String output device

Abstract

The present invention addresses the problem of outputting a character string up to the end of a sentence from sequentially acquired character strings. A character string output device 1 comprises an acquisition unit 12 that sequentially acquires character strings, a determination unit 13 that repeatedly determines whether the end of a sentence is included in character strings to be determined, the determination unit 13 making a determination of a target character string initially acquired by the acquisition unit 12 in the first determination and making a determination of another target character string including the target character string and a character string subsequently acquired by the acquisition unit 12 when the determination unit 13 determines, in the first and subsequent determinations, that the end of a sentence is not included in the target character string in the previous determination, and an output unit 14 that outputs a character string up to the end of a sentence from the target character string when the determination unit 13 determines that the end of a sentence is included in the target character string.

Description

[Technical Field]

[0001] One aspect of the present disclosure relates to a character string output device that outputs character strings acquired sequentially up to the end of a sentence. [Background technology]

[0002] Patent Document 1 listed below discloses a detection device that sequentially acquires and recognizes user utterances, and generates a character string representing the content of the user utterance as the speech recognition result. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Publication No. 2017-058804 Summary of the Invention [Problem to be solved by the invention]

[0004] Generally, in speech recognition, if a silent section continues for a certain period of time, it is determined to be a sentence boundary, and the character string corresponding to the speech up to the silent section is output as a single speech recognition result. Therefore, depending on the speaking style, a sentence that should be a complete sentence may be broken up in the middle. For example, when character strings generated by the above detection device are sequentially acquired, there is a possibility that the acquired character string does not end at the end of the sentence.

[0005] Although the meaning of a sentence should be processed as one sentence, inappropriate characters that do not end at the end of the sentence can be difficult to interpret due to the mechanisms of speech recognition, and in the case of translation, for example, the translation results may not be what was intended. Therefore, it is desirable to output the characters up to the end of the sentence for each character that is acquired sequentially. [Means for solving the problem]

[0006] A character string output device according to one aspect of the present disclosure includes an acquisition unit that sequentially acquires character strings, and a determination unit that repeatedly determines whether a target character string, which is a character string to be determined, includes an ending of a sentence, wherein in a first determination, the determination is made on the character string first acquired by the acquisition unit as the target string, and in subsequent determinations, if it is determined in the previous determination that the target string does not include an ending of a sentence, the determination unit makes a determination on a new target string that includes the target string and a character string subsequently acquired by the acquisition unit; and an output unit that, if the determination unit determines that the target string includes an ending of a sentence, outputs the character string up to the ending of the target string.

[0007] In this aspect, a determination is made as to whether a target string including a sequentially acquired character string includes an end of a sentence, and if it is determined that the target string includes an end of a sentence, the target string up to the end of the sentence is output. In other words, the sequentially acquired character strings up to the end of the sentence can be output. [Effects of the Invention]

[0008] According to one aspect of the present disclosure, it is possible to output character strings up to the end of a sentence for sequentially acquired character strings. [Brief explanation of the drawings]

[0009] [Figure 1] FIG. 1 is a diagram illustrating an example of a functional configuration of a character string output device according to an embodiment. [Figure 2] FIG. 10 is a diagram illustrating an example of learning data. [Figure 3] FIG. 10 is a diagram illustrating an example of a method for generating learning data. [Figure 4] FIG. 10 is a diagram illustrating an example of learning data to which labels are assigned in sequence labeling. [Figure 5] 10 is a flowchart showing an example (part 1) of a character string output process executed by the character string output device according to the embodiment. [Figure 6] 10 is a flowchart showing an example (part 2) of the character string output process executed by the character string output device according to the embodiment. [Figure 7] 10 is a flowchart showing an example (part 3) of the character string output process executed by the character string output device according to the embodiment. [Figure 8] FIG. 8 is a diagram showing an example of a speech translation result obtained by a conventional technique for the speech recognition result of FIG. 7. [Figure 9] FIG. 1 is a diagram illustrating an example of machine translation according to the prior art. [Figure 10] FIG. 1 is a diagram illustrating an example of machine translation by a character string output device according to an embodiment. [Figure 11] FIG. 1 is a diagram illustrating a problem (part 1) of speech recognition using conventional technology. [Figure 12] FIG. 10 is a diagram illustrating a problem (part 2) of speech recognition using conventional technology. [Figure 13] FIG. 2 is a diagram illustrating an example of a hardware configuration of a computer used in the character string output device according to the embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0010] Hereinafter, embodiments of the present disclosure will be described in detail with reference to the drawings. In the description of the drawings, the same elements are designated by the same reference numerals, and duplicate explanations will be omitted. Furthermore, the embodiments of the present disclosure in the following description are specific examples of the present invention, and the present invention is not limited to these embodiments unless otherwise specified to limit the present invention.

[0011] A character string output device 1 according to the embodiment is a computer device that outputs a character string. The character string is composed of one or more characters. In the embodiment, the character string is assumed to be in Japanese, but is not limited to this and may be in a language other than Japanese, such as English.

[0012] The character string output device 1 acquires a character string, which is a recognition result, from a voice recognition device that performs voice recognition or a character recognition device that performs character recognition. The voice recognition device or the character recognition device performs voice recognition or character recognition in real time on sequentially input voice or characters. Therefore, the character string output device 1 acquires a character string, which is a recognition result, sequentially every time voice recognition or character recognition is performed.

[0013] Here, as described above, the sequentially acquired character strings may not end at the end of a sentence due to the mechanisms of voice recognition or character recognition. The end of a sentence is the final part of a sentence. The end of a sentence may also be represented by a sentence end symbol, which is a symbol indicating the end of a sentence. Examples of sentence end symbols in Japanese include ".", "!", and "?". The character string output device 1 outputs the character strings up to the end of the sentence, even if the sequentially acquired character strings do not end at the end of a sentence. Details of the functions of the character string output device 1 will be explained below.

[0014] Fig. 1 is a diagram showing an example of the functional configuration of a character string output device 1. As shown in Fig. 1, the character string output device 1 includes a storage unit 10, a learning unit 11, an acquisition unit 12 (acquisition unit), a determination unit 13 (determination unit), an output unit 14 (output unit), and a translation unit 15 (translation unit).

[0015] Each functional block of the character string output device 1 is assumed to function within the character string output device 1, but is not limited to this. For example, some of the functional blocks of the character string output device 1 may function in a computer device different from the character string output device 1, and connected to the character string output device 1 through a network, while appropriately transmitting and receiving information to and from the character string output device 1. Also, some functional blocks of the character string output device 1 may be omitted, multiple functional blocks may be integrated into one functional block, or one functional block may be separated into multiple functional blocks.

[0016] Hereinafter, each function of the character string output device 1 shown in FIG. 1 will be described.

[0017] The storage unit 10 stores any information used in calculations in the character string output device 1 and the results of the calculations in the character string output device 1. The information stored by the storage unit 10 may be referred to by each function of the character string output device 1 as appropriate.

[0018] The storage unit 10 may store a sentence-ending symbol insertion model that, when a string without a sentence delimiter (or sentence-ending symbol), which is a symbol that separates sentences, is input, outputs a string with a sentence delimiter (or sentence-ending symbol) inserted. Examples of sentence delimiters include ",", ".", "!", and "?". For example, when the sentence-ending symbol insertion model receives the string "Now, the meeting will begin" without a sentence delimiter, it outputs the string "Now, the meeting will begin." with the sentence delimiter inserted. The sentence-ending symbol insertion model may be generated by existing technology.

[0019] The sentence-ending symbol insertion model may be a trained model that has been (machine) trained based on training data that is a pair of strings without sentence delimiters (or sentence-ending symbols) and strings with sentence delimiters (or sentence-ending symbols).

[0020] FIG. 2 is a diagram showing an example of training data. In the training data shown in FIG. 2, character strings without sentence delimiters and character strings with sentence delimiters are associated as pairs. Note that the training data shown in FIG. 2 shows an example of all or part of one sentence (text), but is not limited to this and may also include all or part of two or more sentences. In the training data, character strings without sentence delimiters can be regarded as input data, and character strings with sentence delimiters can be regarded as training data.

[0021] The training data may be a set of extracted strings, which are partial strings extracted from strings containing sentence delimiters (or sentence-ending symbols), and the extracted strings themselves. The extracted strings may be partial strings obtained by dividing a string containing sentence delimiters (or sentence-ending symbols) into words and splitting them at random positions, for example.

[0022] FIG. 3 is a diagram showing an example of a method for generating training data. As shown in FIG. 3, four extracted strings, "Now, let's start the meeting," "Now," "the meeting," and "start," are extracted from original data, which is a string with sentence delimiters, "Now, let's start the meeting," "Now," "the meeting," and "start." Then, for each of the four extracted strings, strings, "Now, let's start the meeting," "Now," "the meeting," and "start," are generated with the sentence delimiters removed. In FIG. 3, the pair of "Now, let's start the meeting" and "Now, let's start the meeting," the pair of "Now" and "Now," the pair of "the meeting" and "the meeting," and the pair of "start" and "start." are training data.

[0023] A string of characters with a sentence delimiter (or a sentence ending symbol) included in the training data may be given a label in sequence labeling indicating whether or not a sentence delimiter (or a sentence ending symbol) follows each word that makes up the string. In this case, a string of characters without a sentence delimiter (or a sentence ending symbol) included in the training data may be divided into words. By using training data that has been given labels in sequence labeling, machine learning can be performed as a sequence labeling task to predict which word will be followed by which sentence delimiter.

[0024] 4 is a diagram showing an example of learning data to which labels are assigned in sequence labeling. In FIG. 4, for example, in the character string "Sate, Kaigi wo Hajimaru" (Let's start the meeting), which has a sentence delimiter, the word "Sate" is given a label " <comma>" is assigned to the words "kaigi" and "wo", respectively, indicating that they are not followed by a sentence delimiter. <o>" is added to the word "hajime" and the label " <period>" has been given.

[0025] The learning unit 11 generates a sentence-end symbol insertion model. More specifically, the learning unit 11 performs (machine) learning based on learning data that is a pair of a character string without a sentence delimiter (or a sentence-end symbol) and a character string with a sentence delimiter (or a sentence-end symbol), and generates a sentence-end symbol insertion model as a trained model. Alternatively, the learning unit 11 may perform (machine) learning based on the various types of learning data described above to generate a sentence-end symbol insertion model. Furthermore, the learning unit 11 may generate the learning data itself based on the learning data generation method described above.

[0026] The learning unit 11 causes the storage unit 10 to store the generated sentence-final symbol insertion model. Note that the sentence-final symbol insertion model stored by the storage unit 10 may not be generated by the learning unit 11, but may be a model similarly generated by another device and acquired via a network.

[0027] The acquiring unit 12 sequentially acquires character strings. The acquiring unit 12 may sequentially acquire character strings via a network from another device connected to the character string output device 1 via a network or the like. The acquiring unit 12 may sequentially acquire character strings input sequentially by a user of the character string output device 1 using an input device 1005 (described later) such as a keyboard. The acquiring unit 12 may sequentially acquire character strings that do not (may) end at the end of a sentence, or character strings that (may) have a sentence segmentation error where the sentence is segmented incorrectly.

[0028] The acquisition unit 12 may sequentially acquire character strings that are the recognition results of the voice recognition or character recognition. For example, as described above, the acquisition unit 12 may sequentially acquire character strings that are the recognition results of the voice recognition or character recognition performed in real time by a voice recognition device or a character recognition device.

[0029] The acquiring unit 12 may store the sequentially acquired character strings in the storage unit 10 or may output the character strings to the determining unit 13.

[0030] The determination unit 13 repeatedly determines whether a target string, which is a string to be determined, includes a sentence ending. In the first determination, the determination unit 13 may perform the determination using the string first acquired (output) by the acquisition unit 12 as the target string. In the first or subsequent determinations, if the determination unit 13 has determined in the previous determination that the target string does not include a sentence ending, the determination unit 13 may perform the determination on a new target string including the target string and a string subsequently acquired (output) by the acquisition unit 12 (a string formed by combining the target string and the string subsequently acquired (output) by the acquisition unit 12 (in that order)). In the first or subsequent determinations, if the determination unit 13 has determined in the previous determination that the target string includes a sentence ending, the determination unit 13 may perform the determination on a new target string including a string after the sentence ending in the target string and a string subsequently acquired (output) by the acquisition unit 12 (a string formed by combining the string after the sentence ending in the target string and the string subsequently acquired (output) by the acquisition unit 12 (in that order).

[0031] The determination unit 13 may determine that the target character string contains a sentence ending if the target character string ends with the sentence ending, and may determine that the target character string does not contain a sentence ending if the target character string does not end with the sentence ending.

[0032] The determination unit 13 may determine whether a sentence ending is included based on whether a sentence ending symbol is present in a string obtained by inputting the target string from which the sentence ending symbol has been removed into the sentence ending symbol insertion model.When making the determination, the determination unit 13 may replace the target string with a string obtained by inputting the target string from which the sentence ending symbol has been removed into the sentence ending symbol insertion model.

[0033] A specific example of the process performed by the determination unit 13 will be described later.

[0034] The determination unit 13 may cause the storage unit 10 to store the determination result, or may output the determination result to the output unit 14.

[0035] When the determination unit 13 determines that the target character string includes the end of a sentence, the output unit 14 outputs the character string from the beginning of the sentence to the end of the sentence of the target character string. The output unit 14 may output (display) the character string to an output device 1006 (described later) such as a display, may output to the translation unit 15, or may output (transmit) the character string to another device via a network.

[0036] The translation unit 15 outputs the result of machine-translating the character string output by the output unit 14. The machine translation uses existing technology. The translation unit 15 may output (display) the result on an output device 1006 (described later) such as a display, or may output (transmit) the result to another device via a network.

[0037] Next, examples of various character string output processes executed by the character string output device 1 will be described with reference to FIGS.

[0038] FIG. 5 is a flowchart showing an example (part 1) of the character string output process executed by the character string output device 1. First, the user utters "Therefore, this happens. Therefore, it will rain." Next, the speech recognition device sequentially performs speech recognition and outputs a speech recognition result A "Therefore." Next, the acquisition unit 12 first acquires the speech recognition result A "Therefore." The speech recognition result A "Therefore." is the character string acquired first by the acquisition unit 12. Next, the determination unit 13 removes the sentence delimiter from the acquired speech recognition result A "Therefore," to generate a sentence delimiter-free character string "Therefore" (step S1a). Next, the determination unit 13 inputs the generated sentence delimiter-free character string "Therefore" into the sentence-end symbol insertion model, thereby generating a sentence delimiter-inserted character string "Therefore," in which a sentence delimiter is inserted (step S2a).

[0039] Next, the determination unit 13 determines whether the generated sentence delimiter insertion string "Therefore," includes a sentence end (initial determination), and determines that it does not include a sentence end (step S3a). Next, since the determination unit 13 has determined that the sentence end is not included, the determination unit 13 waits for the next speech recognition result (step S3a). Next, the acquisition unit 12 acquires speech recognition result B "This is what will happen, rain." (following speech recognition result A "Therefore."). Next, the determination unit 13 generates a combined string "Therefore, this is what will happen, rain." by combining, in that order, the sentence delimiter insertion string "Therefore," (which is a string obtained by inputting the sentence delimiter removal string "Therefore," from which the sentence end symbols have been removed, into the sentence-end symbol insertion model that was replaced by the determination unit 13 when the initial determination was made) and speech recognition result B "This is what will happen, rain." subsequently acquired by the acquisition unit 12 (step S3a).

[0040] Next, the determination unit 13 removes the sentence delimiter from the generated combined string "Therefore, this is what happens, so rain will." to generate a sentence delimiter-free string "Therefore, this is what happens, so rain will." (Step S4a). Next, the determination unit 13 inputs the generated sentence delimiter-free string "Therefore, this is what happens, so rain will." into the sentence-end symbol insertion model to generate a sentence delimiter-inserted string "Therefore, this is what happens, so rain will." (Step S5a).

[0041] Next, the determination unit 13 determines (second determination) whether the generated sentence delimiter symbol-inserted string "Therefore, this is what happens. Therefore, the rain" contains an end of the sentence, and determines that it does. Next, in response to the determination by the determination unit 13 that the end of the sentence is contained, the output unit 14 outputs the string "Therefore, this is what happens." (from the beginning of the sentence to the end of the sentence) of the sentence delimiter symbol-inserted string "Therefore, this is what happens. Therefore, the rain" (the string obtained by inputting the sentence delimiter symbol-removed string "Therefore, this is what happens. Therefore, the rain" from which the sentence-end symbols have been removed into the sentence-end symbol insertion model replaced by the determination unit 13 when making the second determination) as the speech recognition correction result (step S6a). On the other hand, when the judgment unit 13 judges in the second judgment that the end of the sentence is included, the judgment unit 13 continues (repeatedly) judges the string obtained by combining, in the order, the string "Therefore, it will be like this. Therefore, it will rain" that comes after the end of the sentence in the sentence delimiter insertion string "Therefore, it will be like this. Therefore, it will rain," and the string acquired (output) by the acquisition unit 12 (after the speech recognition result B "It will be like this, it will be like this.") (step S7a).

[0042] FIG. 6 is a flowchart showing an example (part 2) of the character string output process executed by the character string output device 1. First, the user utters "Therefore, this happens. Therefore, it will rain." Next, the speech recognition device sequentially performs speech recognition and outputs the speech recognition result C "Therefore." Next, the acquisition unit 12 first acquires the speech recognition result C "Therefore." The speech recognition result C "Therefore." is the character string acquired first by the acquisition unit 12. Next, the determination unit 13 removes the sentence delimiter from the acquired speech recognition result C "Therefore," to generate a sentence delimiter-removed character string "Therefore" (step S1b). Next, the determination unit 13 inputs the generated sentence delimiter-removed character string "Therefore" into the sentence-end symbol insertion model, thereby generating a sentence delimiter-inserted character string "Therefore," in which the sentence delimiter is inserted (step S2b).

[0043] Next, the determination unit 13 determines whether the generated sentence delimiter insertion string "Therefore," ends with a sentence end (initial determination), and determines that the end is not a sentence end, i.e., that a sentence end is not included (step S3b). Next, since the determination unit 13 has determined that a sentence end is not included, the determination unit 13 waits for the next speech recognition result (step S3b). Next, the acquisition unit 12 acquires the speech recognition result D "This is how it will turn out, so it will rain." (following the speech recognition result C "Therefore."). Next, the determination unit 13 generates a combined string "Therefore, this is how it will turn out, so it will rain." by combining, in that order, the sentence delimiter insertion string "Therefore," (which is a string obtained by inputting the sentence delimiter removal string "Therefore," from which the sentence end symbols have been removed, into the sentence-end symbol insertion model that the determination unit 13 replaced when making the initial determination), and the speech recognition result D "This is how it will turn out, so it will rain." subsequently acquired by the acquisition unit 12 (step S3b).

[0044] Next, the determination unit 13 removes the sentence delimiter from the generated combined string "Therefore, this is what happens, so it will rain" to generate a sentence delimiter-free string "Therefore, this is what happens, so it will rain" (step S4b). Next, the determination unit 13 inputs the generated sentence delimiter-free string "Therefore, this is what happens, so it will rain" to the sentence-end symbol insertion model to generate a sentence delimiter-inserted string "Therefore, this is what happens, so it will rain" in which the sentence delimiter is inserted (step S5b).

[0045] Next, the determination unit 13 determines (second determination) whether the end of the generated sentence delimiter symbol-inserted string "Therefore, this is what happens. Therefore, it will rain." is the end of the sentence, and determines that the end is the end of the sentence, i.e., that the end of the sentence is included. Next, in response to the determination by the determination unit 13 that the end of the sentence is included, the output unit 14 outputs the string "Therefore, this is what happens. Therefore, it will rain." from the beginning of the sentence to the end of the sentence (the string obtained by inputting the sentence delimiter-removed string "Therefore, this is what happens. Therefore, it will rain" from the beginning to the end of the sentence, which is the string obtained by inputting the sentence delimiter-removed string "Therefore, this is what happens. Therefore, it will rain" from the beginning to the end of the sentence, as the speech recognition correction result (step S6b). If the judgment unit 13 judges in the second or subsequent judgment that the end is not the end of a sentence, i.e., that the end of a sentence is not included, the judgment unit 13 continues (repeatedly) making judgments on the string obtained by combining, in that order, the sentence delimiter insertion string and the string subsequently acquired (output) by the acquisition unit 12 (after the speech recognition result D ``This is what will happen, it will rain.'') (step S7b).

[0046] FIG. 7 is a flowchart showing an example (part 3) of the character string output process executed by the character string output device 1. First, the user utters "Therefore, this happens. Therefore, it will rain." Next, the speech recognition device sequentially performs speech recognition and outputs the speech recognition result E "Therefore." Next, the acquisition unit 12 first acquires the speech recognition result E "Therefore." The speech recognition result E "Therefore." is the character string acquired first by the acquisition unit 12. Next, the determination unit 13 removes punctuation from the acquired speech recognition result E "Therefore," to generate a punctuation-removed character string "Therefore" (step S1c). Next, the determination unit 13 inputs the generated punctuation-removed character string "Therefore" into a punctuation insertion model (functionally similar to a sentence-end symbol insertion model), thereby generating a punctuation-inserted character string "Therefore," in which the punctuation is inserted (step S2c).

[0047] Next, the determination unit 13 determines whether the generated punctuation-inserted character string "Therefore," ends with a period (initial determination), and determines that the end is not a period, i.e., that the end of a sentence is not included (step S3c). Next, since the determination unit 13 has determined that the end of a sentence is not included, the determination unit 13 waits for the next speech recognition result (step S3c). Next, the acquisition unit 12 acquires the speech recognition result F "This is how it will be, so it will rain." (following the speech recognition result E "Therefore."). Next, the determination unit 13 generates a combined character string "Therefore, it will be, so it will rain." by combining, in that order, the punctuation-inserted character string "Therefore," (which is a character string obtained by inputting the punctuation-removed character string "Therefore," into the punctuation insertion model that was replaced by the determination unit 13 when the initial determination was made) and the speech recognition result F "This is how it will be, so it will rain." subsequently acquired by the acquisition unit 12 (step S3c).

[0048] Next, the determination unit 13 removes punctuation from the generated combined character string "Therefore, this is how it happens, therefore, it will rain" to generate a punctuation-removed character string "Therefore, this is how it happens, therefore, it will rain" (step S4c). Next, the determination unit 13 inputs the generated punctuation-removed character string "Therefore, this is how it happens, therefore, it will rain" to the punctuation insertion model to generate a punctuation-inserted character string "Therefore, this is how it happens, therefore, it will rain" in which punctuation is inserted (step S5c).

[0049] Next, the determination unit 13 determines (second determination) whether the generated punctuation-inserted string "Therefore, this happens. Therefore, it will rain." ends with a period, and determines that the string is a period, i.e., that the end of a sentence is included. Next, since the determination unit 13 has determined that the end of a sentence is included, the output unit 14 outputs the string "Therefore, this happens. Therefore, it will rain." (which is the string obtained by inputting the punctuation-removed string "Therefore, this happens. Therefore, it will rain" into the punctuation insertion model that was replaced by the determination unit 13 when the second determination was made) from the punctuation-inserted string "Therefore, this happens. Therefore, it will rain." up to the period (from the beginning of the sentence to the last) as the speech recognition correction result (step S6c). Next, the translation unit 15 outputs the speech translation result "Therefore, this happens. Therefore, it will rain.", which is the result of machine-translating the speech recognition correction result output by the output unit 14.

[0050] Figure 8 is a diagram showing an example of a speech translation result using conventional technology for the speech recognition result of Figure 7. As shown in Figure 8, when speech recognition result E "Therefore." and speech recognition result F "This is what will happen, and it will rain." are directly speech translated using conventional technology, the speech translation result output is "Therefore. This is what will happen, and it will rain."

[0051] Next, the effects of the character string output device 1 according to the embodiment will be described.

[0052] The character string output device 1 includes an acquisition unit 12 that sequentially acquires character strings, and a determination unit 13 that repeatedly determines whether a target character string, which is a character string to be determined, contains a sentence ending. In the initial determination, the character string first acquired by the acquisition unit 12 is used as the target character string, and in subsequent determinations, if it is determined in the previous determination that the target character string does not contain a sentence ending, the determination unit 13 determines a new target character string that contains the target character string and a character string subsequently acquired by the acquisition unit 12. The output unit 14 outputs the character string up to the sentence ending of the target character string if the determination unit 13 determines that the target character string contains a sentence ending. With this configuration, it is determined whether a target character string containing sequentially acquired character strings contains a sentence ending, and if it is determined that the target character string contains a sentence ending, the character string up to the sentence ending of the target character string is output. That is, it is possible to output the character string up to the sentence ending of the sequentially acquired character strings.

[0053] Furthermore, according to the character string output device 1, the determination unit 13 may determine that the target character string contains an end of a sentence if the end of the target character string is an end of a sentence, and may determine that the target character string does not contain an end of a sentence if the end of the target character string is not an end of a sentence. With this configuration, only the end of the target character string is subject to determination, which simplifies the process and enables the process to be performed at a higher speed.

[0054] Furthermore, according to the character string output device 1, in the first or subsequent determination, if it is determined in the previous determination that the target character string includes the end of the sentence, the determination unit 13 may perform a determination on a new target character string that includes the character string after the end of the sentence in the target character string and the character string subsequently acquired by the acquisition unit 12. With this configuration, the character strings sequentially acquired by the acquisition unit 12 can be reliably processed without missing any of them.

[0055] Furthermore, according to the character string output device 1, the determination unit 13 may determine whether a sentence ending is included based on whether a character string obtained by inputting a target character string from which a sentence ending symbol has been removed into a sentence ending symbol insertion model that outputs a character string in which a sentence ending symbol has been inserted when a character string without a sentence ending symbol is input. With this configuration, determination can be made on a more accurate character string corrected based on the sentence ending symbol insertion model, thereby enabling more accurate processing.

[0056] Furthermore, according to the character string output device 1, the determination unit 13 may replace the target character string with a character string obtained by inputting the target character string from which the sentence-final symbols have been removed into the sentence-final symbol insertion model when making the determination. With this configuration, it is possible to finally output a more accurate character string corrected based on the sentence-final symbol insertion model.

[0057] Furthermore, according to the character string output device 1, the sentence-final symbol insertion model may be a trained model trained based on training data that is a pair of character strings without a sentence-final symbol and character strings with a sentence-final symbol. With this configuration, it is possible to more reliably generate a sentence-final symbol insertion model that produces more accurate output.

[0058] Furthermore, according to the character string output device 1, the learning data may be a set of a character string obtained by removing a sentence-end symbol from an extracted character string, which is a part of a character string extracted from a character string having a sentence-end symbol. With this configuration, learning data can be easily generated.

[0059] Furthermore, according to the character string output device 1, a character string having a sentence-final symbol included in the learning data may be assigned a label in sequence labeling indicating whether or not a sentence-final symbol follows each word constituting the character string. With this configuration, machine learning can be performed as a sequence labeling task, thereby enabling more accurate processing.

[0060] Furthermore, according to the character string output device 1, the acquisition unit 12 may sequentially acquire character strings that are the recognition results of voice recognition or character recognition. With this configuration, the recognition results of voice recognition or character recognition can be targeted.

[0061] Furthermore, the character string output device 1 may further include a translation unit 15 that outputs the results of machine translation of the character string output by the output unit 14. This configuration can broaden the range of applications in the real world.

[0062] The character string output device 1 uses a technology for correcting speech recognition results by determining sentence break positions.

[0063] This section explains the problems inherent in conventional speech recognition. When a silent interval continues for a certain period of time, it is determined to be a sentence boundary. The speech up to the silent interval is output as a single speech recognition result. Therefore, depending on the speaking style, a sentence may be broken into segments in the middle of what should be a single sentence. Furthermore, when a sentence is broken into segments, periods are often inserted at the end of the segmented segments. Such speech recognition results, which should be processed as a single sentence but have been split into multiple sentences by the speech recognition mechanism and have inappropriate punctuation, are difficult to interpret and can result in unintended translations, for example. Therefore, a technology is needed for speech recognition systems to combine speech recognition results that have been split into multiple sentences into a single sentence and to reinsert punctuation correctly. Furthermore, speech translation systems, for example, must translate speech recognition results sequentially without delay and output the translation results. Therefore, it is undesirable to wait for the next speech recognition result to determine whether a sentence needs to be combined (because the current speech recognition result and translation result cannot be output while waiting for the next speech recognition result). Therefore, what is needed is a technology that uses only one sentence from the current speech recognition result, determines whether it should be combined with the next speech recognition result (whether it is a sentence that should have been one sentence but has been split into multiple sentences), and waits for the next speech recognition result to combine them only if combining is necessary.

[0064] For example, a specific example of speech translation of an utterance such as "Therefore, it becomes like this." will be described. In the prior art, machine translation of a speech recognition result that is split into multiple sentences due to the speech recognition mechanism and has inappropriate punctuation positions results in an unintended result (see FIG. 9). On the other hand, the character string output device 1 (see FIG. 10) uses only the speech recognition result G, determines that it is not a sentence break (it should be combined with the next speech recognition result), and performs processing to combine it with the next speech recognition result and correctly re-insert sentence break symbols such as punctuation marks. This allows the speech recognition result and machine translation result to be obtained as intended by the user. If the speech recognition result G is determined to be a sentence break, there is no need to wait for the next speech recognition result, and no delay occurs. The sentence break determination is performed by re-inserting sentence break symbols such as punctuation marks using a sentence break insertion model, and determining whether the end of the result is a sentence end symbol (such as a period).

[0065] There is a conventional technology that automatically determines whether a speech recognition result indicates the end of a speech (whether it is a sentence break). This technology divides the speech recognition result into punctuation marks and performs the determination using an end-of-speech determination model. Although this technology can determine the position of a sentence break, it has the following two problems. (1) Depending on the position of the punctuation marks inserted by the speech recognition engine, it may not be possible to correctly determine the position of sentence breaks. (2) Depending on the position of punctuation marks inserted by the speech recognition engine, the position of punctuation marks may be inappropriate when joining sentences.

[0066] The above problem (1) will be explained below (see Figure 11). When there is an error in the punctuation in the speech recognition results and two sentences are included in one speech recognition result, it is not possible to infer the correct sentence break position. For example, as shown in Figure 11, it is not possible to determine that the two sentences "Therefore, this is what happens" and "Therefore, it will rain" are included. It is not possible to determine that the break occurs after "This is what happens." In other words, it depends on the position of the punctuation inserted by the speech recognition engine.

[0067] The above problem (2) will be explained below (see Figure 12). If there is an error in the punctuation marks in the speech recognition results, the sentence will have punctuation marks in inappropriate positions when joined. As shown in Figure 12, the punctuation marks after joining are inappropriate. In other words, it depends on the positions of punctuation marks inserted by the speech recognition engine.

[0068] According to the character string output device 1, all punctuation marks inserted by the speech recognition engine are deleted, and punctuation marks are reinserted using a self-created model to determine sentence breaks, so processing can be performed independently of the positions of punctuation marks inserted by the speech recognition engine, thereby solving the above-mentioned problem.

[0069] There is a conventional technique for predicting the position of punctuation marks. This technique can predict the position of punctuation marks, but it has the following problems. (1) The target data is not sequentially input data that often contains errors, such as voice recognition results (no training data with cut-off parts is used). (2) This technology only predicts the position of punctuation marks and does not support other symbols (such as "!" and "?"). (3) It only processes punctuation marks and does not perform sentence boundary determination or sentence joining.

[0070] The character string output device 1 uses broken sentences as learning data, making it strong at predicting the position of punctuation marks in broken sentences. It is also possible to insert other sentence delimiters in addition to punctuation marks (such as "!" and "?"). It is also a technology that applies punctuation mark insertion to determine sentence delimiters and join sentences.

[0071] Punctuation position prediction technology exists as a conventional technique. This technology can determine the correct position of punctuation marks between two sentences, but it has the following problems. (1) This technology involves inputting two sentences and determining whether a period is needed between them, which causes delays because it has to wait for the next input. (2) It does not correspond to any symbols or positions other than periods between two sentences. (3) It only processes periods and does not determine sentence boundaries or join sentences.

[0072] The character string output device 1 can make a judgment based on just one sentence, so there is no delay. In addition to periods, it is also possible to insert other sentence delimiters (such as "," "!" and "?"). This technology also applies punctuation mark insertion to judge sentence delimiters and join sentences.

[0073] As described above, the character string output device 1 has the following advantages. (1) It is possible to determine whether a sentence needs to be joined using only one speech recognition result. In other words, there is no delay while waiting for the next speech recognition result. (2) It is possible to determine the position of a sentence break without relying on punctuation marks or other sentence break symbols inserted by the speech recognition engine. In other words, the accuracy of sentence break determination is high. (3) It is possible to insert sentence delimiters such as punctuation marks into the combined sentence in an appropriate position, without relying on the sentence delimiters inserted by the speech recognition engine. In other words, it is possible to correct the combined sentence into a more natural sentence. (4) It is possible to insert not only punctuation marks but also sentence separators such as "?" and "!". This means that the accuracy of sentence separator determination is high, and the combined sentence can be corrected to a more natural sentence.

[0074] The character string output device 1 may have the following configuration.

[0075] [Item number 1] A step of removing all sentence delimiters (such as ",", ".", "!", and "?") from the speech recognition results and inserting sentence delimiters using a sentence delimiter insertion model generated by machine learning of training data that pairs sentences with sentence delimiters with sentences from which the sentence delimiters have been removed; a step of determining that the sentence delimiter symbol-inserted speech recognition result does not end with a sentence end symbol (".", "!", or "?") and combining the result with the next speech recognition result; removing all sentence delimiters from the combined speech recognition results again and inserting sentence delimiters using a sentence delimiter insertion model; repeating this process until the end of the sentence delimiter symbol insertion speech recognition result becomes a sentence end symbol; a step of determining the sentence ending with the sentence end symbol as a speech recognition correction result; A method for correcting a speech recognition result, comprising:

[0076] [Item number 2] In the step of determining the end of the sentence, if the end of the sentence is not a sentence end symbol but the sentence contains a sentence end symbol, the part from the beginning of the sentence to the sentence end symbol is output as a speech recognition correction result, and the part from the sentence end symbol to the end of the sentence is combined with the next speech recognition result. Item 1. A method for correcting a speech recognition result.

[0077] [Item number 3] In the step of generating a sentence delimiter insertion model by machine learning, a sentence with a sentence delimiter is used as the original data, and a process of "dividing the sentence into words and dividing it at random positions (to deal with inappropriate sentence divisions in speech recognition)" and a process of "removing the sentence delimiter" are performed to generate a pair of sentences with the sentence delimiter removed. Item 1. A method for correcting a speech recognition result.

[0078] [Item number 4] In the step of generating a sentence delimiter insertion model by machine learning, machine learning is performed on the "sentences with sentence delimiters removed" divided into words as a sequence labeling task to predict which sentence delimiter will be inserted after which word, using sentence delimiter positions in the "sentences with sentence delimiters" as training data. Item 1. A method for correcting a speech recognition result.

[0079] [Item No. 5] The method for correcting speech recognition results is applied to recognize speech from meetings, telephone calls, etc., and the speech recognition results are corrected to have appropriate sentence divisions and symbols, which are then machine-translated to produce speech translation results. Item 1. A speech translation method according to item 1. The character string output device 1 may have the following configuration. [Item No. 11] an acquisition unit that sequentially acquires character strings; A determination unit that repeatedly determines whether a target character string that is a character string to be determined includes an end of a sentence, In the initial determination, the character string acquired first by the acquisition unit is used as a target character string for determination; a determination unit that, in a determination after the first time, if it is determined in the previous determination that the target character string does not include a sentence ending, performs a determination on a new target character string that includes the target character string and a character string subsequently acquired by the acquisition unit; an output unit that outputs a character string up to the end of the sentence from the target character string when the determination unit determines that the target character string includes an end of the sentence; A character string output device comprising: [Item No. 12] the determination unit determines that a sentence ending is included if the target character string ends at the end of a sentence, and determines that a sentence ending is not included if the target character string ends at a non-end of a sentence; Item 12. The character string output device according to item 11. [Item No. 13] In a determination after the first determination, if it is determined in the previous determination that the target character string includes an end of a sentence, the determination unit performs a determination on a new target character string that includes a character string after the end of the sentence in the target character string and a character string subsequently acquired by the acquisition unit. Item 12. The character string output device according to item 11. [Item No. 14] the determination unit determines whether a sentence ending is included or not based on whether a sentence ending symbol is present in a string obtained by inputting a target string from which a sentence ending symbol has been removed to a sentence ending symbol insertion model that outputs a string in which a sentence ending symbol is inserted when a string without a sentence ending symbol is input; 14. The character string output device according to any one of items 11 to 13. [Item No. 15] the determination unit replaces the target character string with a character string obtained by inputting the target character string from which the sentence-final symbol has been removed into a sentence-final symbol insertion model when the determination is made; Item 15. The character string output device according to item 14. [Item No. 16] The sentence-final symbol insertion model is a trained model trained based on training data that is a pair of strings without sentence-final symbols and strings with sentence-final symbols. 16. The character string output device according to item 14 or 15. [Item No. 17] The training data is a set of a string obtained by removing a sentence-ending symbol from a string that is extracted from a string containing a sentence-ending symbol, and the extracted string. Item 17. The character string output device according to item 16. [Item No. 18] For strings containing sentence-final symbols included in the training data, a label is assigned in sequence labeling to indicate whether or not a sentence-final symbol follows each word in the string. 18. The character string output device according to item 16 or 17. [Item No. 19] the acquiring unit sequentially acquires character strings that are recognition results of speech recognition or character recognition; 19. The character string output device according to any one of items 11 to 18. [Item No. 20] a translation unit that outputs a result of machine translation of the character string output by the output unit, 20. The character string output device according to any one of items 11 to 19.

[0080] In the embodiments, Japanese is used as an example language, but the technology can also be used for languages ​​other than Japanese. In the embodiments, "speech recognition" is used as the target, but the technology is not limited to speech recognition and is effective for "sequentially input text containing sentence segmentation errors."

[0081] The block diagrams used to explain the above embodiments show functional blocks. These functional blocks (components) are realized by any combination of at least one of hardware and software. Furthermore, the method for realizing each functional block is not particularly limited. That is, each functional block may be realized using a single device that is physically or logically coupled, or may be realized using two or more physically or logically separated devices that are directly or indirectly connected (for example, using wires, wirelessly, etc.) and these multiple devices. The functional block may also be realized by combining the single device or multiple devices with software.

[0082] Functions include, but are not limited to, judgment, determination, judgment, calculation, computation, processing, derivation, investigation, search, confirmation, reception, transmission, output, access, resolution, selection, election, establishment, comparison, assumption, expectation, consideration, broadcasting, notifying, communicating, forwarding, configuring, reconfiguring, allocation, mapping, and assignment. For example, a functional block (component) that performs transmission is called a transmitting unit or transmitter. As mentioned above, there are no particular limitations on how these functions are implemented.

[0083] For example, the character string output device 1 according to an embodiment of the present disclosure may function as a computer that performs processing of the character string output method of the present disclosure. Fig. 13 is a diagram showing an example of the hardware configuration of the character string output device 1 according to an embodiment of the present disclosure. The character string output device 1 described above may be physically configured as a computer including a processor 1001, a memory 1002, a storage 1003, a communication device 1004, an input device 1005, an output device 1006, a bus 1007, etc.

[0084] In the following description, the term "device" can be interpreted as a circuit, a device, a unit, etc. The hardware configuration of the character string output device 1 may be configured to include one or more of the devices shown in the drawings, or may be configured to exclude some of the devices.

[0085] Each function in the character string output device 1 is realized by loading predetermined software (programs) onto hardware such as the processor 1001 and memory 1002, causing the processor 1001 to perform calculations, control communication via the communication device 1004, and control at least one of reading and writing data in the memory 1002 and storage 1003.

[0086] The processor 1001 controls the entire computer by running, for example, an operating system. The processor 1001 may be configured by a central processing unit (CPU) including an interface with peripheral devices, a control device, an arithmetic unit, a register, etc. For example, the learning unit 11, the acquisition unit 12, the determination unit 13, the output unit 14, and the translation unit 15 described above may be realized by the processor 1001.

[0087] The processor 1001 also reads programs (program codes), software modules, data, etc. from at least one of the storage 1003 and the communication device 1004 into the memory 1002 and executes various processes in accordance with the programs. The programs used are those that cause a computer to execute at least some of the operations described in the above-described embodiments. For example, the learning unit 11, the acquisition unit 12, the determination unit 13, the output unit 14, and the translation unit 15 may be implemented by a control program stored in the memory 1002 and running on the processor 1001, and similar implementations may be made for other functional blocks. While the above-described various processes have been described as being executed by one processor 1001, they may also be executed simultaneously or sequentially by two or more processors 1001. The processor 1001 may be implemented by one or more chips. The programs may also be transmitted from a network via a telecommunications line.

[0088] The memory 1002 is a computer-readable recording medium and may be configured, for example, by at least one of a read-only memory (ROM), an erasable programmable ROM (EPROM), an electrically erasable programmable ROM (EEPROM), a random access memory (RAM), etc. The memory 1002 may also be called a register, a cache, a main memory (primary storage device), etc. The memory 1002 can store executable programs (program codes), software modules, etc. for implementing a wireless communication method according to an embodiment of the present disclosure.

[0089] Storage 1003 is a computer-readable recording medium, and may be composed of at least one of, for example, an optical disk such as a CD-ROM (Compact Disc ROM), a hard disk drive, a flexible disk, a magneto-optical disk (e.g., a compact disk, a digital versatile disk, a Blu-ray disc), a smart card, a flash memory (e.g., a card, a stick, a key drive), a floppy disk, a magnetic strip, etc. Storage 1003 may also be referred to as an auxiliary storage device. The above-mentioned storage medium may be, for example, a database, a server, or other appropriate medium including at least one of memory 1002 and storage 1003.

[0090] The communication device 1004 is hardware (transmission / reception device) for communicating between computers via at least one of a wired network and a wireless network, and is also referred to as, for example, a network device, a network controller, a network card, or a communication module. The communication device 1004 may be configured to include a high-frequency switch, a duplexer, a filter, a frequency synthesizer, etc. to realize at least one of frequency division duplex (FDD) and time division duplex (TDD). For example, the learning unit 11, the acquisition unit 12, the determination unit 13, the output unit 14, and the translation unit 15 described above may be realized by the communication device 1004.

[0091] The input device 1005 is an input device (for example, a keyboard, a mouse, a microphone, a switch, a button, a sensor, etc.) that receives input from the outside. The output device 1006 is an output device (for example, a display, a speaker, an LED lamp, etc.) that outputs to the outside. The input device 1005 and the output device 1006 may be integrated into one device (for example, a touch panel).

[0092] Furthermore, each device, such as the processor 1001 and the memory 1002, is connected by a bus 1007 for communicating information. The bus 1007 may be configured using a single bus, or may be configured using different buses between each device.

[0093] Furthermore, the character string output device 1 may be configured to include hardware such as a microprocessor, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a programmable logic device (PLD), or a field programmable gate array (FPGA), and some or all of the functional blocks may be realized by the hardware. For example, the processor 1001 may be implemented using at least one of these pieces of hardware.

[0094] Notification of information is not limited to the aspects / embodiments described in this disclosure, and may be performed using other methods.

[0095] Each aspect / embodiment described in the present disclosure may be applied to at least one of systems using LTE (Long Term Evolution), LTE-Advanced (LTE-A), SUPER 3G, IMT-Advanced, 4G (4th generation mobile communication system), 5G (5th generation mobile communication system), FRA (Future Radio Access), NR (New Radio), W-CDMA (registered trademark), GSM (registered trademark), CDMA2000, UMB (Ultra Mobile Broadband), IEEE 802.11 (Wi-Fi (registered trademark)), IEEE 802.16 (WiMAX (registered trademark), IEEE 802.20, UWB (Ultra-Wideband), Bluetooth (registered trademark), or other appropriate systems, and next-generation systems extended based on these. Furthermore, a combination of multiple systems (e.g., a combination of at least one of LTE and LTE-A with 5G, etc.) may also be applied.

[0096] The order of the procedures, sequences, flowcharts, etc. of each aspect / embodiment described in this disclosure may be changed unless it is consistent. For example, the methods described in this disclosure present elements of various steps using an example order, and are not limited to the particular order presented.

[0097] Input and output information may be stored in a specific location (for example, memory) or may be managed using a management table. Input and output information may be overwritten, updated, or added to. Output information may be deleted. Input information may be sent to another device.

[0098] The determination may be made based on a value represented by one bit (0 or 1), a Boolean value (true or false), or a numerical comparison (e.g., comparison with a predetermined value).

[0099] Each aspect / embodiment described in this disclosure may be used alone, in combination, or switched depending on the implementation. Furthermore, notification of predetermined information (e.g., notification that "X is true") is not limited to being done explicitly, but may be done implicitly (e.g., by not notifying the predetermined information).

[0100] Although the present disclosure has been described in detail above, it is clear to those skilled in the art that the present disclosure is not limited to the embodiments described herein. The present disclosure can be implemented in modified and altered forms without departing from the spirit and scope of the present disclosure as defined by the claims. Therefore, the description of the present disclosure is intended to be illustrative and does not have any limiting meaning on the present disclosure.

[0101] Software shall be construed broadly to mean instructions, instruction sets, code, code segments, program code, programs, subprograms, software modules, applications, software applications, software packages, routines, subroutines, objects, executable files, threads of execution, procedures, functions, etc., whether referred to as software, firmware, middleware, microcode, hardware description language, or otherwise.

[0102] Software, instructions, information, etc. may also be transmitted or received over a transmission medium. For example, if software is transmitted from a website, server, or other remote source using wired technologies (such as coaxial cable, fiber optic cable, twisted pair, Digital Subscriber Line (DSL)), and / or wireless technologies (such as infrared, microwave), then these wired and / or wireless technologies are included within the definition of transmission media.

[0103] The information, signals, etc. described in this disclosure may be represented using any of a variety of different technologies. For example, data, instructions, commands, information, signals, bits, symbols, chips, etc. that may be referred to throughout the above description may be represented by voltages, currents, electromagnetic waves, magnetic fields or magnetic particles, optical fields or photons, or any combination thereof.

[0104] In addition, terms explained in this disclosure and terms necessary for understanding this disclosure may be replaced with terms having the same or similar meanings.

[0105] As used in this disclosure, the terms "system" and "network" are used interchangeably.

[0106] Furthermore, the information, parameters, etc. described in this disclosure may be expressed using absolute values, may be expressed using relative values ​​from a predetermined value, or may be expressed using other corresponding information.

[0107] The names used for the above parameters are not limiting in any way, and furthermore, the mathematical formulas etc. using these parameters may differ from those explicitly disclosed in this disclosure.

[0108] As used in this disclosure, the terms "determining" and "determining" may encompass a wide variety of actions. "Determining" and "determining" may include, for example, judging, calculating, computing, processing, deriving, investigating, looking up, searching, inquiring (e.g., searching in a table, database, or other data structure), ascertaining, and the like. "Determining" and "determining" may also include receiving (e.g., receiving information), transmitting (e.g., sending information), input, output, accessing (e.g., accessing data in memory), and the like. Furthermore, "judgment" and "decision" can include regarding resolving, selecting, choosing, establishing, comparing, etc. as having been "judged" or "decided." In other words, "judgment" and "decision" can include regarding some action as having been "judged" or "decided." Furthermore, "judgment (decision)" can be interpreted as "assuming," "expecting," "considering," etc.

[0109] The terms "connected," "coupled," or any variation thereof, refer to any direct or indirect connection or coupling between two or more elements, and may include the presence of one or more intermediate elements between two elements that are "connected" or "coupled" to each other. The coupling or connection between elements may be physical, logical, or a combination thereof. For example, "connected" may be read as "access." As used in this disclosure, two elements may be considered to be "connected" or "coupled" to each other using one or more wires, cables, and / or printed electrical connections, as well as electromagnetic energy having wavelengths in the radio frequency range, microwave range, and optical (both visible and invisible) range, as some non-limiting and non-exhaustive examples.

[0110] As used in this disclosure, the phrase "based on" does not mean "based only on," unless expressly stated otherwise. In other words, the phrase "based on" means both "based only on" and "based at least on."

[0111] As used in this disclosure, any reference to an element using a designation such as "first," "second," etc. does not generally limit the quantity or order of those elements. These designations may be used in this disclosure as a convenient method of distinguishing between two or more elements. Thus, a reference to a first and a second element does not imply that only two elements may be employed or that the first element must in some way precede the second element.

[0112] The "means" in the configuration of each of the above devices may be replaced with "part," "circuit," "device," etc.

[0113] When used in this disclosure, the terms "include," "including," and variations thereof are intended to be inclusive, similar to the term "comprising." Furthermore, when used in this disclosure, the term "or" is not intended to be an exclusive or.

[0114] In this disclosure, where articles are added by translation, such as a, an, and the in English, the disclosure may include that the nouns following these articles are in the plural form.

[0115] In the present disclosure, the term "A and B are different" may mean "A and B are different from each other." The term may also mean "A and B are each different from C." Terms such as "separate" and "coupled" may also be interpreted in the same way as "different." [Explanation of symbols]

[0116] 1...character string output device, 10...storage unit, 11...learning unit, 12...acquisition unit, 13...determination unit, 14...output unit, 15...translation unit, 1001...processor, 1002...memory, 1003...storage, 1004...communication device, 1005...input device, 1006...output device, 1007...bus.< / period> < / o> < / comma>

Claims

1. an acquisition unit that sequentially acquires character strings; A determination unit that repeatedly determines whether a target character string that is a character string to be determined includes an end of a sentence, In the initial determination, the character string acquired first by the acquisition unit is used as a target character string for determination; a determination unit that, in a determination after the first time, if it is determined in the previous determination that the target character string does not include a sentence ending, performs a determination on a new target character string that includes the target character string and a character string subsequently acquired by the acquisition unit; an output unit that outputs a character string up to the end of the sentence from the target character string when the determination unit determines that the target character string includes an end of the sentence; Equipped with In a determination after the first determination, if it is determined in the previous determination that the target character string includes an end of a sentence, the determination unit performs a determination on a new target character string that includes a character string after the end of the sentence in the target character string and a character string subsequently acquired by the acquisition unit. A string output device.

2. an acquisition unit that sequentially acquires character strings; A determination unit that repeatedly determines whether a target character string that is a character string to be determined includes an end of a sentence, In the initial determination, the character string acquired first by the acquisition unit is used as a target character string for determination; a determination unit that, in a determination after the first time, if it is determined in the previous determination that the target character string does not include a sentence ending, performs a determination on a new target character string that includes the target character string and a character string subsequently acquired by the acquisition unit; an output unit that outputs a character string up to the end of the sentence from the target character string when the determination unit determines that the target character string includes an end of the sentence; Equipped with the determination unit determines whether a sentence ending is included or not based on whether a sentence ending symbol is present in a string obtained by inputting a target string from which a sentence ending symbol has been removed to a sentence ending symbol insertion model that outputs a string in which a sentence ending symbol is inserted when a string without a sentence ending symbol is input; A string output device.

3. the determination unit determines that a sentence ending is included if the target character string ends at the end of a sentence, and determines that a sentence ending is not included if the target character string ends at a non-end of a sentence; 3. The character string output device according to claim 2.

4. In a determination after the first determination, if it is determined in the previous determination that the target character string includes an end of a sentence, the determination unit performs a determination on a new target character string that includes a character string after the end of the sentence in the target character string and a character string subsequently acquired by the acquisition unit.

3. The character string output device according to claim 2.

5. the determination unit replaces the target character string with a character string obtained by inputting the target character string from which the sentence-final symbol has been removed into a sentence-final symbol insertion model when the determination is made; 3. The character string output device according to claim 2.

6. The sentence-final symbol insertion model is a trained model trained based on training data that is a pair of strings without sentence-final symbols and strings with sentence-final symbols.

3. The character string output device according to claim 2.

7. The training data is a set of a string obtained by removing a sentence-ending symbol from a string that is extracted from a string containing a sentence-ending symbol, and the extracted string.

7. The character string output device according to claim 6.

8. For strings containing sentence-final symbols included in the training data, a label is assigned in sequence labeling to indicate whether or not a sentence-final symbol follows each word in the string.

7. The character string output device according to claim 6.

9. the acquiring unit sequentially acquires character strings that are recognition results of speech recognition or character recognition; 3. The character string output device according to claim 1.

10. a translation unit that outputs a result of machine translation of the character string output by the output unit, 3. The character string output device according to claim 1.

Images