JPH0552506B2

JPH0552506B2 -

Info

Publication number: JPH0552506B2
Application number: JP57170189A
Authority: JP
Inventors: Fumio Togawa; Kenichi Funabashi
Original assignee: Sharp Corp
Current assignee: Sharp Corp
Priority date: 1982-09-28
Filing date: 1982-09-28
Publication date: 1993-08-05
Also published as: JPS5958492A

Description

[Detailed description of the invention]

＜技術分野＞本発明は認識方式の改良に関し、更に詳細には
例えば文節等の一区切りの音声等の一区切りの認
識すべき情報を音韻、かな、音節、文節等のより
細分化された単位要素で認識する認識装置に適用
可能な認識装置に関するものである。＜従来技術＞文節等の一区切りの音声等を音韻、かな、音節
等のより細分化された単位で認識する場合、従来
一般的には入力された認識すべき一区切りの音声
情報等を例えば音響処理して音韻、音節等の単位
毎の特徴ベクトル入力パターンを得ると共に、こ
の入力パターンと予め記憶されている標準パター
ンとのマツチングを行つて入力された情報を候補
単位列として類似度の高いものから出力し、この
出力された候補単位列と文節等の辞書の内容とを
照合して入力された情報に対する文節等の一区切
りの情報を認識している。しかし、このような方法によれば、類似度の高
いものから出力される候補単位列の全てについて
辞書照合処理を行う必要があり、その処理時間が
長くなり、正しい文節等を認識する確度が向上せ
ず、結果的に全体の認識に要する処理量が膨大な
ものになつていた。＜目的＞本発明は、上記従来の欠点を除去した認識装置
を提供することを目的とし、正しい文節等の一区
切りの認識すべき情報を認識する確度を向上させ
ると共に、結果的に全体の認識に要する処理量を
減少させることの出来る認識装置を提供するもの
である。＜実施例＞以下、本発明の認識装置を文節等の一区切りの
音声を音節等のより細分化された単位要素で認識
する認識装置に適用した例を実施例として説明す
る。本発明の実施例によれば、文節等の一区切りの
音声等の認識すべき情報を音韻、かな、音節等の
より細分化されたＮ個の単位要素で認識する認識
装置において、単位要素毎に認識された音節等の
複数個の候補から信頼度の高い組合せ順に候補列
を作成して辞書照合等の処理を行い、妥当な文字
列等の単位要素列を認識結果として出力する場
合、上記の辞書に対応した言語に含まれる文節等
の文字列（単位要素列）について、予め（Ｎ＋
１）個の文字（端子要素）間の接続関係であるＭ
次の遷移関係を記述した遷移行列を設け、上記の
候補列の作成において、この遷移行列を用いて文
字（単位要素）の非遷移関係を積極的に活用し
て、音響処理で得られる候補列のうち、文字（単
位要素）の遷移が不可能な候補列は除外して、次
の高次の辞書照合等の処理量の削減を図ることが
出来るように構成されている。まず、本発明の実施例の説明に先立ち、本発明
の認識方式に用いられる単位要素間の接続関係で
ある遷移関係を示した遷移行列について説明す
る。一般に日本語文章は、全てかな文字で表現した
場合、かな文字列に対応した音節列で表現でき
る。例えば文節「地球の」は“ち”“きゆ”“う”
“の”という４個の単音節といわれる単位要素か
ら成り立つている。２つの音節間の接続関係
（“ち”から“きゆ”、“きゆ”から“う”、“う”か
ら“の”）を、日本語全て、あるいは特定の分野、
話題における文章等について調べると接続（遷
移；以下遷移ということばを使う）しない音節対
がある。例えば行の音節の前には“ん”、“つ”以
外はこない。また“にや”は語頭にこないし、
“へ”（へと発声するもの）は語尾にこない。このような文節を構成する音節の１次の遷移関
係を以下に示す式(1)に従つて記述して、第１図に
示すような遷移行列Ｍ（Ｘ，Ｙ）を作成する。第１図において遷移行列Ｍ（Ｘ，Ｙ）は単位要
素列である文字列の文字Ｘから次の文字Ｙへの遷
移を記述したものであり、単位要素（音節）がＮ
個の場合、（Ｎ＋１）×（Ｎ＋１）の行列であり、
ハード的にはROM等に記憶される。またY₀列に
は各単位要素（１〜Ｎ）が節頭に来るか否かを表
わし、X₀行には各単位要素（１〜Ｎ）が節尾に
来るか否かを表わすデータが書込まれる。例えば“赤い”という文字列の遷移を遷移行列
に書込んだ例を第２図に示す。遷移行列の要素は
０（遷移不可能）か１（遷移可能）の２値のどちら
かで表現され、１ビツトで記憶される。なお、第
２図においては表記“１”以外の行列要素は全て
“０”であり、その表示を省略している。次に遷移行列の作成について、今少し詳細に説
明する。まず遷移行列の作成にあたつて遷移行列メモリ
を“０”に初期セツト〔Ｍ（Ｘ，Ｙ）＝０〕する。次に文字列Ａ＝（a₁，a₂，a₃，……、a_I）但し、Ｉ：列の文字数とした場合、次式(1) Ｍ（０，a₁）＝１，（ｉ＝１）Ｍ（a_i-1，a_i）＝１，（ｉ＝２〜Ｉ）Ｍ（a_I，０）＝１，（ｉ＝Ｉ＋１） ……(1) に従つて、文字列Ａの文字遷移関係を遷移行列Ｍ
（Ｘ，Ｙ）に書込む。同様に認識対象となる文字
列の全てについて遷移関係を書込み遷移行列（１
次）の作成を完了する。このようにして作成された具体的な遷移行列
（１次）Ｍ（Ｘ，Ｙ）の列を第３図に示している。
この第３図より明らかなように例えば（Ｘ，Ｙ）
＝（え、く）のビツト位置が“１”であるため、
“え”から“く”への遷移が存在し、また（Ｘ，
Ｙ）＝（え、け）のビツト位置が“０”であるた
め、“え”から“け”への遷移が存在しないこと
を表わしている。上記は１次の遷移であるが、２次遷移、更には
一般にＭ次へ拡張したＭ次遷移行列も同様に次式
(2)に従つて作成することが出来る。Ｍ次遷移行列：Ｍ（X₁，X₂，X₃，……，M_M，
Ｙ），（Ｎ＋１）^M+1次元Ｍ（a_i-M，a_i-(M-1)，……，a_i）＝１，（ｉ＝１〜
Ｉ＋１） ……(2) 但し０＞Ｉのときａ＝０本発明の実施例は、この遷移しない音節の非遷
移関係を積極的に活用して、入力された文節音声
を音節毎に処理して認識する場合、複数個の音節
候補の時系列から信頼度の高い組合せ順に候補列
を作成する候補列出力部において、上記第３図に
示したような遷移行列を参照して遷移不可能な音
節遷移を含む候補列は除外して、遷移可能な候補
列のみ、次の高次の辞書照合等の処理を行うよう
にしたものである。次に本発明の実施例を図面を参照して説明す
る。第４図は候補文字ラテイスから候補列を作成す
る候補列出力部に上記の遷移行列に基く認識処理
を適用した装置のブロツク図である。第４図において、文節音声入力部１に入力され
た音声情報は次段の音響処理・比較部２に入力さ
れる。この音響処理・比較部２は従来公知のもの
であり、例えば文節音声入力部１に入力された文
節音声信号が音響処理部２により単音節毎に特徴
抽出処理が行なわれ、各単音節毎の特徴パターン
が同処理部２内のバツフアに一時記憶される。一
方記憶装置３には各単音節毎の標準パターンP_i
（ｉ＝１〜Ｎ）が記憶されており、この標準パタ
ーンP_iが順次読出されて処理・比較部２において
該処理部内のバツフアに記憶された入力音声の入
力特徴パターンとのマツチング計算が行なわれ、
最も近似したものが第１候補として、また順次近
似したものが次候補として選出され、その結果が
候補音節ラテイスメモリ４に記憶される。上記候補音節ラテイスメモリ４に記憶された複
数個の候補音節の時系列は候補列作成部５及び遷
移行列メモリ６を備えた候補列出力部７に入力さ
れ、該候補列出力部７において、遷移行列メモリ
６の内容を参照して遷移不可能な音節遷移を含む
候補列は除外して、遷移可能な候補列のみ、信頼
度の高い組合せ順に作成され、この候補列と辞書
８に記憶された文節とが辞書照合部９により照合
され、一致すればその結果が文節出力部１０に出
力されるように構成されている。次に上記候補列出力部７で実行されている遷移
行列を用いた候補音節列作成動作について、第５
図に示す遷移行列を用いた候補列作成の処理ブロ
ツク図を参照して説明する。上記第４図に示した音響処理・比較部２から出
力された複数個の候補音節の時系列を記憶する候
補音節ラテイスメモリ４の内容をもとに、候補音
節列作成部１１において信頼度の高い順に候補列
が作成され、その結果が候補音節列バツフア１２
に一次記憶される。この候補音節列バツフア１２
の記憶された候補音節列は遷移行列参照部１３に
おいてメモリ６に記憶された遷移行列：Ｍ（Ｘ，
Ｙ）を参照して、遷移可能か不可能かを次式(3)に
よつて判定部１４において判定し、可能な候補列
のみ候補音節列書込み部１５を介して候補音節列
出力バツフア１６に記憶していく。今第ｊ番目の候補音節列を Aj＝（a₁，a₂，……，a_I）但し、a_i：第ｉ番目の音節番号Ｉ：列の音節数とした場合、判定部１４による遷移行列Ｍ（Ｘ，
Ｙ）を用いた候補列否定はＭ（０，a₁）＝０（ｉ＝１）Ｍ（a_i-1，a_i）＝０，（ｉ＝２〜Ｉ）Ｍ（a_I，０）＝０，（ｉ＝Ｉ＋１） ……(3) のいずれか一つが成立した場合に成される。この(3)式において、いずれか一つが成立した遷
移不可能な音節列を含んだ候補音節列は除外さ
れ、次の候補音節列について同様の判定を行な
い、遷移可能な候補音節列のみが出力バツフア１
６に記憶される。今、一文節音声として「国民は」を入力した場
合、音響処理・比較部２の処理により候補音節ラ
テイスメモリ４に次表の如き候補音節が時系列に
記憶される。 <Technical Field> The present invention relates to the improvement of recognition methods, and more specifically, the present invention relates to the improvement of recognition methods, and more specifically, the present invention relates to the improvement of recognition methods, and more specifically, the present invention relates to the improvement of recognition methods. The present invention relates to a recognition device that can be applied to a recognition device that performs recognition. <Prior art> When recognizing one segment of speech, such as a phrase, in more subdivided units such as phonemes, kana, syllables, etc., conventionally, the input speech information of one segment to be recognized is generally processed by, for example, acoustic processing. to obtain a feature vector input pattern for each unit such as phoneme, syllable, etc. This input pattern is matched with a pre-stored standard pattern, and the input information is used as a candidate unit string, starting from the one with the highest degree of similarity. The output candidate unit string is compared with the contents of a dictionary of phrases, etc., to recognize a section of information, such as a phrase, for the input information. However, according to this method, it is necessary to perform dictionary matching processing on all candidate unit sequences output from those with high similarity, which increases the processing time and improves the accuracy of recognizing correct clauses, etc. As a result, the amount of processing required for overall recognition was enormous. <Purpose> The present invention aims to provide a recognition device that eliminates the above-mentioned drawbacks of the conventional art, and improves the accuracy of recognizing a section of information such as a correct phrase, and as a result improves overall recognition. The present invention provides a recognition device that can reduce the amount of processing required. <Example> Hereinafter, an example in which the recognition apparatus of the present invention is applied to a recognition apparatus that recognizes one segment of speech such as a phrase into more subdivided unit elements such as syllables will be described as an example. According to an embodiment of the present invention, in a recognition device that recognizes information to be recognized, such as a segment of speech such as a phrase, using N unit elements further divided into phonemes, kana, syllables, etc., each unit element is If you want to create a candidate string in the order of combinations with high reliability from multiple candidates of recognized syllables, etc., perform processing such as dictionary matching, and output a valid unit element string such as a character string as a recognition result, use the above method. For character strings (unit element strings) such as clauses included in languages that are compatible with the dictionary, (N+
1) M, which is the connection relationship between characters (terminal elements)
A transition matrix that describes the following transition relationships is provided, and in creating the above candidate sequences, this transition matrix is used to actively utilize the non-transition relationships of characters (unit elements) to obtain candidate sequences obtained by acoustic processing. Among them, candidate strings in which character (unit element) transitions are impossible are excluded, thereby reducing the amount of processing required for the next higher-order dictionary matching, etc. First, prior to describing embodiments of the present invention, a transition matrix showing a transition relationship, which is a connection relationship between unit elements used in the recognition method of the present invention, will be described. In general, if a Japanese sentence is expressed entirely in kana characters, it can be expressed as a syllable string corresponding to the kana character string. For example, the phrase ``earth'' is ``chi'', ``kiyu'', and ``u''.
It is made up of four monosyllable unit elements called "no". The connection relationship between two syllables (from “chi” to “kiyu”, from “kiyu” to “u”, from “u” to “no”) can be studied in all of Japanese, or in specific fields,
If you look at sentences in topics, there are syllable pairs that do not connect (transition; hereinafter we will use the term transition). For example, the syllables in a row are preceded by nothing other than "n" or "tsu." Also, “niya” does not come at the beginning of the word,
“He” (pronounced “he”) does not come at the end of the word. The first-order transition relationship of the syllables constituting such a bunsetsu is described according to the following equation (1), and a transition matrix M(X, Y) as shown in FIG. 1 is created. In Figure 1, the transition matrix M(X, Y) describes the transition from character X to the next character Y in a character string that is a unit element string, and the unit elements (syllables) are N.
, it is a (N+1)×(N+1) matrix,
In terms of hardware, it is stored in ROM etc. In addition, the _Y0 column indicates whether each unit element (1 to N) comes at the beginning of the clause, and the _X0 row contains data indicating whether each unit element (1 to N) comes at the end of the clause. written. For example, FIG. 2 shows an example in which the transition of the character string "red" is written in a transition matrix. The elements of the transition matrix are expressed as either 0 (transition not possible) or 1 (transition possible), and are stored as 1 bit. In FIG. 2, all matrix elements other than the notation "1" are "0", and their display is omitted. Next, the creation of the transition matrix will be explained in a little more detail. First, when creating a transition matrix, the transition matrix memory is initially set to "0" [M(X,Y)=0]. Next, character string A = (a ₁ , a ₂ , a ₃ , ..., a _I ) However, when I is the number of characters in the string, the following formula (1) M (0, a ₁ ) = 1, (i =1) M(a _i-1 , a _i )=1, (i=2~I) M(a _I ,0)=1, (i=I+1) ...According to (1), character string A The character transition relationship of is expressed as a transition matrix M
Write to (X, Y). Similarly, write transition relationships for all character strings to be recognized and transition matrix (1
Complete the creation of the following). FIG. 3 shows the columns of a concrete transition matrix (first order) M(X, Y) created in this way.
As is clear from this figure 3, for example (X, Y)
Since the bit position of = (E, KU) is “1”,
There is a transition from “e” to “ku”, and (X,
Since the bit position of Y)=(E, KE) is "0", this indicates that there is no transition from "E" to "KE". The above is a first-order transition, but the second-order transition, and moreover, the M-order transition matrix that is generally extended to the M-order is also expressed by the following formula.
It can be created in accordance with (2). M-order transition matrix: M(X ₁ , X ₂ , X ₃ , ..., M _M ,
Y), (N+1) ^M+1 dimension M(a _iM , a _i-(M-1) , ..., a _i )=1, (i=1~
I+1) ...(2) However, when 0 > I, a=0 The embodiment of the present invention actively utilizes this non-transitional relationship between syllables that do not transition, and processes the input syllable speech syllable by syllable. When recognizing a syllable, the candidate string output unit that creates a candidate string in the order of highly reliable combinations from the time series of multiple syllable candidates refers to the transition matrix shown in Figure 3 above to determine which transitions are impossible. Candidate sequences that include syllable transitions are excluded, and only transitional candidate sequences are subjected to the next higher-order dictionary matching process. Next, embodiments of the present invention will be described with reference to the drawings. FIG. 4 is a block diagram of an apparatus in which recognition processing based on the above-mentioned transition matrix is applied to a candidate string output unit that creates a candidate string from candidate character lattice. In FIG. 4, speech information input to the phrase speech input section 1 is input to the next stage acoustic processing/comparison section 2. This acoustic processing/comparison section 2 is conventionally known, and for example, the acoustic processing section 2 performs feature extraction processing on a syllable by monosyllable for the syllable speech signal input to the syllable speech input section 1. The characteristic pattern is temporarily stored in a buffer within the processing section 2. On the other hand, the standard pattern P _i for each single syllable is stored in the storage device 3.
(i = 1 to N) are stored, and this standard pattern P _i is sequentially read out and the processing/comparison unit 2 performs a matching calculation with the input feature pattern of the input voice stored in the buffer in the processing unit. Re,
The most approximated one is selected as the first candidate, and the sequentially approximated ones are selected as the next candidates, and the results are stored in the candidate syllable latex memory 4. The time series of the plurality of candidate syllables stored in the candidate syllable latex memory 4 is inputted to a candidate sequence output unit 7 having a candidate sequence creation unit 5 and a transition matrix memory 6, and in the candidate sequence output unit 7, a transition matrix With reference to the contents of the memory 6, candidate strings that include syllable transitions that cannot be transitioned are excluded, and only transitional candidate strings are created in the order of combinations with high reliability, and this candidate string and the phrases stored in the dictionary 8 are created. are compared by the dictionary matching unit 9, and if they match, the result is output to the clause output unit 10. Next, regarding the candidate syllable string creation operation using the transition matrix executed by the candidate string output unit 7, the fifth
The process for creating candidate columns using the transition matrix shown in the figure will be explained with reference to the block diagram. Based on the contents of the candidate syllable latex memory 4 that stores the time series of a plurality of candidate syllables output from the acoustic processing/comparison section 2 shown in FIG. Candidate strings are created in order, and the results are stored in a candidate syllable string buffer 12.
is temporarily stored in This candidate syllable string buffer 12
The stored candidate syllable string is stored in the transition matrix reference unit 13 as a transition matrix: M(X,
Y), the determination unit 14 determines whether the transition is possible or not using the following equation (3), and only possible candidate sequences are sent to the candidate syllable sequence output buffer 16 via the candidate syllable sequence writing unit 15. I will remember it. Now, the j-th candidate syllable string is Aj = (a ₁ , a ₂ , ..., a _I ), where a _i is the i-th syllable number and I is the number of syllables in the string, then the transition by the determination unit 14 Matrix M(X,
Candidate sequence negation using Y) is M(0, a ₁ ) = 0 (i = 1) M (a _i-1 , a _i ) = 0, (i = 2 ~ I) M (a _I , 0) =0, (i=I+1)...This is done when any one of (3) holds true. In this equation (3), candidate syllable strings that include non-transitionable syllable strings in which any one of them is true are excluded, the same judgment is made for the next candidate syllable string, and only transitional candidate syllable strings are output. Batsuhua 1
6 is stored. Now, when "Kokuminwa" is inputted as a single sentence speech, candidate syllables as shown in the following table are stored in chronological order in the candidate syllable latex memory 4 through processing by the acoustic processing/comparison section 2.

【表】このメモリ４に記憶された音節ラテイスを基
に、信頼度の高い順に候補列が作成され、遷移行
列：Ｍ（Ｘ，Ｙ）を参照して作成された候補列が
遷移可能なもののみが出力され、この例の場合に
は候補音節列が次の如く出力される。[Table] Based on the syllable lattice stored in this memory 4, candidate strings are created in order of reliability, and the candidate strings created by referring to the transition matrix: M(X, Y) are transitionable ones. In this example, the candidate syllable string is output as follows.

【表】遷移行列を参照しない従来方式によれば信頼度
の最も高い候補列として「GOKUPINWA」が出
力されにことになるが、本発明方式によれば、こ
の候補列の音節の遷移例えば“KU”から“PI”
が遷移不可能であると遷移行列：Ｍ（Ｘ，Ｙ）を
用いて判断され、以後の辞書照合処理から除外さ
れる。以上の遷移行列は１次遷移であるが、２次遷
移、更には一般的なＭ次遷移まで同じ手法で拡張
することができる。なおＭ次の遷移行列の作成は上述の式(2)に従
い、候補音節列の否定は次に示す式(4)によつて行
うことが出来る。即ち、Ｍ次遷移行列：Ｍ（X₁，X₂，X₃，……，
M_M，Ｙ）への拡張の場合、第ｊ候補列をAj＝
（a₁，a₂，……，a_I）とするとＭ（a_i-M，a_i-(M-1)，……，a_i）＝０（ｉ＝１〜Ｉ
＋１） ……(4) （但し０，＞Ｉのときａ＝０）のいずれか一つが成立した場合に否定が成され
る。なお、Ｍの次数を大きくとれば、候補音節列の
限定が強くなり、本発明装置による効果は大きく
なる。上記した、本発明装置による認識対象は文節に
限らず、音節、単語、文章でもよく、また細分化
された単位は音節に限らず、音韻、単語でもよ
い。またアルフアベツト等の文字列でもよい。本発明は一般に認識対象語を構成する細分化し
た単位の遷移関係の存在する文字列であれば適用
可能である。＜効果＞以上の如く、本発明によれば確度高く、正しい
候補列を抽出することが出来るため、正しい文節
等を認識する確度が高くなり、結果的に高次の辞
書照合等の処理量を減少させることが出来る。[Table] According to the conventional method that does not refer to the transition matrix, "GOKUPINWA" would be output as the candidate string with the highest reliability, but according to the method of the present invention, the syllable transition of this candidate string, for example, "KU ” to “PI”
It is determined that the transition is not possible using the transition matrix: M(X, Y), and is excluded from the subsequent dictionary matching process. Although the above transition matrix is a first-order transition, it can be extended to a second-order transition and even a general M-order transition using the same method. Note that the M-order transition matrix can be created according to the above equation (2), and the candidate syllable string can be negated using the following equation (4). That is, M-order transition matrix: M(X ₁ , X ₂ , X ₃ , ...,
M _M , Y), the j-th candidate column is Aj =
If (a ₁ , a ₂ , ..., a _I ), then M (a _iM , a _i-(M-1) , ..., a _i ) = 0 (i = 1 ~ I
+1) ...(4) (However, when 0,>I, a=0) Negation is achieved if any one of the following holds true. Note that, if the degree of M is increased, the candidate syllable strings are more limited, and the effect of the device of the present invention becomes greater. The objects to be recognized by the apparatus of the present invention described above are not limited to phrases, but may be syllables, words, and sentences, and the subdivided units are not limited to syllables, but may also be phonemes or words. It may also be a character string such as alphanumeric characters. The present invention is generally applicable to character strings in which there is a transition relationship between subdivided units constituting a recognition target word. <Effects> As described above, according to the present invention, since it is possible to extract correct candidate sequences with high accuracy, the accuracy of recognizing correct phrases, etc. is increased, and as a result, the amount of processing such as high-level dictionary matching can be reduced. It can be reduced.

[Brief explanation of drawings]

第１図は１次遷移行列を示す図、第２図は文字
列の遷移を書込んだ遷移行列例を示す図、第３図
は文節文字列の遷移行列例を示す図、第４図は本
発明の実施された認識装置の構成を示すブロツク
図、第５図は本発明に係る候補列作成処理ブロツ
ク図である。２……音響処理・比較部、４……候補音節ラテ
イスメモリ、５……候補列作成部、６……遷移行
列メモリ、７……候補列出力部、８……辞書メモ
リ、９……辞書照合部。 Figure 1 shows a linear transition matrix, Figure 2 shows an example of a transition matrix in which character string transitions are written, Figure 3 shows an example of a transition matrix for bunsetsu character strings, and Figure 4 shows an example of a transition matrix in which character string transitions are written. FIG. 5 is a block diagram showing the configuration of a recognition apparatus in which the present invention is implemented. FIG. 5 is a block diagram of a candidate sequence creation process according to the present invention. 2... Acoustic processing/comparison unit, 4... Candidate syllable latex memory, 5... Candidate string creation section, 6... Transition matrix memory, 7... Candidate string output section, 8... Dictionary memory, 9... Dictionary collation Department.

Claims

[Claims] 1. In a device that recognizes one section of information to be recognized using N syllables that are further divided into N syllables, a predetermined sequence of unit syllables to be recognized (N
+1) an inter-syllable connection information memory for storing connection/disconnection information between unit syllables; a means for acoustically processing and comparing input speech to generate candidate syllable lattices; means for generating candidate sequences based on the generated candidate sequences; means for generating only connectable candidate sequences from the generated candidate sequences based on the inter-syllable connection information in the memory; dictionary checking of the generated connectable candidate sequences, etc. What is claimed is: 1. A recognition device comprising means for performing the processing and outputting a valid unit character string as a recognition result.