JPH02289900A - Japanese voice input assisting device - Google Patents

Abstract

PURPOSE:To facilitate the operation for confirming an input sentence by extracting a vague part automatically from the conversion result of KANA (Japanese syllabary)-KANJI(Chinese character) conversion, retrieving reference information as to the words of the vague part, and displaying it together with the conversion result. CONSTITUTION:The very vauge part which is generated when an input operator inputs a Japanese sentence received in voice after KANA-KANJI conversion is detected by using a case pattern storage part 5, a reference information storage part 6 stored with examples, equivalent expressions, and meaning explanations of the words of the vague part is retrieved, and explanatory sentences of the respective words are generated based upon those pieces of reference information and displayed. Consequently, the input operator confirms the intention of a speaker efficiently.

Description

[Detailed Description of the Invention] [Industrial Application Field] The present invention... A Japanese language that supports tasks in which the input operator confirms the input sentence to the speaker using only voice, such as when an input operator different from the speaker inputs the input text by dictation or over the phone. Concerning voice input support devices. [Conventional technology and problems] A speaker conveys a Japanese sentence to an input operator using a means that can transmit information only by voice, such as a telephone, and the input operator converts the Japanese sentence into kana-kanji based on the utterance at a location different from the speaker. When converting and inputting, ambiguities such as homonyms may occur. In such cases, the input operator often confirms the speaker's intention by thinking of example sentences or semantic explanations using the ambiguous word and conveying them to the speaker. However, the input operator may not be able to understand the speaker's intent. If an unknown term is used, it may not be possible to present a valid example sentence to the human operator, making it impossible to provide a semantic explanation. It may take a lot of effort for the speaker to convey his/her intention to the input operator. Conventionally, Japanese word processors have provided a function to present synonyms and semantic descriptions as a human support and elaboration support function, but these functions are provided on the assumption that the author will input the information himself/herself. 9 Problems when the speaker and the human operator are different, that is, there is no guidance function that automatically extracts parts with a high degree of ambiguity or a function that presents reference information for those parts. was there. [Means for solving the problems] The present invention. A speaker conveys a Japanese sentence to an input operator using a means that can transmit information only by voice, such as a telephone, and the input operator inputs the Japanese sentence based on the utterance at a location different from the speaker. Ambiguity such as homophones may occur, but in such cases, ambiguous parts are automatically removed in the conversion result to help the input operator confirm the speaker's intention (Japanese sentence). By extracting the words, searching for reference information (examples, paraphrases, semantic explanations, etc.) for those words and displaying them along with the conversion results. The purpose is to make it easier to check input sentences. [Example] An example of the present invention will be described below with reference to the drawings. 1 is an input display section that displays alphanumeric and kana input functions and conversion results and reference information; 2 is an input display control section that controls input display section 1; and 3 is an input display section that outputs all candidates. Kana-kanji conversion department. 4 is an ambiguous part extraction unit that extracts parts with a high degree of ambiguity based on case patterns and intersentence causality rules; 5 is a case pattern storage unit; 6 is a unit that extracts the relationship between the reference information of each word and the search key along with a flag for each word. A reference information storage unit 7 that stores reference information searches for examples, paraphrase expressions, and semantic explanations using words written in kanji as keys, and a reference information storage unit 6 that uses search keys assigned to the reference information of each word. A reference that generates an explanatory text based on a flag indicating the relationship with reference information. Information retrieval section 8 is a device control section that controls the entire device. below. The operation of the device of the present invention will be explained using example sentences. When the input operator receives the input sentence ``Gogotanari Toukitsuunitsu'' over the telephone or the like, the input display unit 1 inputs the kana character string manually, and the kana-kanji conversion unit 3 converts the kana character string into kana-kanji. In this case, candidates as shown in FIG. 3 are obtained, and the ambiguous part extraction section 4 refers to the case pattern storage section 5 and extracts the ambiguous part from the conversion result.

The case pattern storage section 5 stores case patterns as shown in FIG. For each verb included in the conversion result. The case pattern in the conversion result is compared with the case pattern stored in the case pattern storage section 5. This is because the case patterns of ``tsuki'' do not match. The luminosity as a candidate is lowered, and the case of ``r'' in ``attached'' is concrete. Although it remains a candidate because it can take on a very wide semantic category such as abstract objects, there is a pattern in which the locative case is added to the ni case in the case of ``arriku'', and the ``ni'' case is the case in all cases of the conversion result. is the place name. Therefore, ``arrive'' and ``tukuku''
It was determined that there was little ambiguity between ``Tsu'' and ``Tsu'', and ``Aruku'' was determined to be the first candidate. Reference information for the word is not automatically displayed. On the other hand, the kana-kanji conversion results for "Tokyo" each include place names, and likelihood cannot be assigned depending on the case pattern. It is determined that the parts have a high degree of ambiguity, and reference information is searched for the words contained in these parts. The reference information storage unit 6 stores reference information for each word along with a flag indicating the meaning of each word. The reference information search unit 7 searches for reference information for the word and generates an explanatory text using those flags. ``Kyobe'' is stored for ``Kyo'', and ``etc.'' for ``etc. It detects this and generates explanatory sentences such as ``Kyo, meaning Kyoto,'' and ``etc, meaning etc.'' In addition, for "Ukyo", a flag indicating that "Kyoto" is a high-ranking place name is stored, and in this case, an explanation such as "Ukyo in Kyoto" is generated. Figure 3 shows an example of screen display. Conversion results and reference information for parts judged to have a high degree of ambiguity (Tokyo: the capital Tokyo, etc.: "etc." meaning "etc.", Kyoto:
Kyo means Kyoto. Ukyo: Ukyo in Kyoto) is displayed, and the input operator confirms the speaker's intention by conveying this reference information to the speaker. In addition, in the case of proper nouns such as personal names, there are many homophones, so for each kanji character that makes up a proper noun, common words and kunyomi containing those kanji characters are stored in the reference information storage unit 6. It has been done. If there are homophones in the conversion result, the input display system 1n section 2 displays their reference information on the input display section l. An example is shown below.
Example: Koji [Happiness, Meiji] Kouji [Tagayazu Tsukasa] Kouji [Yasu number 2] Furthermore, any word that the ambiguous part extraction unit 4 has judged to have a low ambiguity can also be displayed with the cursor on the input display unit 1. etc., and by giving a reference information search instruction, the input display control unit 2 detects the word and transmits a reference information search request to the device system 12I1 section 8,
The device control section 8 transmits a search request together with a search key to the reference information search section 7. The reference information of the word is displayed by executing a search for the reference information and returning the reference information to the manual display control unit 2. [Effects of the Invention] As explained above, according to the present invention. It uses case patterns to detect parts with a high degree of ambiguity that occur when the input operator converts Japanese sentences received by voice into kana-kanji and inputs them, and also uses example 2 paraphrase expressions of those words.
Search the reference information storage unit that stores semantic explanations. Furthermore.
An explanatory sentence for each word is generated and presented based on the reference information, which has the advantage of streamlining the work of six operators to confirm the speaker's intention. Furthermore, although the present invention has described an embodiment in which a human operator intervenes, a voice recognition device is used instead of the input operator to recognize the human sentence, generate a confirmation sentence for the conversion result, and generate the 611 recognition sentence. It is also possible to synthesize speech and convey it to the speaker. It has the advantage of being able to unattend the input of Japanese that is conveyed by voice.

[Brief explanation of drawings]

FIG. 1 is a Bronok diagram showing one embodiment of the present invention, and FIG. 2 is an example of a case pattern storage unit. FIG. 3 is an embodiment of a display screen. l...Input display section. 2...Manual display control unit, 3...Kana-Kanji conversion unit, 4...Ambiguous part extraction unit, 5...Case pattern storage unit. 6...Reference information storage unit, 7...Reference information search unit, 8...Device control unit. Patent applicant Nippon Telegraph and Telephone Corporation

Claims (1)

[Claims] An input means capable of inputting alphanumeric and kana characters, and a display including kana-kanji conversion results including all candidates for the input kana character string, usage examples, paraphrase expressions, and semantic explanations of the converted words. An input display section that has a display means configured to display, an input display control section that controls the input display section, a kana-kanji conversion section that outputs all candidates for the input kana character string, and a conversion result. ambiguous part extraction unit that extracts parts with a high degree of ambiguity, and reference information that includes usage examples, paraphrases, and semantic explanations for each word, as well as flags that indicate the semantic relationship between each word and the reference information. A reference information storage unit searches for reference information of the word using the word written in kanji as a search key, and searches for the word using the searched reference information and a flag indicating the semantic relationship between the word as the search key and the reference information. 1. A Japanese voice input support device, comprising: a reference information search unit that generates an explanatory text.