JP5553037B2 - Text input support system, text input support device, reference information creation device, and program - Google Patents

Description

  The present invention relates to a text input support system, a text input support device, a reference information creation device, and a program.

  For example, in a document creation apparatus that creates a document using a computer, a word that will be inserted next by a user is predicted for a sentence that is being input in order to reduce the typing input load by the user. Some of them have an input support function that is presented to the user as a completion candidate and the word selected by the user from the completion candidates is inserted into the sentence.

Here, various inventions have been proposed for the input support function in the document creation apparatus as described above.
For example, a pen input computer predicts a character string that will be input after text input and presents one or more so that the user does not have to input the character string. An invention has been proposed (see Patent Document 1).
For example, it is an invention that learns a small-sized dictionary and a statistical language model from an existing text corpus so that prediction input can be performed even on hardware with a small amount of memory such as a mobile phone or a PDA, and uses that to predict input. An invention has been proposed in which it is checked whether a similar expression appears in a text corpus, a dictionary, or a language model for the last several character strings of input text, and a subsequent character string is predicted. (See Patent Document 2).
For example, an invention has been proposed in which the priority order is adjusted so that a character string of a reply source text appears at the top as a predicted input candidate when creating a text when replying to an e-mail (see Patent Document 3).

Japanese Patent Laid-Open No. 10-154033 JP 2006-216044 A JP 2006-344039 A

  The present invention provides a candidate based on a connection with a sentence preceding the sentence being input when presenting a candidate that is predicted to be input subsequent to the sentence being input by the user who is creating the sentence. The purpose is to propose a technology that can present

  The present invention according to claim 1 is a computer-specific function for analyzing a sentence preceding a sentence being input and identifying a word included in the sentence in a sentence being created based on an operation input by a user. From the storage unit that stores two consecutive sentences included in an existing sentence in association with a word included in a preceding sentence as a preceding word, and a word included in a succeeding sentence as a subsequent word, the specific function A search function for searching for a subsequent word associated with an antecedent word that matches the word specified by, and a presentation function for presenting the subsequent word searched by the search function to a user. It is a program.

  According to a second aspect of the present invention, in the first aspect of the present invention, the storage unit includes a group of preceding words in which each preceding word included in a preceding sentence in an existing sentence is grouped in sentence units and the preceding word. A sentence including the preceding word group and a sentence appearance amount indicating the appearance degree of the sentence including the preceding word group in the existing sentence together with the association information with each subsequent word included in the sentence following the sentence including the group. The word appearance amount indicating the appearance degree of the subsequent word in the sentence that follows is stored, and the search function is the subsequent corresponding to the preceding word group that matches the word group included in the sentence preceding the sentence being input. A search is made for a word, and a sentence appearance amount relating to the preceding word group and a word appearance amount relating to a combination of the preceding word group and the subsequent word are searched, and the presenting function is a subsequent word searched by the search function. For the sentence searched with the subsequent word The ratio of word appearance amount is presented in descending order with respect to the amount, which is a program, characterized in that.

  According to a third aspect of the present invention, in the first aspect of the present invention, the storage means is included in each preceding word included in a sentence preceding in an existing sentence and a sentence following the sentence including the preceding word. Along with association information with each subsequent word, a sentence appearance amount indicating the appearance degree of a sentence including the preceding word in an existing sentence, and an appearance degree of the subsequent word in a sentence following the sentence including the preceding word The amount of word appearance is stored, and the search function searches for a subsequent word corresponding to each preceding word that matches each word included in a sentence preceding the sentence being input, and a sentence related to the preceding word The amount of appearance and the word appearance amount related to the combination of the preceding word and the subsequent word are searched, and the presentation function uses the subsequent word searched by the search function as a word for the sentence appearance amount searched together with the subsequent word. High ratio of appearance amount Presented in the order, it is a program, characterized in that.

  According to a fourth aspect of the present invention, in the present invention according to the third aspect, the storage means includes association information between each preceding word and each subsequent word included in a sentence following the sentence including the preceding word. Information on the case of the preceding word is stored, the specifying function specifies a case of a word included in a sentence preceding the sentence being input, and the search function is used for a sentence preceding the sentence being input. It is a program characterized by searching for a subsequent word associated with a preceding word having a matching word and case.

  According to the fifth aspect of the present invention, a computer analyzes two consecutive sentences included in an existing sentence, a word included in a preceding sentence is a preceding word, and a word included in a succeeding sentence is a subsequent word. For the sentence being entered in the sentence being created based on the input function specified by the user and the operation input by the user, the subsequent word corresponding to the preceding word that matches the word contained in the sentence preceding the sentence is given to the user. The program for realizing a storage function for storing the preceding word and the succeeding word specified by the specifying function in association with each other and storing them in the storage means for the sentence input support process to be presented.

  The present invention according to claim 6 is the present invention according to claim 5, wherein for each preceding word group in which each preceding word included in the preceding sentence in the existing sentence is grouped in sentence units, the preceding word in the existing sentence. A sentence appearance amount indicating the appearance degree of a sentence including a group is calculated, and for each combination of each preceding word group and each subsequent word included in a sentence following the sentence including the preceding word group, the preceding word group is calculated. The computer further realizes a calculation function for calculating a word appearance amount indicating the appearance degree of the subsequent word in a sentence subsequent to the sentence including the word, and the word included in the sentence preceding the sentence being input in the sentence input support process In order to present the subsequent words corresponding to the preceding word group matching the group in descending order of the ratio of the word appearance amount related to the subsequent word to the sentence appearance amount related to the preceding word group, the storage function Group and the preceding word group Together with correspondence information between each subsequent word included in statement that follows, the stores the calculated sentence occurrence amount calculated by the function and word appearance amount in the storage means, a program, characterized in that.

  According to a seventh aspect of the present invention, in the present invention according to the fifth aspect, for each preceding word included in the preceding sentence in the existing sentence, a sentence appearance indicating a degree of appearance of the sentence including the preceding word in the existing sentence. For each combination of each preceding word and each subsequent word included in the sentence that follows the sentence that includes the preceding word, the degree of appearance of the subsequent word in the sentence that follows the sentence that includes the preceding word is calculated. The computer further realizes a calculation function for calculating a word appearance amount to indicate, and in the sentence input support process, subsequent words corresponding to respective preceding words that match each word included in a sentence preceding the sentence being input. The memory function is included in a sentence following each preceding word and a sentence including the preceding word in order of increasing the ratio of the word appearance amount related to the succeeding word to the sentence appearing amount related to the preceding word. Each successor and Together with correspondence information, the stores the calculated sentence occurrence amount calculated by the function and word appearance amount in the storage means, a program, characterized in that.

  According to an eighth aspect of the present invention, in the present invention according to the seventh aspect, the calculation function calculates a sentence appearance amount for each preceding word, and for each sentence including the preceding word, the number of words in the sentence is large. In addition, the calculation of the word appearance amount for each combination of the preceding word and the succeeding word is performed for each sentence including the succeeding word and the sentence following the sentence including the preceding word. The program is characterized by adding a smaller value as the number of words in a sentence increases.

  The present invention according to claim 9 is the present invention according to claims 7 and 8, wherein the specific function is included in a case of a preceding word included in a sentence preceding in an existing sentence and a sentence following the sentence. A program that identifies a case of a subsequent word, and the calculation function increases a value to be added to a word appearance amount related to the subsequent word for a subsequent word having a case that matches the specified preceding word. It is.

  According to a tenth aspect of the present invention, in the present invention according to the seventh to ninth aspects, the specifying function specifies a case of a preceding word included in a preceding sentence in an existing sentence, and in the sentence input support process, In order to present the word included in the sentence preceding the sentence being input and the subsequent word corresponding to the preceding word having the matching case, the storage function is provided for each preceding word and the sentence following the sentence including the preceding word. A program characterized in that information on the case of the preceding word is stored in the storage unit together with association information with each subsequent word included.

  The present invention according to claim 11 analyzes two consecutive sentences included in an existing sentence, specifies a word included in a preceding sentence as a preceding word, and specifies a word included in a succeeding sentence as a subsequent word. A reference information creating unit having a first specifying unit, a storage unit that stores the preceding word and the subsequent word specified by the first specifying unit in association with each other, and is being created based on an operation input by the user In the sentence, the second identification unit that analyzes the sentence preceding the sentence being input and identifies the word included in the sentence, and the preceding word that matches the word identified by the second identification unit A text input support unit comprising: search means for searching for a subsequent word from the storage means; and presentation means for presenting the subsequent word searched by the search means to a user. A text input support system .

  According to a twelfth aspect of the present invention, in a sentence being created based on an operation input by a user, a specifying unit that analyzes a sentence preceding the sentence being input and identifies a word included in the sentence, Two consecutive sentences included in a sentence are specified by the specifying unit from a storage unit that stores a word included in a preceding sentence as a preceding word and a word included in a succeeding sentence as a subsequent word in association with each other. Search means for searching for a subsequent word associated with the preceding word matching the word, and presentation means for presenting the subsequent word searched by the search means to the user. This is a text input support device.

  The present invention according to claim 13 analyzes two consecutive sentences included in an existing sentence, specifies a word included in a preceding sentence as a preceding word, and specifies a word included in a succeeding sentence as a subsequent word. For the sentence being entered in the sentence being created based on the user's operation input by the specifying means, the subsequent word corresponding to the preceding word that matches the word included in the sentence preceding the sentence is presented to the user. And a storage means for storing the preceding word and the succeeding word specified by the specifying means in association with each other in order to store them in the storage means.

  According to the first, fifth, and eleventh to thirteenth aspects of the present invention, when presenting a candidate that is predicted to be input subsequent to a sentence being input by a user who is creating a sentence, Compared to a case where the sentence is not applied, it is possible to present a candidate word based on the connection with the sentence preceding the sentence being input and to support the input.

  According to the second and sixth aspects of the present invention, a word included in a sentence following an existing sentence whose word group matches with a sentence preceding the sentence being input follows the existing sentence. It is possible to present in the order according to the conditional probability that the word appears in the sentence.

  According to the third and seventh aspects of the present invention, a word included in a sentence following an existing sentence having a matching word with a sentence preceding the sentence being input follows the existing sentence. It is possible to present in the order according to the conditional probability that the word appears in the sentence.

  According to the fourth and tenth aspects of the present invention, it is possible to search for candidate words that are predicted to be input in succession in consideration of the case of the word included in the sentence being input.

  According to the present invention of claim 8, the sentence appearance amount and the word appearance amount used for calculating the conditional probability can be calculated in consideration of the word weight in the sentence as a basis thereof.

  According to the present invention of claim 9, it is possible to calculate the word appearance amount used for calculating the conditional probability taking into account that words of the same case tend to appear in successive sentences.

It is a figure which shows the functional block of the 1st structural example of the text input assistance system which concerns on one Embodiment of this invention. It is a figure which illustrates the result of morphological analysis. It is a figure which illustrates the dictionary concerning the 1st example of composition. It is a figure which shows the functional block of the 2nd structural example of the text input assistance system which concerns on one Embodiment of this invention. It is a figure which illustrates the dictionary which concerns on a 2nd structural example. It is a figure which shows the functional block of the 3rd structural example of the text input assistance system which concerns on one Embodiment of this invention. It is a figure which illustrates the result of a parsing. It is a figure which illustrates the dictionary which concerns on a 3rd structural example. It is a figure which shows the functional block of the 4th structural example of the text input assistance system which concerns on one Embodiment of this invention. It is a figure which illustrates the hardware constitutions of the computer which operate | moves as a text input assistance system which concerns on one Embodiment of this invention. It is a figure explaining a text input support function. It is a figure explaining a text input support function.

Prior to specific description of the present invention, the text input support function will be described with reference to FIGS.
In the example of FIG. 11, a character string “in a diffusion weighted image” is input to a text input unit 51 where a user inputs a text sentence to create a sentence. At this time, with the text input support function, a character string that is likely to follow the character string before the current cursor position is estimated, and each character string obtained by the estimation is used as a complement candidate in the child window 52. Are presented (displayed). The user can select one of the supplement candidates presented in the child window 52 by operating the keyboard, the mouse, or the like, and can add the supplement candidate character string without typing. In other words, it is possible to create a sentence with less effort than typing the character string of the candidate for completion.

Here, consider a scene in which one or more sentences are input and then a subsequent sentence is input.
In FIG. 12, when the user inputs several characters of the next sentence after inputting one sentence in the conventional sentence input support function, the character string that will be input after that is estimated and presented. The scene is illustrated. In the example of FIG. 12, in the text input unit 61, the first sentence “A prominent high signal area is seen in the upper part of the right lung as a T1-weighted image” and the beginning of the second sentence “T of the part” are displayed. Are input, and as a complement candidate character string to be added to the continuation of the second sentence (the sentence being input), “1 emphasis image”, “2 emphasis image”, “1 region” are displayed in the child window 52. ”Are presented. That is, when the user inputs “T of the part”, the text input support function allows the user to input two to three words or several characters (“T of the part” in this example) immediately before the user inputs. With reference to the column, supplemental candidates such as “1 weighted image”, “2 weighted image”, “1 region”, which are estimated to appear in succession, are presented.

In presenting the above-mentioned complementation candidates, the conventional text input support function, for example, an N-gram model, for example, a score at which the Nth word appears when the immediately preceding N−1 words appear (eg, probability) Is statistically estimated from a large amount of text (sentences) prepared in advance.
However, the “T1-weighted image” has already been described in the previous sentence (first sentence) preceding the sentence (second sentence) being input, and the description regarding the “T1-weighted image” is not repeated. Hard to think. Further, in the MRI (Magnetic Resonance Imaging) imaging method, a “T1 weighted image” and a “T2 weighted image” are usually obtained at the same time, and descriptions of both images are often made. Therefore, in this case, it is desirable that “2 emphasized image”, which is a character string in the latter half of “T2 emphasized image”, is displayed at the top as a complement candidate.
Therefore, in the present invention, a word that is estimated to be input to the user is (character string) using the information of the word that is input to the sentence preceding the sentence being input (positionally preceding sentence). Can be presented more effectively.

A first configuration example of a text input support system according to an embodiment of the present invention will be described.
FIG. 1 shows functional blocks of the text input support system of the first configuration example.
The text input support system of the present example includes a text input support unit 30 that presents a candidate for completion based on a word included in a sentence preceding the sentence being input for a text that is being created based on an operation input by a user; The reference information creating unit 10 creates a dictionary (dictionary DB 21) that is referred to by the text input support unit 30 when presenting the complement candidates based on existing text.

  The reference information creation unit 10 according to the first configuration example includes a sentence division unit 12 that divides sentences included in a plurality of existing documents stored in the document DB 11 into sentence units, morphologically analyzes the target sentence, An inter-sentence N-gram analyzing unit 13 that analyzes a relationship between sentences (hereinafter referred to as inter-sentence N-gram) obtained by the sentence dividing unit 12 using a morpheme analyzing unit 14 that identifies words included in And a sentence-word count unit 16 for calculating a score for determining the priority order of the words to be listed, and a dictionary created as a result of processing by these function units is stored in the dictionary DB 21.

  Moreover, the text input support unit 30 according to the first configuration example precedes the text acquisition unit 31 that acquires the text (text) being created by the user, and the text being input in the text acquired by the text acquisition unit 31. The sentence (previous sentence) obtained by the previous sentence acquiring part 32 is obtained by using the preamble acquiring part 32 for acquiring a sentence (previous sentence) and the morpheme analyzing part 14 for morphologically analyzing the target sentence and specifying words included in the sentence. Based on the analysis result of the preamble analysis unit 33 to be analyzed and the analysis result of the preamble analysis unit 33, the candidate DB enumeration unit 35 that searches the dictionary DB 21 and identifies a word that is a candidate for complementation and the candidate for completion obtained by the complementation candidate enumeration unit 35 Functional units such as a complement candidate evaluation unit 37 that determines the priority order for each word, and a complement candidate presentation unit 38 that presents the complement candidate words in the priority order determined by the complement candidate evaluation unit 37 It has, to help the user of the text input by adding the words that have been selected from among the presented completions in continuation of the sentence in the input.

Here, the intergram N-gram will be described. The inter-sentence N-gram in this example has a score (for example, score (W | sentence 1,..., Score) in which the word W is present in the Nth sentence following the N−1th sentence. Sentence N-1)) is calculated and used for the presentation of the complementary candidates.
Hereinafter, the case where N = 2 (that is, the case where only the immediately preceding sentence is considered with respect to the sentence being input) will be described by focusing on independent words (words having meaning alone such as nouns and verbs).
For example, the frequency at which a certain sentence s appears in a large amount of sentences (sentence appearance frequency) is cs (s), and the frequency at which the word W appears in the next sentence after the sentence s appears (word appearance frequency) is cw ( s, W), the score relating to the presentation of the complement candidate is the ratio of the word appearance frequency cw (s, W) to the sentence appearance frequency cs (s), that is, the word in the next sentence in which a certain sentence s appears An estimated value p (W | s) = cw (s, W) / cs (s) of the conditional probability that W appears can be used.

The sentence appearance frequency cs (s) and the word appearance frequency cw (s, W) are obtained in advance by the following procedure by each function of the reference information creation unit 10.
First, prepare a large amount of text. This can use external resources such as Web texts and electronic newspaper articles, and texts created by each user who uses the text input support system of this example. In this example, it is assumed that these sentences are collected in advance and stored in the document DB 11. Then, sentences are extracted from these sentences in units of two consecutive sentences, and the following processing is performed.

  For example, the i-1st sentence in a certain sentence is "a high signal area is prominent in the T1-weighted image in the upper right lung." If it is not seen, ”when these sentences are divided into sentence units by the sentence dividing unit 12 and morphological analysis is performed by the morpheme analyzing unit 14, each sentence is included in the sentence as illustrated in FIG. Is identified. The morpheme analysis unit 14 of this example performs part-of-speech assignment as well as word division. In the example of FIG. 2, the sentence (i-1) which is the i-1th sentence and the sentence (i) which is the ith sentence. For each sentence, the word breaks are shown as blanks, and the independent words are underlined.

The inter-sentence N-gram analysis unit 13, based on the analysis result by the morphological analysis unit 14, a preceding word group in which each word appearing in the sentence (i-1) (hereinafter referred to as a preceding word) is grouped in sentence units, Each word appearing in i) (hereinafter referred to as a subsequent word) is identified, and the sentence-word counting unit 16 generates a sentence appearance of a sentence including the preceding word group in each sentence in the document DB 11 for each preceding word group. The frequency cs (s) is calculated, and the word appearance frequency cw (s, W) of the subsequent word in the sentence following the sentence including the preceding word group is calculated for each combination of each preceding word group and each subsequent word. calculate.
The sentence-word counting unit 16 in this example calculates a sentence appearance frequency cs (s) by adding a predetermined value (1 in this example) for each sentence including the corresponding preceding word group. Also, a word appearance frequency cw (s, W) is calculated by adding a predetermined value (1 in this example) for each combination of the corresponding preceding word group and succeeding word.

  As a result of the processing described above, a dictionary of intergram N-grams as illustrated in FIG. 3 is created and stored in the dictionary DB 21. In this example, as shown in FIG. 3A, the first dictionary in which the word appearance frequency cw (s, W) related to the correspondence (combination) between the preceding word group and the succeeding word is set, and FIG. As shown in FIG. 2, a second dictionary in which the sentence appearance frequency cs (s) related to the preceding word group is set is created.

Next, as an example of sentence input support using intergram N-grams, the user enters the sentence "A prominent high-signal area is visible in the upper part of the right lung with a T1-weighted image. The case will be described.
First, from the sentence being created by the user acquired by the text acquisition unit 31, the previous sentence acquisition unit 32 extracts a sentence preceding the sentence at the cursor position or the sentence in which the last character is input in the sentence. In the case of this example, the sentence “A prominent high signal area is seen in the T1-weighted image in the upper right lung” is extracted.
Then, the preamble analysis unit 33 performs a morpheme analysis on the sentence obtained by the preamble acquisition unit 32 by the morpheme analysis unit 14 and converts it into an expression for performing a database search for the dictionary DB 21. In the case of this example, it is converted into the word group “right lung—upper part—T1-weighted image—famous—high signal area—seen”.

Next, the complement candidate listing unit 35 searches the dictionary DB 21 using the word group obtained by the preamble analysis unit 33, specifies the subsequent word associated with the preceding word group matching the word group, The sentence appearance frequency cs (s) related to the preceding word group and the word appearance frequency cw (s, W) related to the combination of the preceding word group and the subsequent word are acquired.
Then, the complement candidate evaluation unit 37 calculates score = cw (s, W) / cs (s) for each subsequent word specified by the complement candidate listing unit 35, and the calculated score is equal to or greater than the threshold value. One or a plurality of words W is specified as a candidate for completion.

  Thereafter, the completion candidate presentation unit 38 displays each word of the completion candidate in the score order in the child window displayed near the cursor position (or the input position of the last character) in the text input unit (lower right part in this example). Enumerate and present it so that it can be selected by the user. The word presented as a candidate in the child window is added after the cursor position (or the input position of the last character) when selected by the user.

  Note that the database search using the word group obtained by the previous sentence analysis unit 33 may be performed while ignoring the arrangement order of the words included in the word group in the sentence. You may make it perform. When searching in consideration of the order of arrangement, information on the order of arrangement of the preceding words included in the preceding word group may be added to each preceding word group in the dictionary.

A second configuration example of the text input support system according to an embodiment of the present invention will be described.
FIG. 4 shows functional blocks of the text input support system according to the second configuration example.
The text input support system of this example has a configuration in which a word-word appearance count unit 17 is provided instead of the sentence-word appearance count unit 16 in the reference information creation unit 10 of the first configuration example.

  In the second configuration example, when a sentence s includes the words w1,..., Wk, the frequency at which sentences including wj appear in a large amount of sentences for all j satisfying 1 ≦ j ≦ k. (Sentence appearance frequency) is cs (wj), and the frequency (word appearance frequency) that the word W appears in the next sentence in which the sentence appears is cw (wj, W). The sum of the ratio of the word appearance frequency cw (wj, W) to the sentence appearance frequency cs (wj) is used.

  In the reference information creation unit 10 according to the second configuration example, the inter-sentence N-gram analysis unit 13 determines each preceding word that appears in the sentence (i-1) based on the analysis result by the morpheme analysis unit 14, and the sentence (i). And the word-word counting unit 17 calculates a sentence appearance frequency cs (w) of a sentence including the preceding word in each sentence in the document DB 11 for each preceding word, For each combination of each preceding word and each subsequent word, the word appearance frequency cw (w, W) of the subsequent word in the sentence following the sentence including the preceding word is calculated.

  In the word-word counting unit 17, a value determined by a predetermined formula for each sentence including the corresponding preceding word (in this example, a value obtained by dividing 1 by the number k of words in the sentence (1 / k)). Is added to calculate the sentence appearance frequency cs (w), and for each combination of the corresponding preceding and succeeding words, a value determined by a predetermined formula (in this example, 1 for the number of words k in the sentence) The word appearance frequency cw (w, W) is calculated by adding the value obtained by dividing ((1 / k)). That is, as the number of words in the sentence increases, a smaller value is added to calculate the sentence appearance frequency and the word appearance frequency, and as the number of words co-occurring in the same sentence increases, the value to be added is discounted. The weight of the word in each sentence is reflected in the sentence appearance frequency and the word appearance frequency.

  As a result of the above processing, a dictionary of intergram N-grams as illustrated in FIG. 5 is created and stored in the dictionary DB 21. In this example, as shown in FIG. 5A, the first dictionary in which the word appearance frequency cw (w, W) related to the correspondence (combination) between the antecedent word and the subsequent word is set, and in FIG. As shown, a second dictionary in which the sentence appearance frequency cs (w) related to the preceding word is set is created.

  In the text input support unit 30 according to the second configuration example, the previous sentence analysis unit 33 performs a morpheme analysis on the sentence obtained by the previous sentence acquisition unit 32 by the morpheme analysis unit 14 and converts it into an expression for performing a database search for the dictionary DB 21. To do. In this example, w1 = “right lung”, w2 = “upper part”, w3 = “T1-weighted image”, w4 = “famous”, w5 = “high signal area”, w6 = “seen” word wj ( 1 ≦ j ≦ k) is obtained.

Next, the complement candidate enumeration unit 35 searches the dictionary DB 21 using each word wj obtained by the preamble analysis unit 33 and specifies subsequent words respectively associated with the preceding words that match each word wj. The sentence appearance frequency cs (wj) related to the preceding word and the word appearance frequency cw (wj, W) related to the combination of the preceding word and the subsequent word are acquired.
Then, the complementation candidate evaluation unit 37 calculates a score for each subsequent word specified by the complementation candidate listing unit 35 according to the following (Equation 1), and the calculated score is one or more words that are equal to or greater than a threshold value. Specify W as a candidate for completion. Thereafter, the supplement candidate presentation unit 38 lists each candidate word in the order of score and presents it in a selectable manner by the user.

A third configuration example of the text input support system according to an embodiment of the present invention will be described.
FIG. 6 shows functional blocks of the text input support system according to the third configuration example.
The text input support system of this example has a configuration in which a syntax analysis unit 15 is added to the reference information creation unit 10 of the second configuration example.

In the above-described second configuration example, when creating the database (see FIG. 5) of the sentence appearance frequency cs (wj) and the word appearance frequency cw (wj, W), 1 / k was added uniformly. In the third configuration example, the syntactic analysis unit 15 performs syntactic analysis of the upper and lower two sentences to identify the dependency structure between words in the sentence, and adjusts the value to be added according to the result.
That is, in the third configuration example, (1) the subject or other case is often common between consecutive sentences, and (2) there is an expression pattern according to the flow of text description (example) 1; “Koke” is likely to appear in the next sentence after “Run”, but the reverse is unlikely to occur, eg, “T2 weighted image” is likely to appear after “T1 weighted image”). Based on the results of the syntactic analysis of the upper and lower two sentences, for combinations where the antecedent word and the succeeding word match, add to the word appearance frequency cw (wj, W) related to the succeeding word Increase the value to be

  In the reference information creation unit 10 according to the third configuration example, the inter-sentence N-gram analysis unit 13 determines each preceding word that appears in the sentence (i-1) based on the analysis result by the morpheme analysis unit 14, and the sentence (i). And the word-word counting unit 17 calculates a sentence appearance frequency cs (w) of a sentence including the preceding word in each sentence in the document DB 11 for each preceding word, For each combination of each preceding word and each subsequent word, the word appearance frequency cw (w, W) of the subsequent word in the sentence following the sentence including the preceding word is calculated. At this time, based on the analysis result by the syntax analysis unit 15, for a combination in which the case of the preceding word and the subsequent word match, the value added to the word appearance frequency cw (wj, W) related to the subsequent word is compared with other values. Adjust so that it becomes larger.

  For example, the syntax analysis result of each sentence illustrated in FIG. 2 can be expressed as shown in FIG. FIG. 7A illustrates the syntax analysis result of the sentence (i-1), and FIG. 7B illustrates the syntax analysis result of the sentence (i). The dependency structure between words is indicated by arrows. . According to FIGS. 7A and 7B, the sentence (i-1) has the main character “high signal area”, the procedural “upper right lung”, and the character “T1-weighted image” for the predicate “seen”. Sentence (i) has a dependency relationship with the predicate “seen” as the main character “abnormal”, the case “part”, and the character “T2 weighted image”. . In FIG. 7, the detailed dependency structure is omitted.

  Here, the words that make up each case, namely, the predicates “seen” and “seen”, the main cases “high signal area” and “abnormal”, the procedural “upper right lung” and “part”, the class “ The correspondence between the “T1-weighted image” and the “T2-weighted image” is specified and added so that the word appearance frequency cw (wj, W) related to the subsequent word in the pair (combination of the preceding word and the subsequent word) becomes large. Adjust the value. In this example, the first dictionary in which the word appearance frequency cw (w, W) related to the correspondence (combination) between the antecedent word and the subsequent word is set as the subsequent word in the combination having the same case as illustrated in FIG. The word appearance frequency cw (wj, W) is set to be doubled, but is not limited to this. For example, the value to be added is changed according to the type of matching cases. It may be.

A fourth configuration example of the text input support system according to an embodiment of the present invention will be described.
FIG. 9 shows functional blocks of a text input support system according to the fourth configuration example.
The text input support system of this example has a configuration in which a syntax analysis unit 35 is added to the text input support unit 30 of the third configuration example.
In the fourth configuration example, the word included in the sentence preceding the sentence being input and the succeeding word matched with the preceding word matching the case are presented as the complement candidates. In addition, in the dictionary DB 21 for searching for a complement candidate, as illustrated in FIG. 8, in the first dictionary in which the word appearance frequency cw (w, W) related to the correspondence (combination) between the preceding word and the succeeding word is set, Assume that the case of the preceding word (and the case of the subsequent word) is further stored in association with each other.

  In the sentence input support unit 30 according to the fourth configuration example, the preamble analysis unit 33 performs syntax analysis on the analysis result by the morpheme analysis unit 14 using the syntax analysis unit 35 to identify the case of each word, and the dictionary It converts into the expression for performing the database search with respect to DB21. As a result, for example, w1 = “right lung (legal)”, w2 = “upper part (legal)”, w3 = “T1 weighted image (grit)”, w4 = “famous”, w5 = “high signal area A word wj (1 ≦ j ≦ k) with case information “(main case)” and w6 = “seen (predicate)” is obtained.

Next, the complement candidate enumeration unit 35 searches the dictionary DB 21 using each word wj obtained by the preamble analysis unit 33, and selects each subsequent word associated with each word wj and the preceding word that matches the case. At the same time, the sentence appearance frequency cs (wj) related to the preceding word and the word appearance frequency cw (wj, W) related to the combination of the preceding word and the subsequent word are acquired.
Then, the complementation candidate evaluation unit 37 calculates a score for each subsequent word specified by the complementation candidate enumeration unit 35 according to (Equation 1) described above, and the calculated score is equal to or greater than a threshold value. Specify W as a candidate for completion. Thereafter, the supplement candidate presentation unit 38 lists each candidate word in the order of score and presents it in a selectable manner by the user.

  Here, the text input support system according to each configuration example described above includes a reference information creation device that operates as the reference information creation unit 10 and a text input support device that operates as the text input support unit 30 in separate devices. The dictionary created by the reference information creation device is configured to be distributed to each text input support device. However, the present invention is not limited to this. For example, each text input support device is held in the reference information creation device. The reference information creation unit 10 and the text input support unit 30 may be provided in an integrated apparatus.

  Moreover, in each structural example mentioned above, it is set as the structure which shows a complementation candidate based on two continuous sentences, However, It is good also as a structure which presents a complementation candidate based on three or more continuous sentences. That is, for example, in the case where the word wx included in the sentence preceding the sentence being input (the previous sentence) and the word wy included in the preceding sentence (the previous sentence), the existing sentence When a sentence sy including the word wy exists and a sentence sx including the word wx exists after the sentence sy, a word appearing in the sentence next to the sentence sx is presented as a complement candidate.

  Further, in each configuration example described above, the subsequent word associated with the preceding word is specified as a complementation candidate on condition that the word included in the sentence preceding the sentence being input matches the preceding word in the dictionary DB 21. However, for example, even if the words are different in expression, if they are synonyms or related words, they are regarded as matching words, and the subsequent words associated with the corresponding predecessors are identified as the candidate for completion. You may make it do.

  Further, in each configuration example described above, since it is assumed that the number of words specified as the complement candidates is relatively small, for example, the complement candidates obtained by each configuration example are replaced with the complement candidates obtained by the conventional method. May be merged (synthesized) and presented. Furthermore, the complement candidate scores obtained by the respective configuration examples and the complement candidate scores obtained by the conventional method may be summed, and the complement candidates may be presented in the order of the scores.

  Further, in the first to fourth configuration examples described above, in order to support document input in a wide range of fields, it is possible to collect a sentence without specifying a field and create a dictionary that is used for presenting supplement candidates. For example, based on medical documents such as medical records, a dictionary that is used exclusively when creating medical documents may be created, or based on internal documents such as reports and used exclusively when creating internal documents It is also possible to collect a specific type of document in which a fixed sentence is used, such as creating a dictionary, and create a dictionary exclusively used when creating the type of document. Moreover, you may make it create a dictionary for oneself based on the document which each user created by himself.

FIG. 10 shows the computer of the reference information creation device that operates as the reference information creation unit 10 and the text input support that operates as the text input support unit 30 in the text input support system according to the first to fourth configuration examples. The hardware configuration of the computer of the apparatus is illustrated.
The computer of this example includes a CPU (Central Processing Unit) 41 that performs various arithmetic processing, a RAM (Random Access Memory) 42 that is a work area of the CPU 41, a ROM (Read Only Memory) 43 that records basic control programs, and the like. Main storage device, auxiliary storage device such as HDD (Hard Disk Drive) 44 for storing programs and various data according to an embodiment of the present invention, display device for displaying and outputting various information, and used for input operations by an operator Hardware resources such as an input / output I / F 45 that is an interface with input devices such as operation buttons and a touch panel, and a communication I / F 46 that is an interface for performing wired or wireless communication with other devices. ing.
The program according to the embodiment of the present invention is read from the auxiliary storage device 44 and the like, loaded into the RAM 42, and executed by the CPU 41, thereby realizing the above-described functional units on the computer.

Note that the program according to the embodiment of the present invention is based on, for example, a computer according to this example depending on a format read from an external storage medium such as a CD-ROM storing the program or a format received via a communication network. Set to
Moreover, it is not restricted to the aspect which implement | achieves each function part by software configuration like this example, You may make it implement | achieve each function part with a dedicated hardware resource.

  11: Document DB, 12: Sentence division unit, 13: Inter-sentence N-gram analysis unit, 14: Morphological analysis unit, 15: Syntax analysis unit, 16: Word-sentence appearance count unit, 17: Word-word appearance count unit, 21: Dictionary DB, 31: Text acquisition unit, 32: Preamble acquisition unit, 33: Preamble analysis unit, 34: Morphological analysis unit, 35: Syntax analysis unit, 36: Completion candidate enumeration unit, 37: Completion candidate evaluation unit, 38 : Completion candidate presentation section

Claims (13)

On the computer,
In a sentence being created based on an operation input by a user, a specific function for analyzing a sentence preceding the sentence being input and identifying a word included in the sentence,
With respect to two consecutive sentences included in an existing sentence, a word included in a preceding sentence is used as a preceding word, and a word included in a succeeding sentence is stored in association as a subsequent word. A search function that searches for subsequent words associated with an antecedent that matches the identified word;
A presentation function for presenting subsequent words retrieved by the search function to a user;
A program to realize The storage means associates each preceding word group in which each preceding word included in a preceding sentence in an existing sentence is grouped in sentence units with each subsequent word included in a sentence following the sentence including the preceding word group. Along with the information, a sentence appearance amount indicating the appearance degree of the sentence including the preceding word group in the existing sentence, and a word appearance amount indicating the appearance degree of the subsequent word in the sentence following the sentence including the preceding word group are stored. And
The search function searches for a subsequent word corresponding to a preceding word group that matches a word group included in a sentence preceding the sentence being input, and includes a sentence appearance amount related to the preceding word group, the preceding word group, and the Search the word appearance amount related to the combination with the succeeding word,
The presenting function presents subsequent words searched by the search function in descending order of the ratio of the word appearance amount to the sentence appearance amount searched together with the subsequent word.
The program according to claim 1. The storage means stores the antecedent word in the existing sentence together with the correspondence information between each antecedent word included in the sentence preceding the existing sentence and each subsequent word included in the sentence subsequent to the sentence including the antecedent word. Storing the sentence appearance amount indicating the appearance degree of the sentence including, and the word appearance amount indicating the appearance degree of the subsequent word in the sentence subsequent to the sentence including the preceding word,
The search function searches for a subsequent word corresponding to each preceding word that matches each word included in a sentence preceding the sentence being input, and also includes a sentence appearance amount related to the preceding word and the preceding word and the subsequent word. Search for word appearance amount related to combination with word,
The presenting function presents subsequent words searched by the search function in descending order of the ratio of the word appearance amount to the sentence appearance amount searched together with the subsequent word.
The program according to claim 1. The storage means stores information on the case of the preceding word together with association information between each preceding word and each subsequent word included in a sentence following the sentence including the preceding word,
The specifying function specifies a case of a word included in a sentence preceding the sentence being input,
The search function searches for a subsequent word associated with a preceding word that matches a word and a case included in a sentence preceding the sentence being input.
The program according to claim 3. On the computer,
A specific function for analyzing two consecutive sentences included in an existing sentence, specifying a word included in a preceding sentence as a preceding word, and specifying a word included in a succeeding sentence as a subsequent word,
Sentence input support that presents subsequent words corresponding to a preceding word that matches a word included in a sentence preceding the sentence for the sentence being input based on an operation input by the user to the user A storage function that associates and stores the antecedent word and the subsequent word specified by the specific function in the storage means for processing;
A program to realize For each preceding word group that summarizes each preceding word included in the preceding sentence in the existing sentence in sentence units, a sentence appearance amount indicating the appearance degree of the sentence including the preceding word group in the existing sentence is calculated, and each For each combination of a preceding word group and each subsequent word included in a sentence following the sentence including the preceding word group, a word appearance amount indicating the appearance degree of the subsequent word in the sentence following the sentence including the preceding word group The computer further realizes a calculation function for calculating
In the sentence input support process, the subsequent word corresponding to the preceding word group that matches the word group included in the sentence preceding the sentence being input is represented as the word occurrence related to the subsequent word with respect to the sentence appearance amount related to the preceding word group. In order to present in the order of the ratio of the quantities, the storage function is calculated by the calculation function together with association information between each preceding word group and each subsequent word included in a sentence following the sentence including the preceding word group. Storing the generated sentence appearance amount and word appearance amount in the storage means;
The program according to claim 5. For each preceding word included in the preceding sentence in the existing sentence, the sentence appearance amount indicating the appearance degree of the sentence including the preceding word in the existing sentence is calculated, and subsequent to each preceding word and the sentence including the preceding word For each combination with each subsequent word included in the sentence to be performed, the computer further realizes a calculation function for calculating a word appearance amount indicating the appearance degree of the subsequent word in the sentence subsequent to the sentence including the preceding word,
In the sentence input support process, a word corresponding to each subsequent word corresponding to each preceding word that matches each word included in the sentence preceding the sentence being input is represented as the word appearance related to the subsequent word relative to the sentence appearance amount related to the preceding word. The storage function is calculated by the calculation function together with association information between each preceding word and each subsequent word included in a sentence following the sentence including the preceding word, in order to present in order of the ratio of the quantities. Storing the sentence appearance amount and the word appearance amount in the storage means;
The program according to claim 5. The calculation function calculates the sentence appearance amount for each preceding word by adding a smaller value as the number of words in the sentence increases for each sentence including the preceding word. The calculation of the word appearance amount for each combination is performed by adding a smaller value as the number of words in the sentence increases for each sentence including the subsequent word that is a sentence subsequent to the sentence including the preceding word.
The program according to claim 7. The specifying function specifies a case of a preceding word included in a sentence preceding in an existing sentence and a case of a subsequent word included in a sentence following the sentence,
The calculation function increases a value to be added to a word appearance amount related to the succeeding word for a succeeding word whose case matches the identified preceding word,
The program according to claim 7 or 8, wherein The specifying function specifies a case of a preceding word included in a preceding sentence in an existing sentence,
In the sentence input support processing, in order to present a subsequent word corresponding to a preceding word that matches a word and a case included in a sentence preceding the sentence being input, the storage function is configured to display each preceding word and the preceding word. Storing information on the case of the preceding word in the storage unit together with the association information with each subsequent word included in the sentence subsequent to the sentence including the sentence;
The program according to any one of claims 7 to 9, characterized in that: A first specifying means for analyzing two consecutive sentences included in an existing sentence, specifying a word included in a preceding sentence as a preceding word, and specifying a word included in a succeeding sentence as a following word; A storage means for storing the preceding word and the succeeding word specified by the specifying means in association with each other;
In a sentence being created based on an operation input by a user, the sentence is identified by a second identification unit that analyzes a sentence preceding the sentence being input and identifies a word included in the sentence, and the second identification unit Sentence input support comprising: search means for searching a subsequent word associated with a preceding word that matches a word from the storage means; and presentation means for presenting the subsequent word searched by the search means to a user And
A text input support system characterized by comprising: In a sentence being created based on an operation input by a user, a specifying means for analyzing a sentence preceding the sentence being input and identifying a word included in the sentence;
With respect to two consecutive sentences included in an existing sentence, a word included in a preceding sentence is used as a preceding word, and a word included in a succeeding sentence is stored in association as a subsequent word. Search means for searching for subsequent words associated with the preceding word that matches the identified word;
Presenting means for presenting subsequent words retrieved by the retrieval means to the user;
A text input support device characterized by comprising: A means for analyzing two consecutive sentences included in an existing sentence, specifying a word included in a preceding sentence as a preceding word, and specifying a word included in a succeeding sentence as a subsequent word;
Sentence input support that presents subsequent words corresponding to a preceding word that matches a word included in a sentence preceding the sentence for the sentence being input based on an operation input by the user to the user Storage means for associating and storing the antecedent word and the subsequent word specified by the specifying means in the storage means for processing;
A reference information creating apparatus comprising:

Images