JP6109664B2 - Program, apparatus and method for estimating specific feelings for synonyms between language systems - Google Patents

Description

  The present invention relates to a technology of a dialogue system interposed between two language systems.

  Conventionally, there is a communication system technology for automatically interpreting or translating in order to support a dialogue between two language systems. According to this technique, a bilingual server is interposed between at least two terminals, and a text input from one terminal can be translated using a dictionary database and output to the other terminal.

  For example, there is a technique for supporting communication by estimating a difference in values from the other party and an affirmation level for a conversation topic (see, for example, Patent Document 1). According to this technique, the strength of agreement and disagreement of a phrase used for replying a conversation is estimated. In particular, the degree of divergence of opinions (values) is detected from the polarity of responses (degrees of affirmation / denial) to the conversation partner's statements. In addition, there is a technology that absorbs the cultural background of the source language by the culture of the target language and accurately translates the expression expressed in the source language into the expression of the target language (see, for example, Patent Document 2). ).

JP 2006-209332 A Japanese Patent Laid-Open No. 06-348751

  When talking with two or more people, there are various differences in each person's native language, growing environment, and past experience. It does not matter whether they are face-to-face / not face-to-face, real-time conversation / mail. As long as language is used, there is often a gap (difference) in the recognition of objects pointed to by the same word or clause / sentence.

  In general, if the language systems are different, even if they are synonymous words through translation, the images pointed to reflect the differences in history and culture, and it is likely to give a different impression. For example, Japanese “castle” and European “castle” have different impressions. Also, even those who talk in the same national language may be subject to the same language due to differences in regional characteristics (for example, dialects) or individual attributes (gender, age, education ...). There is a gap in recognition as well. However, if dialogues are advanced without fully understanding these different impressions, the communication itself is often frustrated.

  According to the existing parallel translation system, it is possible to obtain a translation result that is synonymous in terms of syntax and semantics and that is correct by referring to the dictionary database. However, there is no guarantee that the listener receiving it will have a similar image to the subject of the topic. This is because there is a culture-specific mapping relationship between the language and the instruction target.

  FIG. 1 is an explanatory diagram illustrating an example in which a native language is generated in communication between interlocutors in a conventional bilingual system.

  According to FIG. 1, a Japanese user A and a German user B are interacting with each other using a terminal 2 and a bilingual system. The bilingual system converts, for example, Japanese speech received from a Japanese terminal 2A into Japanese text by speech recognition, translates the Japanese text into German text, and translates the German text into German by speech synthesis. The voice is converted into a word voice, and the German voice is transmitted to the German terminal 2B. On the other hand, German speech received from the German terminal 2B is converted into German text by speech recognition, the German text is translated into Japanese text, and the Japanese text is translated into Japanese by speech synthesis. The voice is converted to voice and the Japanese voice is transmitted to the Japanese terminal 2A. The following describes examples of communication gaps that may arise between Japanese and German.

(S11) Assume that German B utters “Mahlzeit!”, Which is a greeting at noon, to terminal 2B. This voice is translated through the system and transferred to the terminal 2A. Here, in a general German-Japanese dictionary, it is translated as “Mahlzeit”-> “When eating”.
(S12) Japanese A utters “At the time of a meal!” From the terminal 2A, and is confused as “Separately at the time of a meal! For Germans, this is just a general daytime greeting, but the Japanese cannot understand it, creating a communication gap.

(S13) Next, it is assumed that the Japanese A utters “Yes, it's about noon” to the terminal 2A. This voice is translated through the system and transferred to the terminal 2B. Here, in a typical Japanese-German dictionary, it is translated as "Yes, it's about time"->"Ja, richtig! Es ist bald Mittag."
(S14) German B is uttered “Ja, richtig! Es ist bald Mittag.” From terminal 2B and is confused as “I just gave a greeting at noon…?”. For Germans, it is difficult to understand why such a response has occurred, and a communication gap has arisen.

(S15) Next, assume that Japanese A sneezes. Japanese A thinks, “I feel cold now, but ...” Assume that the sneezing voice of Japanese A is transferred from terminal 2A to terminal 2B via the network and is heard by German B as it is.
(S16) Assume that German B utters “Gasundheit!”, Which means “precious” to terminal 2B. This voice is translated through the system and transferred to the terminal 2A. Here, in a general German-Japanese dictionary, it is translated as “Gasundheit”-> “health”.
(S17) Japanese A is uttered “Healthy!” From terminal 2A and is confused as “???”. For the Germans, they only worked hard, but the Japanese could not understand it, creating a communication gap.

  A typical example of such a communication gap is based on the fact that the listener cannot imagine the existence of a culture-dependent usage on the speaker's side simply by lexicographically translating specific words in the speaker's language system. it is conceivable that. On the other hand, if the result of the translation is translated, there is a transmission intention that exceeds that, and a further gap is generated.

  Here, can the inventors of the present application reduce a communication gap caused by a general translation using a dictionary? I thought. In other words, I thought that it would be sufficient for the listener to be able to estimate the emotional difference for the translated phrases between language systems. Note that none of the techniques described in Patent Documents 1 and 2 described above estimate a communication gap caused by a difference in emotion related to language.

  Then, an object of this invention is to provide the program, apparatus, and method which estimate the specific emotion with respect to a synonym phrase between language systems.

According to the present invention, when the first word / phrase of the first language system and the second word / phrase of the second language system are in a parallel translation relationship or a correspondence relationship, the user who contacts the second word / phrase is specified. A program that causes a computer installed in the device to function as if it is estimated whether to recognize emotions,
Corpus collection and storage means including a number of reference sentences based on the second language system;
Corpus search means for searching for a reference sentence including the second word using the corpus collection and storage means;
A co-occurrence frequency calculating means for calculating a co-occurrence frequency between the first word and the specific emotion word for the retrieved reference sentence;
Co-occurrence frequency for a particular emotion word to be detected is, cause the computer to function as a specific emotion determination means for determining whether or Jo Tokoro threshold higher,
The second word / phrase determined to be true by the specific emotion determination means makes the computer function so as to presume that the user recognizes the specific emotion.
According to another embodiment of the program of the present invention,
The co-occurrence frequency calculating means further calculates a co-occurrence frequency between the first word and each of the plurality of emotion words defined in advance for the retrieved reference text,
The specific emotion determination means determines whether the co-occurrence frequency for the specific emotion word to be detected is higher than a co-occurrence frequency of other emotion words by a predetermined threshold or more.
It is also preferable to make the computer function.

According to another embodiment of the program of the present invention,
It is also preferable to cause the computer to function so that the first language system and the second language system are in the same national language and have a corresponding relationship.

According to another embodiment of the program of the present invention,
The corpus search means preferably causes the computer to function to search for a reference sentence including the first word and the second word.

According to another embodiment of the program of the present invention,
A specific emotion word is a noun or a verb derived from it,
It is also preferable that the computer function so that the plurality of emotion words include at least the specific emotion word.

According to another embodiment of the program of the present invention,
It is also preferable to make the computer function so that the specific emotion words are the nouns “confused”, “embarrassed”, “question” or the verbs “prone”, “confused”, “don't know”.

According to another embodiment of the program of the present invention,
The co-occurrence frequency calculating means preferably causes the computer to function so as to calculate the co-occurrence frequency for each of the first word and each emotion word based on the mutual information criterion.

According to another embodiment of the program of the present invention,
It is also preferable to cause the computer to function so that the text of the second language system stored in the corpus collection storage means is text published by a Web server, a blog server, or an SNS (Social Networking Service) server.

According to another embodiment of the program of the present invention,
Alarm phrase storage means for storing the first and second phrases determined to be true by the specific emotion determination means;
As specific emotion clarification means for clearly indicating to the user who is in contact with the second phrase that the translation is based on the specific emotion word when the parallel translation relationship and the correspondence relationship between the first phrase and the second phrase are used. It is also preferable to make the computer function.

According to another embodiment of the program of the present invention,
The specific emotion clarification means causes the computer to function as a comment to the user, with a part or all of a sentence of the second language system searched by the corpus search means or a link destination address as an annotation together with the second word / phrase. Is also preferable.

According to the present invention, when the first word / phrase of the first language system and the second word / phrase of the second language system are in a parallel translation relationship or a correspondence relationship, the user who contacts the second word / phrase is specified. A device for estimating whether or not to recognize emotions,
Corpus collection and storage means including a number of reference sentences based on the second language system;
Corpus search means for searching for a reference sentence including the second word using the corpus collection and storage means;
A co-occurrence frequency calculating means for calculating a co-occurrence frequency between the first word and the specific emotion word for the retrieved reference sentence;
Co-occurrence frequency for a particular emotion word to be detected is, and a specific emotion determination means for determining whether or Jo Tokoro threshold higher,
The second word / phrase determined to be true by the specific emotion determination means is estimated to cause the user to recognize the specific emotion.

According to the present invention, when the first word / phrase of the first language system and the second word / phrase of the second language system are in a parallel translation relationship or a correspondence relationship, the user who contacts the second word / phrase is specified. A specific emotion estimation method for a device that estimates whether or not to recognize emotions,
The apparatus has a corpus collection and accumulation unit including a large number of reference sentences based on the second language system,
The device
A first step of searching for a reference sentence including a second word using a corpus collection and storage unit;
A second step of calculating a co-occurrence frequency between the first phrase and the specific emotion word for the retrieved reference sentence;
Co-occurrence frequency for a particular emotion word to be detected is, performs a third step of determining whether or Jo Tokoro threshold higher,
The second word / phrase determined to be true in the third step is estimated to cause the user to recognize a specific emotion.

  According to the program, apparatus, and method of the present invention, it is possible to estimate a specific emotion for a synonym phrase between language systems.

It is explanatory drawing showing the example in which a native language has arisen in communication between the dialog persons about the dialog system in a prior art. It is a sequence diagram of a system in the present invention. It is an example of the screen displayed on the terminal in this invention. It is a function block diagram of the parallel translation server in this invention. It is explanatory drawing of operation | movement of the co-occurrence frequency calculation part in this invention.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

  The present invention can automatically detect a phrase that is likely to cause a communication gap between language systems, and can explicitly display a speaker's intention meant by the phrase on a listener's terminal. By allowing the listener to understand the intention of the speaker, the communication gap that occurs on the listener can be reduced.

  First, in order to clarify the operational effects of the present invention, a description will be given in comparison with the sequence in the prior art of FIG.

FIG. 2 is a sequence diagram of the system according to the present invention.
FIG. 3 is an example of a screen displayed on the terminal according to the present invention.

In FIG. 3, the sequence numbers are described in correspondence with those in FIG.
(S11) Assume that German B utters “Mahlzeit!”, Which is a greeting at noon, to terminal 2B. This voice is translated through the system and transferred to the terminal 2A. Here, in a general German-Japanese dictionary, it is translated as “Mahlzeit”-> “When eating”.
(S12) Japanese A utters “At the time of meal!” From terminal 2A. At this time, according to the present invention, “* Note light greeting during lunch” is displayed on the display of the terminal 2A (see S12 in FIG. 3). As a Japanese A, you can understand that it is.

(S13) Next, it is assumed that Japanese A also utters “at the time of meal” to terminal 2A. This voice is translated through the system and transferred to the terminal 2B.
(S14) German B is uttered “Mahlzeit!” From terminal 2B, and German B does not feel any sense of incongruity.

(S15) Next, assume that Japanese A sneezes. Japanese A thinks, “I feel cold now, but ...” Assume that the sneezing voice of Japanese A is transferred from terminal 2A to terminal 2B via the network and is heard by German B as it is.
(S16) Assume that German B utters “Gasundheit!”, Which means “precious” to terminal 2B. This voice is translated through the system and transferred to the terminal 2A. Here, in a general German-Japanese dictionary, it is translated as “Gasundheit”-> “health”.
(S17) Japanese A says “Healthy!” From terminal 2A. At this time, according to the present invention, “* notice in case of poor physical condition” is displayed on the display of the terminal 2A (see S37 in FIG. 3). As Japanese A, I can understand that German B worked hard.
(S18) Next, it is assumed that Japanese A also utters “thank you” to terminal 2A. This voice is translated through the system and transferred to the terminal 2B.
(S19) German B is uttered “Danke” from terminal 2B, and German B does not feel any sense of incongruity.

  FIG. 4 is a functional configuration diagram of the translation server in the present invention.

  According to FIG. 4, the bilingual server 1 is interposed between terminals 2 of users who interact in two language systems via a network. The parallel translation server 1 can also collect various sentences from other server groups via the network. The bilingual translation server 1 may interpret a voice conversation between users in real time, or may sequentially translate text such as an email. The parallel translation server 1 is not limited to interpretation or translation between different language systems, and the two language systems may be the same national language. In this case, only the association between words is monitored.

The bilingual translation server 1 according to the present invention, when the first word / phrase of the first language system and the second word / phrase of the second language system are in a bilingual relationship or a correspondence relationship, the user who contacts the second word / phrase Whether or not recognizes a specific emotion. Here, the difference in language system is not limited to a difference in national language such as Japanese <-> foreign languages (English or German), but may be a difference in dialects in the same national language. In addition, it is assumed that the “specific emotion” is a puzzled emotion such as “confused”, “embarrassed”, “prone”, “confused”. This is because such feelings create a communication gap.
First language system: Language system on the speaker side
The language on the other user's side as seen by the user recognizing the specific emotion
(For example, if you want Japanese to recognize specific emotions, the other party's foreign language)
Second language system: Language system on the listener side
User language to recognize specific emotions
(For example, if you want Japanese to recognize specific emotions, Japanese)

  The parallel translation server 1 includes a communication network 10 connected to the network, an automatic parallel translation unit 11, a corpus collection storage unit 12, a corpus search unit 13, a co-occurrence frequency calculation unit 14, a specific emotion determination unit 15, an alarm phrase The storage unit 16 and the specific emotion clarification unit 17 are included. These functional components can be realized by executing a program that causes a computer installed in the apparatus to function. The processing flow of each functional unit can also be executed as a method.

[Automatic translation unit 11]
The automatic parallel translation unit 11 is an existing parallel translation function and has a function of automatically interpreting or translating different language systems. This may be a real-time interpretation of a voice conversation or a text-based translation such as mail. In addition, the automatic parallel translation part 11 does not necessarily process a different language system, and may be the same language system. In this case, only the association between words is monitored.

  When translating the real-time conversation, the automatic translation unit 11 converts the speech of the first language system received from one terminal into the first text by speech recognition, and converts the first text to the second language system. The second text is translated into speech, the second text is converted into speech by speech synthesis, and the speech of the second language system is transmitted to the other terminal. When text such as e-mail is translated, speech recognition and speech synthesis processes are not required.

  Further, the automatic parallel translation unit 11 outputs the first word / phrase of the first language system and the second word / phrase of the second language system to be the parallel translation to the corpus search unit 13. Each phrase is output by morphological analysis. Morphological analysis is a technology that uses grammar and word dictionaries as information sources, divides sentences written in natural language into morphemes (Morpheme), which are the smallest units that have meaning as languages, and discriminates each part of speech. .

[Corpus Collection Accumulation Unit 12]
The corpus collection / storage unit 12 includes a large number of reference texts based on the second language system (Japanese according to FIG. 2). The corpus collection / storage unit 12 collects various sentences such as blogs from an external server group via a network. The text in the second language system is text published by a Web server, a blog server, or an SNS server. These texts often describe an individual's opinions and impressions on various social or cultural topics. Further, for example, a highly reliable sentence issued by mass media may be used.

  The sentences collected here are preferably called “experience corpus”. An experience corpus is one in which a part of the experience of a third party is described in a text such as a blog. When a communication gap is encountered, there is a high probability that an experience corpus in the native language (second language system, for example, Japanese) will be posted in the form of quoting the first phrase and the second phrase.

[Corpus Search Unit 13]
The corpus search unit 13 uses the corpus collection storage unit 12 to search for a reference sentence including the second word (for example, “when eating” in Japanese). Also, a second language system (Japan in FIG. 2) that includes both the first word (for example, “Mahlzeit” in German) and the second word (“Japanese meal” in S12 in FIG. 2). Word) may be searched.

[Co-occurrence frequency calculation unit 14]
The co-occurrence frequency calculation unit 14 determines between the first reference phrase (for example, “Mahlzeit” in German) and each of a plurality of emotion words defined in advance for the retrieved reference sentence (for example, Japanese sentence). The co-occurrence frequency of is calculated. The co-occurrence frequency calculation unit 14 calculates the co-occurrence frequency for each of the first word and each emotion word based on the mutual information amount criterion.

“Emotion” refers to subjective experiences and attitudes caused by cognition and interpretation of things.
The “specific emotion word” is a noun or a verb derived from the noun, for example, a noun “Puzzled”, “Embarrassed”, “Question”, or a verb “Troubled”, “Confused”, “I don't know”. These phrases are likely to be described in the communication gap experience corpus.
An “emotion word” is a phrase that expresses emotion, such as “joyful”, “angry”, “sad”, “fun”, and “confused”. Here, the plurality of emotion words include at least a specific emotion word (for example, “confused”).

  FIG. 5 is an explanatory diagram of the operation of the co-occurrence frequency calculation unit in the present invention.

  According to FIG. 5, the co-occurrence frequency calculation unit 14 associates the co-occurrence frequency and the mutual information amount for each emotion word for each first word / phrase in the first language system. For example, a correspondence graph as shown in FIG. 5 is generated from about 10,000 words / phrases output from the corpus search unit 13. Here, for example, for German “Mahlzeit”, the co-occurrence frequencies of the emotion words “confused”, “happy”, “angry”, “sad”, and “fun” are accumulated.

Here, the co-occurrence frequency calculation unit 14 further calculates mutual information (Mutual Information). Mutual information (Mutual
information) is a probabilistic method that represents a measure of interdependence between two random variable elements x _i and y _k , and is calculated as follows.
I (x _i , y _k ) = log {p (x _i , y _k ) / p (x _i ) p (y _k )}
p (): Number of appearances For example, suppose that there are the following number of appearances in 10,000 words.
Number of appearances of “Mahlzeit” and emotion word “confused” p = 20 times Number of appearances of “Mahlzeit” p = 24 Number of occurrences of “confused” p = 120 In this case, “Mahlzeit” and the emotion word “confused” The mutual information amount is expressed as follows.
I (Mahlzeit, perplexity) = log {p (Mahlzeit, perplexity) / p (Mahlzeit) p (perplexity)}
= Log ₂ {(20/10000) / ((120/10000) ・ (24/10000))}
≒ 6.118

Thus, the mutual information amount of Mahlzeit for all emotion words is calculated.
I (Mahlzeit, puzzled)
I (Mahlzeit, happy)
I (Mahlzeit, get angry)
I (Mahlzeit, sad)
I (Mahlzeit, fun)
...
In this, it means that it is used combining relatively many, so that mutual information amount is large. In order to calculate an appropriate amount of mutual information, it is assumed that an absolute appearance frequency of a certain level or more is observed using a large number of samples. On the other hand, reliability is ensured by excluding minority cases that do not meet this requirement.

[Specific emotion determination unit 15]
The specific emotion determination unit 15 determines whether the co-occurrence frequency for the specific emotion word to be detected is higher than the co-occurrence frequency of other emotion words by a predetermined threshold or more. Here, the second word / phrase determined to be true is estimated to cause the user to recognize the specific emotion.

  In many cases, those who have experienced the communication gap are expected to have specific emotions such as “confused”, “embarrassed”, “distressed”, “confused”. Therefore, these are specified emotion words. When the mutual information amount of a specific emotion word (eg, “confused”) with respect to the first word (eg, “Mahlzeit” in German) of the first language system is higher than a predetermined threshold, the first word ( For example, “Mahlzeit”) is estimated to have a high relative frequency of causing a communication gap.

[Alarm phrase storage unit 16]
The alarm phrase storage unit 16 stores the first phrase (for example, “Mahlzeit” in German) and the second phrase (for example, “when eating” in Japanese) determined to be true by the specific emotion determination unit 15. To do.

[Specific emotion manifestation part 17]
The specific emotion specifying unit 17 clearly indicates to the user who is in contact with the second phrase (for example, “when eating” in Japanese) that the user is based on the specific emotion word (for example, “confused”). Specifically, the specific emotion clarification unit 17 converts a part or all of a sentence of the second language system searched by the corpus search unit 13 or a link destination address into the second word (for example, “Japanese meal” It is clearly indicated to the user as an annotation together with “ The annotation clearly indicates that a specific emotion “confusing” may occur to the user on the listener side, and the meaning is simply described.

  It is also preferable that the specific emotion clarification unit 17 generates an annotation using the corpus searched by the corpus search unit 13 and stores it in advance. For example, a phrase having the highest co-occurrence frequency for the first phrase and the second phrase may be registered as an annotation. The user on the listener's side can recognize cultural and social differences including differences in habits, etc. by checking the annotation description on the terminal screen, and can avoid or reduce the communication gap caused by it. Since the reliability of the annotation expression is directly linked to the practical value, the output information may be created with reference to a dialogue case database with explanations.

  Finally, according to the embodiment described above, the communication gap between German and Japanese has been described. On the other hand, even in the same national language, there may be a communication gap between the dialect (first language system) and the standard language (second language system).

For example, there are the following examples.
Dialects (Ibaraki, etc.) “Exhausted” “Hard” <-> Standard words “Scared” “Horrible” “Uneasy”
Dialects (Gifu, etc.) “get tired” “Shindo” <-> standard words “Elai” “Respect” “Great”

  In the recording of conversational sentences in which these phrases are used, there are many examples of misunderstandings between dialect speakers and standard speakers due to the difference in meaning despite the same expression. Is done. Therefore, not only between different national languages but also between the same national languages can be displayed on the listener side that there is a possibility of causing a communication gap in the same manner as the present invention. In this case, the bilingual translation server 1 of the present invention simply associates the phrases without interpretation or translation.

  As described above in detail, according to the program, the apparatus, and the method of the present invention, it is possible to estimate a specific emotion for a synonym phrase between language systems. Here, the specific emotion is an emotion such as “confused” on the listener side. This is likely to have a communication gap in terms of words transmitted from the speaker. The present invention only estimates the phrase in which the gap occurs and allows the listener to recognize it (notice it). This alone will facilitate smooth communication between interlocutors with different cultures.

  Various changes, modifications, and omissions of the above-described various embodiments of the present invention can be easily made by those skilled in the art. The above description is merely an example, and is not intended to be restrictive. The invention is limited only as defined in the following claims and the equivalents thereto.

DESCRIPTION OF SYMBOLS 1 Parallel translation server 10 Communication interface 11 Automatic parallel translation part 12 Corpus collection storage part 13 Corpus search part 14 Co-occurrence frequency calculation part 15 Specific emotion judgment part 16 Alarm word phrase storage part 17 Specific emotion clarification part 2 Terminal

Claims (9)

Whether or not the user in contact with the second word recognizes a specific emotion when the first word of the first language system and the second word of the second language system are in a translational relationship or a correspondence relationship A program that causes a computer installed in the device to function so as to estimate
Corpus collection and storage means including a number of reference sentences based on the second language system;
Corpus search means for searching for a reference sentence containing a second word using the corpus collection and storage means;
A co-occurrence frequency calculating means for calculating a co-occurrence frequency between the first word and the specific emotion word for the retrieved reference sentence;
Causing the computer to function as specific emotion determination means for determining whether or not the co-occurrence frequency for the specific emotion word to be detected is higher than a predetermined threshold,
A program for causing a computer to function so as to estimate that the second word / phrase determined to be true by the specific emotion determination means causes the user to recognize the specific emotion.   The program according to claim 1, wherein the computer is caused to function so that the first language system and the second language system are in the same national language and have a corresponding relationship.   The program according to claim 1 or 2, wherein the corpus search means causes a computer to search for a reference sentence including a first word and a second word.   The co-occurrence frequency calculating means further calculates a co-occurrence frequency between the first word and each of a plurality of predefined emotion words for the retrieved reference text,
  The specific emotion determination unit determines whether the co-occurrence frequency for the specific emotion word to be detected is higher than a co-occurrence frequency of other emotion words by a predetermined threshold or more.
The program according to any one of claims 1 to 3, wherein the computer functions as described above. The specific emotion word is a noun or a verb derived therefrom,
The program according to claim 4 , wherein the computer is caused to function such that the plurality of emotion words include at least the specific emotion word. 6. The program according to claim 5, which causes the computer to function so that the specific emotion word is a noun “confused”, “embarrassed”, “question”, or verbs “prone”, “confused”, “don't know”. The co-occurrence frequency calculating means, for each of the first word and each emotion word, by the mutual information criterion, from claim 4, characterized in that causes a computer to function so as to calculate the co-occurrence frequency 6 The program according to any one of the above. Whether or not the user in contact with the second word recognizes a specific emotion when the first word of the first language system and the second word of the second language system are in a translational relationship or a correspondence relationship A device for estimating
Corpus collection and storage means including a number of reference sentences based on the second language system;
Corpus search means for searching for a reference sentence containing a second word using the corpus collection and storage means;
A co-occurrence frequency calculating means for calculating the co-occurrence frequency of the first phrase and the specific emotion word for the retrieved reference sentence;
Specific emotion determination means for determining whether the co-occurrence frequency for a specific emotion word to be detected is higher than a predetermined threshold,
The apparatus that estimates that the second word / phrase determined to be true by the specific emotion determination means causes the user to recognize the specific emotion. Whether or not the user in contact with the second word recognizes a specific emotion when the first word of the first language system and the second word of the second language system are in a translational relationship or a correspondence relationship A method for estimating a specific emotion of a device for estimating
The apparatus includes a corpus collection and accumulation unit including a large number of reference texts based on a second language system,
The device is
A first step of searching for a reference sentence including a second phrase using the corpus collection and storage means;
A second step of calculating a co-occurrence frequency between the first phrase and the specific emotion word for the retrieved reference sentence;
Executing a third step of determining whether or not the co-occurrence frequency for the specific emotion word to be detected is higher than a predetermined threshold,
A specific emotion estimation method, wherein the second phrase determined to be true by the third step is estimated to cause a user to recognize a specific emotion.

Images