JP2005267399A - Speech dictionary search system and speech dictionary search program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a speech dictionary search system and a speech dictionary search program which can reduce false search caused by homonyms. <P>SOLUTION: A speech dictionary search system 10 is equipped with a voice input part 21 to which a voice 20 is inputted, a dictionary part 60 for storing dictionary data, a speech recognition part 22 for generating input character string data 31 corresponding to the inputted voice 20, an object decision part 50 for receiving the input character string data 31 and deciding a candidate for a search object, a dictionary search part 55 for performing the dictionary search by referring to the dictionary data to decide the candidate for the search object, and an output part 70 for outputting the result of the dictionary search. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

  The present invention relates to a voice input type dictionary search system and a voice dictionary search program used in the system.

  A speech recognition device that recognizes input speech and converts it into a character string is known. Here, “speech recognition” is to mechanically extract the linguistic content that the speaker intends to convey from speech.

  The speech recognition apparatus disclosed in Patent Document 1 includes a plurality of speech recognition units based on a unit for inputting speech information including speech, a unit for detecting a situation related to speech recognition, and situation information indicating the detected situation. Means for selecting at least one of the means; Here, each of the plurality of speech recognition means executes a recognition process of the input speech. The information indicating the situation includes at least one of information on the speaker (gender, speed, dialect, etc.), information on the word used, and a noise situation.

  A voice response system is also known. As this voice response system, for example, there is a CTI (Computer Telephony Integration) system. This CTI system can recognize voice input via a telephone and can operate based on the input voice or respond with voice.

  The voice response device disclosed in Patent Document 2 is a device that recognizes an input voice signal and makes a response based on the recognition. The voice response device includes voice recognition means, storage means, execution means, and confirmation means. The voice recognition means recognizes the input voice signal. The storage means stores a dictionary in which a word is associated with information related to the homonym of the word. An execution means performs the response based on the information regarding the homonym for a sound input to the input source of the voice signal when the voice recognition result by the voice recognition means includes a homonym. The confirming means confirms the speech recognition result based on the input source response to the response.

Japanese Patent Laid-Open No. 2003-140991 JP-A-11-153998

  An object of the present invention is to provide a speech dictionary search system and a speech dictionary search program that can reduce erroneous searches due to homonyms.

  Another object of the present invention is to provide a speech dictionary search system and a speech dictionary search program capable of searching for a desired word from a plurality of homophones based on a user's free wording.

  [Means for Solving the Problems] will be described below using the numbers and symbols used in [Best Mode for Carrying Out the Invention]. These numbers and symbols are added in parentheses in order to clarify the correspondence between the description of [Claims] and [Best Mode for Carrying Out the Invention]. However, these numbers and symbols should not be used for the interpretation of the technical scope of the invention described in [Claims].

  The speech dictionary search system (10) according to the present invention includes a speech input unit (21) to which speech (20) is input, a dictionary unit (60) that stores dictionary data (61, 62), and input speech. A speech recognition unit (22) that generates input character string data (31) corresponding to (20), and a target for determining a search target by receiving input character string data (31) and referring to dictionary data (61, 62) A determination unit (50), a dictionary search unit (55) for performing a dictionary search by referring to the dictionary data (61, 62) with respect to the determined search target, and an output unit for outputting the result of the dictionary search ( 70). Moreover, this object determination part (50) is provided with the 1st determination part (51), the homonym extraction part (52), and the 2nd determination part (53). The first determination unit (51) generates first character string data (32) and second character string data from at least one input character string data (31). The first character string data has contents supplementing the contents indicated by the second character string data. A homonym extraction unit (52) refers to the dictionary data (61, 62), detects a plurality of homonyms of the word indicated by the second character string data, and indicates a plurality of homonyms each of the plurality of homonyms. The homonym data (33) is generated. A 2nd determination part (53) determines whether there exists a common point between the characteristic of each of several homophone data (33), and the characteristic of 1st character string data (32). Then, the second determination unit (53) uses the same homonym data (33) having a common point among the plural homonyms data (33) as a search word data (34) indicating the search target. It outputs to a search part (55). The dictionary search unit (55) performs a dictionary search on the search target based on the search word data (34) and the dictionary data (61, 62), and outputs the dictionary search result to the output unit (70). .

  In the speech dictionary search system (10) according to the present invention, the first determination unit (51) includes a second character that follows the character string indicated by one input character string data (31) with the preceding first character string. Split into columns. The first determination unit (51) sets the data indicating the first character string as the first character string data (32), and sets the data indicating the second character string as the second character string data. Here, the first character string includes at least one of a noun, a noun phrase, an adjective, an adjective phrase, a verb, and a verb phrase. The second character string includes a noun or a noun phrase. The first character string has contents that supplementarily explain the contents indicated by the second character string.

  In the speech dictionary search system (10) according to the present invention, the first determination unit (51) uses the input character string data (31) input earlier among the two input character string data (31). Two character string data are set, and input character string data (31) input later is set as first character string data (32). Here, the character string indicated by the first character string data (32) includes at least one of a noun, a noun phrase, an adjective, an adjective phrase, a verb, and a verb phrase. The character string indicated by the second character string data includes a noun or a noun phrase. The first character string data has contents supplementing the contents indicated by the second character string data.

  The speech dictionary search system (10 ′) according to the present invention may further include a first button (81) and a second button (81) connected to the first determination unit (51). At this time, the first determination unit (51) sets the input character string data (31) input when the first button (81) is pressed as the first character string data (32). Further, the first determination unit (51) sets the input character string data (31) input when the second button (81) is pressed as the second character string data.

  In such a speech dictionary search system (10), the second determination unit (53) reads the word indicated by the plural homonym data (33) and the character string indicated by the first character string data (32). It is determined whether there is a common reading between readings. And the 2nd determination part (53) extracts the common part which has a common reading from the character string which 1st character string data (32) shows, and the same sound anomaly data which shows the same word as the common part (33) is output as search term data (34).

  Further, the second determination unit (53) determines whether there is a common kanji character between the character string indicated by each homonym data (33) and the character string indicated by the first character string data (32). Determine if. And this 2nd determination part (53) outputs the homonym synonym data (33) containing a common Chinese character as search word data (34).

  Further, the second determination unit (53) refers to the dictionary data (61, 62), and the first sentence string data (32) indicates the explanation of the word indicated by each homonym data (33). It is determined whether there is a common Kanji character between the character string. Then, the second determination unit (53) outputs the homonym word data (33) corresponding to the explanatory text including the common kanji as the search word data (34).

  Thus, according to the speech dictionary search system (10) according to the present invention, the second character string and the first character string that supplementarily explains the second character string are generated from the input speech (20). The Here, there are no particular restrictions on the words and phrases that should be used in the first character string. The user may describe the intended word using arbitrary words / phrases so that the user can easily explain. This speech dictionary search system (10) refers to “kanji” and “reading” included in the first character string, so that even if the second character string has a plurality of homonyms, You can search for words. That is, the user's intention embedded in the voice (20) or the input character string data (31) is recognized. In this way, erroneous searches due to homonyms are reduced, and a dictionary search can be efficiently performed on input speech.

  The speech dictionary search system (10) according to the present invention may further include a user assistance unit (90) connected to the target determination unit (50) and the output unit (70). When the target determining unit (50) cannot identify the search target, the user auxiliary unit (90) notifies the user of the guidance / guidance / promotion of utterance through the output unit (70). This improves the convenience for the user.

  The speech dictionary search program according to the present invention is executed by a computer in the above-described speech dictionary search system (10). That is, the speech dictionary search program includes (a) a step of generating input character string data (31) corresponding to the input speech (20), and (b) a step of generating at least one input character string data (31). A step of generating one character string data (32) and second character string data, and (c) detecting a plurality of homonyms of the word indicated by the second character string data with reference to dictionary data (61, 62) A step of generating a plurality of homonym data (33) indicating each of the plurality of homonyms, (d) characteristics of each of the homonyms data (33) and first character string data ( 32) determining whether there is a common point with the characteristic of (32), (e) a single homonym word data (33) having a common point among a plurality of homonym word data (33), Search term indicating the target of the search Determining as chromatography data (34), with reference to (f) the dictionary data (61, 62), and a step of performing a dictionary search to the object indicated by the search term data to the computer.

  (B) generating step (b-1) dividing a character string indicated by one input character string data (31) into a preceding first character string and a succeeding second character string; (B-2) setting the data indicating the first character string as first character string data (32); and (b-3) setting the data indicating the second character string as second character string data. Steps. The first character string has contents that supplementarily explain the contents indicated by the second character string.

  The step (b) of generating includes the step (b-4) of setting the input character string data (31) input first among the two input character string data (31) as the second character string data. And (b-5) the step of setting the input character string data (31) input later among the two input character string data (31) as the first character string data (32).

  The step (d) of determining is common between (d-1) reading of the word indicated by the plural homonym data (33) and reading of the character string indicated by the first character string data (32). Determining whether there are any readings. At this time, the determining step (e) includes (e-1) extracting a common part having the common reading from the character string indicated by the first character string data (32), and (e-2 ) Determining the homonym data (33) indicating the same word as the common part as the search word data (34).

  The step (d) of determining is common between the character string indicated by each homonym data (33) and the character string indicated by the first character string data (32). Determining whether there is a Chinese character. At this time, the step of determining (e) includes the step of determining (e-3) the homonym data (33) including the common kanji as the search word data (34).

  Further, in the step (d) of determining, (d-3) the dictionary data (61, 62) is referred to, the explanation of the word indicated by each homonym data (33), and the first character string data ( And 32) determining whether there is a common Chinese character between the character string indicated by 32). At this time, the step of determining (e) includes the step of determining (e-4) the homonym data (33) corresponding to the explanatory text including the common kanji as the search word data (34).

  According to the speech dictionary search system and the speech dictionary search program according to the present invention, it is possible to reduce erroneous searches due to homonyms in speech input type dictionary searches.

  According to the speech dictionary search system and the speech dictionary search program according to the present invention, it is possible to search for a desired word from a plurality of homophones based on a user's free wording.

  A speech dictionary search system and a speech dictionary search program according to the present invention will be described with reference to the accompanying drawings.

(First embodiment)
FIG. 1 is a block diagram showing the configuration of the speech dictionary search system according to the first embodiment of the present invention. In the present embodiment, the voice dictionary search system 10 includes a voice input unit 21, a voice recognition unit 22, a storage unit 30, a target determination unit 50, a dictionary search unit 55, a dictionary unit 60, and an output unit 70.

  The voice 20 is input to the voice input unit 21. An example of the voice input unit 21 is a microphone device. The voice input unit 21 converts the input voice 20 into voice data that is digital data, and outputs the voice data to the voice recognition unit 22.

  The voice recognition unit 22 is connected to the voice input unit 21 and receives voice data from the voice input unit 21. At this time, the voice recognition unit 22 performs sentence voice recognition generally called dictation, and converts the voice uttered by the user in a free sentence into a character string mixed with kana and kanji. That is, the voice recognition unit 22 generates the input character string data 31 from the received voice data using a known voice recognition algorithm (see Patent Document 1). The input character string data 31 is text data corresponding to the input voice 20, and the content of the voice 20 is indicated by a character string. The voice recognition unit 22 outputs the input character string data 31 to the storage unit 30.

  The storage unit 30 stores input character string data 31 output from the speech recognition unit 22. In addition, the storage unit 30 stores supplementary explanation data 32, homonym data 33, search word data 34, and an input text pattern 40 which will be described later. Examples of the storage unit 30 include a memory such as a flash memory and a DRAM. The dictionary unit 60 stores dictionary data. This dictionary data includes a group of word headings 61 and an explanation 62 of the words. The dictionary unit 60 is exemplified by a memory such as a ROM.

  The object determining unit 50 is connected to the storage unit 30 and the dictionary unit 60. The target determining unit 50 reads the input character string data 31 from the storage unit 30. Then, as will be described in detail later, the target determining unit 50 determines a “search target” to be a dictionary search target based on the input character string data 31. Here, the target determining unit 50 refers to the dictionary data stored in the dictionary unit 60. Then, the target determining unit 50 outputs the search term data 34 indicating the determined search target (heading 61) to the storage unit 30.

  The dictionary search unit 55 is connected to the storage unit 30 and reads the search word data 34 from the storage unit 30. The dictionary search unit 55 is connected to the dictionary unit 60 and performs a dictionary search on the search target (heading 61) indicated by the search word data 34 by referring to the dictionary data. Then, the dictionary search unit 55 outputs the result of the dictionary search (heading 61 and corresponding explanatory text 62) to the output unit 70.

  The output unit 70 is connected to the dictionary search unit 55 and notifies the user of the results of the dictionary search described above. As illustrated in FIG. 1, the output unit 70 includes a display unit 72 such as a liquid crystal display or an organic EL display. At this time, the search target (heading 61) and the explanatory text 62 corresponding thereto as a result of the dictionary search are displayed on the display unit 72. The user sees the search result displayed on the display unit 72 to know the meaning of the desired word, or subsequently inputs the voice 20 to the voice input unit 21.

  When a searcher looks up a word in a dictionary, the searcher first looks for a “headline” that matches the word. After finding the headline, the searcher reads the explanatory text corresponding to the headline. However, when a searcher searches for a word spoken by the speech input dictionary search system, it is necessary to consider the possibility that the word has a plurality of homonyms. For example, for the pronunciation word “Kyouka”, there are a plurality of homonyms such as “strengthening”, “subject”, “rage”, “doctorization”, and the like. Therefore, it is necessary to understand which word the searcher uttered is a plurality of homonyms.

  When selecting an intended word from a plurality of homonyms, a supplementary explanation is often added. Specifically, the intended word can be determined by phrases such as “strengthening to strengthen”, “bamboo bamboo and bamboo”, and “object that means goal”. In these examples, the intended words are clearly “strengthening”, “bamboo”, and “subject”. On the other hand, the supplementary explanations are the phrases “strengthen”, “shokutake plum” and “meaning the goal”, respectively. Each of the “intended word” and the “supplementary explanation” is hereinafter referred to as “target word” and “supplementary explanation”.

  In general, the wording used to distinguish homonyms is defined by the type of “supplementary explanation (first character string) + target word (second character string)”. The types of "supplementary explanation" are "adjective (phrase)", "verb (phrase)", "noun (phrase) + no", "noun (phrase) +" means "noun (phrase) +" Type of "". Examples of the “target word” include “noun (phrase)” and “adjective (phrase)”. The input text pattern 40 stored in the storage unit 30 described above is data relating to such a type of pattern (pattern). That is, the input text pattern 40 indicates all of the “supplemental explanation” pattern, the “target word” pattern, and the “supplemental explanation + target word” pattern.

  According to the speech dictionary search system 10 according to the present invention, the target determination unit 50 includes a pattern determination unit 51, a homonym extraction unit 52, and a search target determination unit 53.

  The pattern determination unit 51 reads the input character string data 31 and the input text pattern 40 from the storage unit 30. Then, the pattern determination unit 51 refers to the input text pattern 40 and executes a pattern matching process on the input character string data 31. Specifically, the pattern determination unit 51 determines whether or not the character string indicated by the input character string data 31 (hereinafter referred to as an input character string) matches the above-described “supplemental explanation + target word” pattern. To do. If they match, the pattern determination unit 51 divides the continuous input character string into a preceding “supplementary explanation” and a subsequent “target word”. Thereafter, the pattern determination unit 51 outputs the character string data indicating the supplementary explanation to the homonym extraction unit 52 as the supplementary explanation data 32. The pattern determination unit 51 also outputs character string data indicating the target word to the homonym extraction unit 52 as target word data. In this way, supplementary explanation data 32 and target word data are generated from at least one input character string data 31.

  The homonym extraction unit 52 receives the supplementary explanation data 32 and the target word data from the pattern determination unit 51. The homonym extraction unit 52 detects a plurality of homonyms of the target word indicated by the target word data. Here, the data stored in the dictionary unit 60, particularly the data indicating the heading 61, is referred to. Then, the homonym extraction unit 52 generates a plurality of homonyms data 33 indicating each of the detected plurality of synonyms, and outputs the homonyms data 33 to the storage unit 30. The homonym extraction unit 52 outputs the supplementary explanation data 32 to the storage unit 30.

  The search target determination unit 53 reads the supplementary explanation data 32 and the plurality of homophones data 33 from the storage unit 30. Thereafter, the search target determination unit 53 determines whether or not there is a “common point” between the characteristic of the supplementary explanation data 32 and each characteristic of the plural homophones data 33. The common points of this characteristic include “kanji” and “reading” that the supplementary explanation and the homonym have in common as will be described later. Then, the search target determination unit 53 selects one homonym having the “common point” from the plurality of homonyms, and stores the search word data 34 indicating the one homonym in the storage unit 30. Output.

  Next, the operation of the speech dictionary search system 10 according to the first exemplary embodiment of the present invention will be described in detail with reference to flowcharts and specific examples. 2A and 2B are flowcharts showing the operation of the speech dictionary search system 10 according to the present exemplary embodiment.

(Dictionary search example 1)
FIG. 3 is a diagram illustrating an example of the contents of the storage unit 30 in the present embodiment. That is, in FIG. 3, the contents of the input character string data 31, the supplementary explanation data 32, the homonym synonym data 33, and the search word data 34 are shown. In this example, the voice 20 “Kikagaku no Kyoka” is input to the voice input unit 21 (step S11). Then, the voice recognition unit 22 analyzes and recognizes voice data indicating this voice, and generates input character string data 31 that is text data (step S12). In this case, as shown in FIG. 3, it is assumed that the input character string data 31 “enhancement of training camp” is generated.

  Next, the pattern determination unit 51 refers to the input text pattern 40 and compares and examines the structure of the input character string that is the recognition result (step S13). That is, the pattern determination unit 51 determines whether or not the input character string matches the “supplemental explanation + target word” pattern. In this example, the input character string “strengthening training camp” matches the “noun + noun” pattern (step S14; Yes). In this case, the pattern determination unit 51 divides this input character string into a preceding supplementary explanation and a target word to be followed (step S15). That is, the input character string “strengthening training camp” is divided into “strengthening training camp” that is supplementary explanation and “strengthening” that is the target word. Thereafter, the pattern determination unit 51 outputs the supplementary explanation data 32 and the target word data indicating the character string “strengthening training camp” and the character string “strengthening” to the homonym extraction unit 52, respectively.

  Next, the homonym extraction unit 52 refers to the dictionary data stored in the dictionary unit 60 to detect a plurality of homonyms of the target word (step S16). For example, the homonym extraction unit 52 detects a plurality of homonyms such as “subject”, “rage”, and “teaching” for the target word “enhancement”. Thereafter, the homonym extraction unit 52 generates a plurality of homonym data 33 indicating each of the plurality of homonyms. The above-mentioned supplementary explanation data 32 (“strengthening training camp”) and a plurality of homophones data 33 (“subject”, “strengthening”, “rage”, “teaching”) are output to the storage unit 30 as shown in FIG. Stored in the storage unit 30.

  Next, the search target determination unit 53 compares each of the plurality of homonyms (“subject”, “strengthening”, “rage”, “doctoralization”) with the supplementary explanation “teaching camp” (step S21). First, the search target determination unit 53 determines whether the homonym and the supplementary explanation have a common “reading” (step S22). In this example, the supplementary explanation “strengthening training camp” has a portion having the same reading as “Kyoka”, which is a reading of the homonym (step S22; Yes). In this case, the search target determining unit 53 extracts the portion having the common reading in the supplementary explanation “strengthening training camp”, that is, the word “strengthening”. Thus, the word “enhancement” is adopted as the search target (step S23). That is, the search target determination unit 53 selects a desired word “enhancement” from a plurality of homonyms (“subject”, “enhancement”, “rage”, “education”), and search word data 34 indicating the word “enhancement”. Is output to the storage unit 30. The display unit 72 displays the search result (step S30). In this way, the user is notified of the word intended by the user and the meaning of the word.

(Dictionary search example 2)
FIG. 4 is a diagram showing another example of the contents of the storage unit 30 in the present embodiment. In this example, the voice 20 “Kyoto Kyouka” is input to the voice input unit 21 (step S11). Then, the voice recognition unit 22 analyzes and recognizes voice data indicating this voice, and generates input character string data 31 that is text data (step S12). In this case, as shown in FIG. 4, it is assumed that input character string data 31 of “strengthening subject” is generated. Here, the word that the user intends to search is clearly “strengthening”.

  Next, the pattern determination unit 51 refers to the input text pattern 40 and compares and examines the structure of the input character string that is the recognition result (step S13). In this example, the input character string “strengthening subject” matches the “verb + noun” pattern (step S14; Yes). In this case, the pattern determination unit 51 divides this input character string into a preceding supplementary explanation and a target word to be followed (step S15). In other words, the input character string “Strengthen subject” is divided into a supplementary explanation “Strengthen” and a target word “Subject”.

  Next, the homonym extraction unit 52 refers to the dictionary data stored in the dictionary unit 60 to detect a plurality of homonyms of the target word (step S16). For example, the homonym extraction unit 52 detects a plurality of homonyms such as “enhancement”, “rage”, and “teaching” for the target word “subject”. Thereafter, the homonym extraction unit 52 generates a plurality of homonym data 33 indicating each of the plurality of homonyms. The supplementary explanation data 32 (“strengthen”) and a plurality of homonyms data 33 (“subject”, “enhancement”, “rage”, “teaching”) are output to the storage unit 30 and stored as shown in FIG. Stored in the unit 30.

  Next, the search target determination unit 53 compares each of a plurality of homonyms (“subject”, “enhancement”, “rage”, “doctorization”) with the supplementary explanation “strengthen” (step S21). First, the search target determination unit 53 determines whether the homonym and the supplementary explanation have a common reading (step S22). In this example, there is no common reading (step S22; No). Next, the search target determination unit 53 determines whether the homonym and the supplementary explanation have a common kanji (step 24). “Strengthening” and the supplementary explanation “strengthen” in the plural homonyms have a common Chinese character “strong” (step S24; Yes). In this case, the word “enhancement” having the common Chinese character is adopted as a search target (step S25). That is, the search target determination unit 53 selects a desired word “enhancement” from a plurality of homonyms (“subject”, “enhancement”, “rage”, “education”), and search word data 34 indicating the word “enhancement”. Is output to the storage unit 30. The display unit 72 displays the search result (step S30). In this way, the user is notified of the word intended by the user and the meaning of the word.

(Dictionary search example 3)
FIG. 5 is a diagram illustrating another example of the contents of the storage unit 30 in the present embodiment. In this example, the voice 20 “Gakkou no Kyoka” is input to the voice input unit 21 (step S11). Then, the voice recognition unit 22 analyzes and recognizes voice data indicating this voice, and generates input character string data 31 that is text data (step S12). In this case, it is assumed that the input character string data 31 “school subject” is generated as shown in FIG.

  Next, the pattern determination unit 51 refers to the input text pattern 40 and compares and examines the structure of the input character string that is the recognition result (step S13). In this example, the input character string “school subject” matches the “noun + noun” pattern (step S14; Yes). In this case, the pattern determination unit 51 divides this input character string into a preceding supplementary explanation and a target word to be followed (step S15). That is, the input character string “school subject” is divided into “school” as a supplementary explanation and “subject” as a target word.

  Next, the homonym extraction unit 52 refers to the dictionary data stored in the dictionary unit 60 to detect a plurality of homonyms of the target word (step S16). For example, the homonym extraction unit 52 detects a plurality of homonyms such as “enhancement”, “rage”, and “teaching” for the target word “subject”. Thereafter, the homonym extraction unit 52 generates a plurality of homonym data 33 indicating each of the plurality of homonyms. The supplementary explanation data 32 (“school”) and a plurality of homonym data 33 (“subject”, “enhancement”, “rage”, “teaching”) are output to the storage unit 30 and, as shown in FIG. 30.

  Next, the search target determination unit 53 compares each of the plurality of homonyms (“subject”, “enhancement”, “rage”, “education”) with the supplementary explanation “school” (step S21). First, the search target determination unit 53 determines whether the homonym and the supplementary explanation have a common reading (step S22). In this example, there is no common reading (step S22; No). Next, the search target determination unit 53 determines whether the homonym and the supplementary explanation have a common kanji (step 24). In this example, there is no common kanji (step S24; No). Next, the search target determination unit 53 compares each explanatory sentence of the plural homonyms (“subject”, “enhancement”, “rage”, “education”) with the supplementary explanation “school” (step S26). Here, the dictionary data stored in the dictionary unit 60, that is, the heading 61 and the explanation 62 are referred to.

  FIG. 6 is a diagram showing an example of the contents of the dictionary unit 60 in the speech dictionary search system according to the present invention. In FIG. 6, for example, a heading “enhancement” and its explanation 62 and a heading “subject” and its explanation 62 are shown.

  With reference to such a dictionary, the search target determination unit 53 determines whether or not the explanation sentence 62 of each homonym and the supplementary explanation have a common part (step S27). As shown in FIG. 6, the description 62 of the heading “Department” includes the phrase “school education process”. That is, the explanatory note 62 of the word “subject” and the supplementary explanation “school” have a common part “school” (step S27; Yes). In this case, the word “subject” having the common part “school” is adopted as a search target (step S28). That is, the search target determination unit 53 selects a desired word “subject” from a plurality of homonyms (“subject”, “enhancement”, “rage”, and “teaching”), and the search term data 34 indicating the word “subject”. Is output to the storage unit 30. The display unit 72 displays the search result (step S30). In this way, the user is notified of the word intended by the user and the meaning of the word.

  If there is no common part between the explanation 62 of each homonym and the supplementary explanation (step S27; No), the first candidate is adopted (step S29). That is, the first obtained target word is adopted as a search target. The search target determination unit 53 outputs the search word data 34 indicating the target word to the storage unit 30.

  In the above search examples, when a plurality of types of homonyms include a plurality of candidates having a common part (kanji or reading) with the supplementary explanation, the search target determination unit 53 includes the most common part. Candidates may be selected as dictionary search targets. Alternatively, the search target determination unit 53 may output the search term data 34 indicating the plurality of candidates to the storage unit 30. In this case, a plurality of types of homonyms are displayed on the display unit 72. Based on this result, the user can input a new voice 20 to the voice input unit 21. If the input character string does not match the “supplementary explanation + target word” pattern in step S14 (step S14; No), the recognition result (input character string) is displayed as it is on the display unit 72 (step S17). ).

  The dictionary search process in the speech dictionary search system 10 shown above may be executed by a computer. At this time, the processes in the speech recognition unit 22, the pattern determination unit 51, the homonym extraction unit 52, the search target determination unit 53, and the dictionary search unit 55 are each described by a computer program, and an arithmetic processing unit (not shown) ) Is executed.

  Further, in the speech dictionary search system 10, the output unit 70 described above may include a speech output unit. The voice output unit generates a synthesized voice from the character string and outputs the synthesized voice. At this time, the word heading 61 and the explanation 62 thereof as a search result are notified to the user by synthetic speech. The output unit 70 may output the search result to another system. For example, the output unit 70 may output the search result to the voice reservation system. In this case, the user inputs, for example, the kanji of his / her name freely explaining each character. The output unit 70 sequentially outputs the kanji (heading 61) indicated by the search word data 34 to the voice reservation system.

  As described above, according to the speech dictionary search system 10 according to the present embodiment, the user's intention embedded in the speech 20 (input character string data 31) is recognized. Specifically, a “target word” and a “supplemental explanation” that supplementarily explains the “target word” are extracted from the input character string data 31. By referring to this “supplementary explanation”, an appropriate word considering the user's intention is searched in the dictionary. Here, there are no particular restrictions on the words and phrases to be used in this “supplementary explanation”. The user may describe the “target word” using an arbitrary word / phrase so that the user can easily explain. The speech dictionary search system 10 refers to the “kanji” and “reading” included in the “supplementary explanation”, so that the word intended by the user can be obtained even when the input word has a plurality of homonyms. You can search. As described above, according to the speech dictionary search system 10 according to the present invention, erroneous searches due to homonyms are reduced, and it is possible to efficiently perform a dictionary search on input speech.

(Second embodiment)
When performing a dictionary search by inputting voice, the user can also say “Balance inquiry” as a supplementary explanation after saying “Shokai”, for example. In other words, after the user utters the word to be searched, the user can add the explanation and utter. The second embodiment of the present invention corresponds to such voice input. According to the speech dictionary search system 10 according to the present embodiment, the operation of the pattern determination unit 51 is different from the operation in the first embodiment. Other configurations and operations are the same as those in the first embodiment (see FIG. 1), and the description thereof is omitted as appropriate.

  FIG. 7 is a flowchart showing the operation of the speech dictionary search system 10 according to the present exemplary embodiment. 8A and 8B are diagrams illustrating an example of the contents of the storage unit 30 in the present embodiment. That is, in FIG. 8A and FIG. 8B, the contents of the input character string data 31, the supplementary explanation data 32, the homonym synonym data 33, and the search word data 34 are shown. In this example, the voice 20 “Shokai” is input to the voice input unit 21 (step S31). Then, the voice recognition unit 22 analyzes and recognizes voice data indicating this voice, and generates input character string data 31 that is text data (step S32). In this case, it is assumed that the input character string data 31 “introduction” is generated as shown in FIG. 8A.

  Next, the pattern determination unit 51 checks whether the target word has already been generated and whether the homonym data 33 is registered in the storage unit 30 (step S33). When the homonym data 33 is not yet registered in the storage unit 30 (step S33; No), the pattern determination unit 51 sets this input character string as a “target word” (step S37). That is, the character string “introduction” is registered as the target word. Then, the pattern determination unit 51 outputs target word data indicating the target word to the homonym extraction unit 52.

  Similar to the first embodiment, the homonym extraction unit 52 detects a plurality of homonyms of the target word (step S38). For example, the homonym extraction unit 52 detects a plurality of homonyms, such as “inquiry”, “commercial company”, and “small party”, for the target word “introduction”. Thereafter, the homonym extraction unit 52 generates a plurality of homonym data 33 indicating each of the plurality of homonyms. The plurality of homonym data 33 (“introduction”, “inquiry”, “commercial company”, “sub-committee”) are output to the storage unit 30 and stored in the storage unit 30 as shown in FIG. 8A.

  Since the supplementary explanation data 32 has not yet been generated, the search target determination unit 53 outputs the homonym word data 33 as it is to the storage unit 30 as the search word data 34. Alternatively, the search target determination unit 53 outputs data indicating “first introduction”, which is the first candidate, to the storage unit 30 as the search term data 34. The dictionary search unit 55 performs a dictionary search based on the search word data 34, and the display unit 72 displays the result of the dictionary search (step S39). That is, the display unit 72 displays a plurality of homonyms (“introduction”, “inquiry”, “commercial company”, and “subcommittee”) or the word “introduction”.

  The user can add a supplementary explanation based on the display result. For example, when the user wants to search for the word “inquiry”, the user may utter “Zanda Kasho” as the second utterance (step S31). At this time, as shown in FIG. 8B, it is assumed that the input character string data 31 "balance inquiry" is generated (step S32). After that, as in the previous time, the pattern determination unit 51 confirms whether the target word has already been generated and whether the homonym word data 33 is registered in the storage unit 30 (step S33).

  This time, the homonym data 33 has already been registered in the storage unit 30 (step S33; Yes). In this case, the pattern determination unit 51 refers to the input text pattern 40 and compares and examines the structure of the input character string that is the recognition result (step S34). That is, the pattern determination unit 51 determines whether or not the input character string matches the above-described “supplemental explanation” pattern.

  When the pattern does not match the supplementary description pattern (step S35; No), the pattern determination unit 51 sets the input character string as a “target word” (step S37). If the pattern matches the supplementary explanation pattern (step S35; Yes), the pattern determination unit 51 sets this input character string as “supplementary explanation” (step S36). That is, the character string “Balance inquiry” is registered as the supplementary explanation. Then, supplementary explanation data 32 indicating this supplementary explanation is output to the storage unit 30. At this time, as shown in FIG. 8B, the storage unit 30 stores supplementary explanation data 32 (“balance inquiry”) and a plurality of synonym data 33 (“introduction”, “inquiry”, “company”, “subcommittee”). Is stored.

  Thereafter, the search target determination unit 53 uses a method similar to the method in the first embodiment (steps S21 to S29; see FIG. 2B) to generate a plurality of homonyms (“introduction”, “inquiry”, “company”, “small”). The desired word “inquiry” is selected from “Meeting”). That is, the user's intention is recognized by searching for common kanji and readings. Then, as illustrated in FIG. 8B, the search target determination unit 53 outputs the search term data 34 indicating the word “inquiry” to the storage unit 30. Finally, the display unit 72 displays the search result (step S30). In this way, the user is notified of the word intended by the user and the meaning of the word.

  The dictionary search process in the speech dictionary search system 10 shown above may be executed by a computer. At this time, the processes in the speech recognition unit 22, the pattern determination unit 51, the homonym extraction unit 52, the search target determination unit 53, and the dictionary search unit 55 are each described by a computer program, and an arithmetic processing unit (not shown) ) Is executed.

  As described above, according to the speech dictionary search system 10 according to the second exemplary embodiment of the present invention, the pattern determination unit 51 is the one of the two input character string data 31 that has been input first. Is set as the target word data (synonym synonym data 33), and the later input data is set as the supplementary explanation data 32. As a result, for example, even when the user utters “Shokai” and then utters “Zanda Kashokai” as a supplementary explanation, the desired word “inquiry” is selected as a search target from a plurality of homonyms. . Thus, according to the present embodiment, the same effect as that of the first embodiment can be obtained.

  It should be noted that the first embodiment and the second embodiment can be used in combination. At this time, when it is determined in step S14 shown in FIG. 2A that the input character string does not match the “supplemental explanation + target word” pattern (step S14; No), the pattern determination unit 51 is shown in FIG. Step S33 is executed. That is, the pattern determination unit 51 checks whether the target word has already been generated and whether the homonym data 33 has been registered in the storage unit 30 (step S33). The operations after Step S14 and Step S33 are the same as the operations in the first and second embodiments, respectively. When the supplementary explanation data 32 and the homonym data 33 are registered in the storage unit 30, the operation shown in FIG. 2B is executed.

(Third embodiment)
FIG. 9 is a block diagram showing the configuration of the speech dictionary search system according to the third embodiment of the present invention. 9, the same components as those shown in FIG. 1 are denoted by the same reference numerals, and description thereof will be omitted as appropriate. In the present embodiment, the speech dictionary search system 10 ′ includes an input in addition to the speech input unit 21, speech recognition unit 22, storage unit 30, object determination unit 50, dictionary search unit 55, dictionary unit 60, and output unit 70. Unit 81 and control unit 82.

  The input unit 81 includes a plurality of buttons. The plurality of buttons include a supplement button that is pressed when “supplementary explanation” is input, and an input button that is pressed when “target word” is input. The plurality of buttons (81) are connected to the voice recognition unit 22 and the pattern determination unit 51 via the control unit 82.

  The control unit 82 controls the voice recognition unit 22 to execute the voice recognition process only when the input button or the supplement button is pressed. The control unit 82 controls the operation of the pattern determination unit 51 in accordance with the pressing state of the input button or the supplement button. That is, the pattern determination unit 51 sets the input character string data 31 input when the supplement button is pressed as the supplementary explanation data 32. Further, the pattern determination unit 51 sets the input character string data 31 input when the input button is pressed as the target word data (synonym synonym data 33). The subsequent operation is the same as the operation in the described embodiment (see FIG. 2B). That is, the user's intention is recognized by searching for common kanji and readings. Thus, according to the speech dictionary search system 10 ′ according to the present embodiment, the convenience for the user is improved.

(Fourth embodiment)
FIG. 10 is a block diagram showing the configuration of the speech dictionary search system according to the fourth embodiment of the present invention. 10, the same components as those shown in FIG. 1 are denoted by the same reference numerals, and description thereof will be omitted as appropriate. In the present embodiment, the speech dictionary search system 10 performs a dictionary search by the same method as in the first and second embodiments described above.

  The speech dictionary search system 10 according to the present embodiment further includes a user auxiliary unit 90 connected to the target determining unit 50 and the output unit 70. The target determining unit 50 transmits an auxiliary command signal to the user auxiliary unit 90 when the “search target” cannot be specified. In accordance with the auxiliary command signal, the user auxiliary unit 90 guides or guides the utterance to the user or prompts the next utterance. The guidance / guidance / promotion of these utterances is notified to the user through the output unit 70 (display unit 72).

  For example, in step S14 according to the first embodiment (see FIG. 2A), when the input character string that is the recognition result does not match the “supplementary explanation + target word” pattern (step S14; No), the display unit 72 The recognition result is displayed (step S17), and a phrase indicating guidance / guidance / promotion of utterance is displayed. Specifically, phrases such as “Please speak again”, phrases such as “Please speak in other words”, and “Supplementary explanation + object” patterns are examples: Is displayed on the display unit 72.

  In addition, for example, in step S39 according to the second embodiment (see FIG. 7), a plurality of homonyms are displayed on the display unit 72, and phrases indicating guidance / guidance / promotion of utterance are displayed. . Specifically, a phrase such as “a homonym has been detected” or a phrase “please input supplemental explanation” is displayed on the display unit 72.

  Further, for example, in the first and second embodiments, when the search target determination unit 53 cannot identify “search target” (step S29; see FIG. 2B), the first candidate word is displayed. A phrase indicating guidance / guidance / promotion of utterance is displayed. Specifically, a phrase such as “a homonym has been detected”, a phrase such as “Please speak with another phrase”, or a phrase such as “Please input supplemental explanation again” is displayed on the display unit 72. Is displayed.

  The user can utter again or input a supplementary explanation by looking at the notification displayed on the display unit 72. Thus, according to the speech dictionary search system 10 according to the present embodiment, user convenience is improved.

FIG. 1 is a block diagram showing the configuration of a speech dictionary search system according to the present invention. FIG. 2A is a flowchart showing the operation of the speech dictionary search system according to the first embodiment of the present invention. FIG. 2B is a flowchart showing the operation of the speech dictionary search system according to the first embodiment of the present invention. FIG. 3 is a diagram showing an example of the contents of the storage unit in the speech dictionary search system according to the first embodiment of the present invention. FIG. 4 is a diagram showing another example of the contents of the storage unit in the speech dictionary search system according to the first embodiment of the present invention. FIG. 5 is a diagram showing still another example of the contents of the storage unit in the speech dictionary search system according to the first embodiment of the present invention. FIG. 6 is a diagram showing an example of the contents of the dictionary unit in the speech dictionary search system according to the present invention. FIG. 7 is a flowchart showing the operation of the speech dictionary search system according to the second embodiment of the present invention. FIG. 8A is a diagram showing an example of the contents of the storage unit in the speech dictionary search system according to the second embodiment of the present invention. FIG. 8B is a diagram showing an example of the contents of the storage unit in the speech dictionary search system according to the second embodiment of the present invention. FIG. 9 is a block diagram showing the configuration of the speech dictionary search system according to the third embodiment of the present invention. FIG. 10 is a block diagram showing the configuration of the speech dictionary search system according to the fourth embodiment of the present invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 Voice dictionary search system 20 Voice 21 Voice input part 22 Voice recognition part 30 Storage part 31 Input character string data 32 Supplementary explanation data 33 Homophones synonym data 34 Search word data 40 Input text pattern 50 Target determination part 51 Pattern determination part 52 Anomaly extraction unit 53 Search target determination unit 55 Dictionary search unit 60 Dictionary unit 61 Heading 62 Description 70 Output unit 72 Display unit 81 Input unit 82 Control unit 90 User assistance unit

Claims (16)

An audio input unit for inputting audio;
A dictionary section for storing dictionary data;
A voice recognition unit that generates input character string data corresponding to the input voice;
A target determination unit that receives the input character string data and determines a search target with reference to the dictionary data;
A dictionary search unit that performs a dictionary search by referring to the dictionary data for the determined search target;
An output unit for outputting the result of the dictionary search,
The target determining unit
A first determination unit;
A homonym extraction unit;
A second determination unit, and
The first determination unit generates first character string data and second character string data from at least one of the input character string data,
The homonym synonym extraction unit detects a plurality of homonyms of the word indicated by the second character string data with reference to the dictionary data, and a plurality of homonym data indicating each of the plurality of homonyms Produces
The second determination unit determines whether there is a common point between the characteristic of each of the plurality of homophones data and the characteristic of the first character string data, and among the plurality of homophones data The speech dictionary search system that outputs the homonym data having one common point to the dictionary search unit as search word data indicating the search target. The speech dictionary search system according to claim 1,
The first determination unit divides a character string indicated by one input character string data into a first character string preceding and a second character string following.
The spoken dictionary search system, wherein the first determination unit sets data indicating the first character string as the first character string data, and sets data indicating the second character string as the second character string data. The speech dictionary search system according to claim 2,
The first character string includes at least one of a noun, a noun phrase, an adjective, an adjective phrase, a verb, and a verb phrase,
The second character string includes a noun or a noun phrase. The speech dictionary search system according to claim 1,
The first determination unit sets the input character string data input first among the two input character string data as the second character string data, and sets the input character string data input later as the second character string data. Spoken dictionary search system set as one character string data. The speech dictionary search system according to claim 4,
The character string indicated by the first character string data includes at least one of a noun, a noun phrase, an adjective, an adjective phrase, a verb, and a verb phrase.
The character string indicated by the second character string data includes a noun or a noun phrase. The speech dictionary search system according to claim 1,
A first button and a second button connected to the first determination unit;
The first determination unit sets the input character string data input when the first button is pressed as the first character string data, and the input character input when the second button is pressed. A speech dictionary search system for setting column data as the second character string data. The speech dictionary search system according to any one of claims 1 to 6,
The second determination unit determines whether or not there is a common reading between the reading of the word indicated by the plurality of homonym data and the reading of the character string indicated by the first character string data,
The second determination unit extracts a common part having the common reading from the character string indicated by the first character string data, and the homonym data indicating the same word as the common part is Voice dictionary search system that outputs as search term data. The speech dictionary search system according to any one of claims 1 to 6,
The second determination unit determines whether there is a common kanji character between the character string indicated by each of the homonym data and the character string indicated by the first character string data;
The second determination unit outputs the homonym word data including the common kanji as the search word data. The speech dictionary search system according to any one of claims 1 to 6,
The second determination unit refers to the dictionary data, and whether there is a common kanji character between the explanatory text of the word indicated by each homonym data and the character string indicated by the first character string data Determine whether
The said 2nd determination part outputs the said homonym word data corresponding to the said explanatory note containing the said common kanji as a search word data. The speech dictionary search system according to any one of claims 1 to 9,
A user assistance unit connected to the target determination unit and the output unit;
When the target determining unit cannot identify a search target, the user auxiliary unit notifies the user of guidance / guidance / promotion of utterance through the output unit. In a speech dictionary search system that performs dictionary search with reference to dictionary data for input speech,
(A) generating input character string data corresponding to the input voice;
(B) generating first character string data and second character string data from at least one of the input character string data;
(C) detecting a plurality of homophones of the word indicated by the second character string data with reference to the dictionary data, and generating a plurality of homophones data indicating each of the plurality of homophones; ,
(D) determining whether there is a common point between the characteristics of each of the plurality of homonym data and the characteristics of the first character string data;
(E) determining one homonym term data having the common point among the plurality of homonym term data as search term data indicating the object of the dictionary search;
(F) A speech dictionary search program for causing a computer to execute the step of performing the dictionary search on the target indicated by the search word data with reference to the dictionary data. The speech dictionary search program according to claim 11,
The step (b) of generating includes
(B-1) dividing a character string indicated by one input character string data into a first character string preceding and a second character string following;
(B-2) setting data indicating the first character string as the first character string data;
(B-3) A speech dictionary search program, comprising: setting data indicating the second character string as the second character string data. The speech dictionary search program according to claim 11,
The step (b) of generating includes
(B-4) setting the input character string data input first among the two input character string data as the second character string data;
(B-5) A speech dictionary search program including the step of setting the input character string data input later among the two input character string data as the first character string data. The speech dictionary search program according to any one of claims 11 to 13,
The step (d) of determining includes
(D-1) including a step of determining whether or not there is a common reading between the reading of the word indicated by the plurality of homonym data and the reading of the character string indicated by the first character string data;
The step (e) of determining includes
(E-1) extracting a common part having the common reading from the character string indicated by the first character string data;
(E-2) A speech dictionary search program comprising: determining the homonym data indicating the same word as the common part as the search word data. The speech dictionary search program according to any one of claims 11 to 13,
The step (d) of determining includes
(D-2) including a step of determining whether or not there is a common Chinese character between the character string indicated by each of the homonym data and the character string indicated by the first character string data;
The step (e) of determining includes
(E-3) A speech dictionary search program including a step of determining the homonym data including the common kanji as the search word data. The speech dictionary inspection program according to any one of claims 11 to 13,
The step (d) of determining includes
(D-3) Referring to the dictionary data, whether or not there is a common kanji character between the explanatory text of the word indicated by each homonym data and the character string indicated by the first character string data Including the step of determining,
The step (e) of determining includes
(E-4) A speech dictionary search program including a step of determining, as the search word data, the homonym data corresponding to the explanatory text including the common kanji.

Images