whether or not a sound effect is generated by considering a word such as "yesterday" or "after".
The sound effect index 703 thus obtained is input to the sound effect generator 107, the sound effect to which the index is added is generated, and is input to the sound output device 108. It should be noted that the types of sound effects may be sounds caused by natural phenomena, sounds made by vehicles, sounds made by living things, sounds made by hitting furniture and tableware, sounds made by bells, musical instruments, music, electric appliances, and crowds.

The sound effect thus obtained and the conversation between the characters are combined in the sound output device 108, the place of the scene is extracted in the environment extracting unit 301, and the effector corresponding to the place is combined. Applied to the sound. For example, in a bath, a strong echo is applied.

As described above, according to the present invention, it is possible to introduce human emotions, scenes, and the like when converting a document into sound, and to give a sense of reality to the synthesized sound.

[0013]

According to the present invention, a document input by natural language processing can be analyzed, and the features of the characters appearing in the document, that is, sex, age, and emotion can be extracted by the character feature extraction device. Therefore, it is possible to process the voice using the extracted features of the character rather than converting the content spoken by the character into a voice as it is, and to produce an emotional voice. Similarly, since the scene of the scene can be extracted by using the environment extraction device, the environmental sound of the scene can be reproduced at the same time as the utterance of the character, and the realistic scene can be reproduced by sound. It becomes possible.

[Brief description of drawings]

FIG. 1 is a block diagram of the present invention

FIG. 2 is an explanatory diagram of a document input device.

FIG. 3 is an explanatory diagram of an environment extraction device

FIG. 4 is an explanatory diagram of a character feature extraction device.

FIG. 5 is an explanatory diagram of an emotion table.

FIG. 6 is an explanatory diagram of a character table.

FIG. 7 is an explanatory diagram of an environment table.

[Explanation of symbols]

 101 Keyboard 102 Document Input Device 103 Memory 104 Natural Language Processing Device 105 Environment Extraction Device 106 Character Feature Extraction Device 107 Sound Effect Generation Device 108 Sound Output Device 109 Speech Synthesis Device 201 Document Image Input Device 202 Character Area Extraction Device 203 Character Recognition Device 301 environment extraction unit 302 environment table 401 character feature extraction unit 402 character table 403 emotion table

Claims (10)

[Claims] 1. A natural language processing apparatus for analyzing a document consisting of character codes, and a character feature extraction apparatus for extracting the characteristics of the characters in the document from the analysis result of the document output from the natural language processing apparatus. An information processing apparatus comprising: 2. A natural language processing device for analyzing a document consisting of character codes, and a character feature extraction device for extracting the features of the characters in the document from the analysis result of the document output from the natural language processing device. The character code includes the memory for storing the character code, a voice synthesizer for converting the character code into voice, and a sound output device for outputting the voice synthesized by the voice synthesizer. A reading machine characterized by synthesizing voice after adding the features of. 3. A document image input device for inputting a document as image data, and a character portion cut out from the image data,
The reading machine according to claim 2, further comprising a character recognition device that performs character recognition and converts the character code into a character code. 4. Character features to be extracted are men and women of all ages,
The reading machine according to claim 2 or 3, wherein different characters having the same character characteristics are separately output. 5. The reading machine according to claim 2, 3 or 4, wherein the character feature to be extracted is the emotion of the character, and the emotion type is emotions. 6. The character feature extraction device comprises a character feature extraction unit for extracting character features using the result of the natural language processing device, a character table and an emotion table.
3. The character table includes a character name, a voice quality name, and parameters passed to the voice synthesizer, and the emotion table includes verbs that the character can take and parameters passed to the voice synthesizer. The reading machine according to any one of 5 to 5. 7. An environment extracting device for extracting an environment using a result of a natural language processing device, and a sound effect generating device for generating a sound effect according to the extracted environment, wherein the output of the speech synthesizer. 7. The reading machine according to claim 2, wherein the sound is output to the sound output device at the same time. 8. The reading machine according to claim 7, wherein the environment is a sound of a place and its surroundings. 9. An environment extraction device comprises an environment extraction unit for extracting environment features using a result of a natural language processing device and an environment table, wherein the environment table includes an environment name, a verb associated with the environment name, and a sound effect. The reading machine according to claim 7, which comprises the index of. 10. A sound output device synthesizes an output from the voice synthesizer and an output from the sound effect generator, and applies an effector corresponding to a location extracted by the environment extracting device to the synthesized sound. The reading machine according to claim 7.