JP2008276459A - Input character string prediction device, input character string prediction program and electronic medical chart system - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an input character string prediction device and an input character string prediction program, capable of changing a presentation method of a prediction candidate list according to an input style to improve input efficiency. <P>SOLUTION: This input character string prediction device for generating the prediction candidate list by predicting a character string which a user tries to input based on a character string or a character key-input by the user, and making it be displayed has: an input style estimation means estimating the input style of the user based on a history of the key input of the user; and a prediction candidate display control means controlling a display method of the prediction candidate list based on the input style of the user estimated by the input style estimation means. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to an input character string prediction device and an input character string for generating and displaying a prediction candidate list by predicting a character string to be input by a user based on a character or a character string input by the user. The present invention relates to a prediction program and an electronic medical record system.

  With the spread of information devices such as computers, it has become common to store sentences as digitized text, and the burden of inputting character strings using a keyboard or the like has increased. Similarly, in the medical field, with the popularization of electronic medical records and the like, medical chart input from handwritten keyboards is being performed.

  In the field where quick input is required, there is a desire to reduce the load of character string input. There is an input character string predicting device as means for reducing the load of character string input. The input character string prediction device predicts a word or the like starting with the character or the character string by inputting the first character or character string of the word or the like to be input, and displays the predicted word or the like as a prediction candidate. Then, when the user selects a desired one from the displayed prediction candidates, the prediction candidate selected by the user is automatically input.

  Note that a display priority is set for the prediction candidates, and normally, a predetermined number of prediction candidates having a high display priority are displayed in the order of high display priority. This display priority is set based on, for example, the degree to which the prediction candidate is correct (correctness). This degree of correctness is determined based on the word appearance probability (word occurrence probability) determined by the language model.

  The input character string prediction device reduces the input load because the target character string can be input from a part of the target character string, such as the word to be input, without inputting all the target character strings. it can. It is also possible to predict and display candidate character strings following the selected character string based on the learning history. It is also possible to improve the prediction accuracy by storing a character string whose input is confirmed, such as a prediction candidate selected by the user, as a learning history in time series and referring to the stored learning history at the next and subsequent predictions. .

By the way, in Japanese Patent Laid-Open No. 7-182343, the key input skill of the user is estimated based on the key input history of the user, and a method for generating candidates for kana-kanji conversion is controlled according to the estimated key input skill. Is disclosed. The key input skill is estimated based on the key input speed, the frequency of use of control keys such as the backspace key, and the like. When it is estimated that the key input skill is low, for example, the input “sinyou” is replaced with “shinyo” instead of “shinyo” and a kana-kanji conversion candidate is presented. In addition, people with low key input skills often have confusion among minus "-", long sound "-", hyphen "-", underscore "_", etc., and conversion candidates that include these input characters If the character does not exist, kana-kanji conversion is performed by appropriately replacing the character with another character that is easily confused. On the other hand, for those who have high key input skills, the input contents are trusted and the input characters are not replaced.
JP-A-7-182343

  By the way, this inventor asks a character input person who is inputting characters using the input character string predicting device, and the character input user inputs characters using the input character string predicting device. As a result of observation, it was discovered that there are the following two styles A and B as input styles of the input person.

  Style A: A style in which a relatively short prediction candidate is selected when a relatively small number of characters are input. A user whose input style is style A is a type in which a prediction candidate is expected even with a small number of inputs, and a sentence is composed while synthesizing short words.

  Style B: A style in which a relatively long prediction candidate is selected when a relatively large number of characters are input. A user whose input style is style B is a type that expects a long prediction candidate with an input corresponding to a word or phrase unit.

  For a person whose input style is style A, for example, it is preferable to display a prediction candidate list even if the input is even one character. On the other hand, for a person whose input style is style B, for example, a prediction candidate list is displayed when the number of input characters exceeds a certain length or when the input character string is in units of words and phrases. It is preferable. As described above, it is considered that the input efficiency is higher when the prediction candidate list presentation method is changed according to the input style.

  An object of the present invention is to provide an input character string prediction device, an input character string prediction program, and an electronic medical record system that can change the method of presenting a prediction candidate list according to the input style and can improve the input efficiency.

  According to the first aspect of the present invention, an input character string for generating and displaying a prediction candidate list by predicting a character string to be input by the user based on a character or a character string key-input by the user In the prediction device, an input style estimation means for estimating a user input style based on a user key input history, and a prediction candidate list display method based on the user input style estimated by the input style estimation means A prediction candidate display control means for controlling is provided.

  According to a second aspect of the present invention, in the input character string predicting device according to the first aspect, the input style estimating means inputs a relatively small number of characters in the user input style based on the user key input history. It is estimated whether it is the first input style for selecting a relatively short prediction candidate at the time, or the second input style for selecting a relatively long prediction candidate when a relatively large number of characters are input. It is characterized by being.

  According to a third aspect of the present invention, in the input character string prediction apparatus according to the second aspect, the prediction candidate display control means is estimated by the input style estimation means that the user input style is the first input style. If the input style of the user is estimated to be the second input style by the means for displaying the prediction candidate list and the input style estimation means each time a reading is input, Means is provided for displaying the prediction candidate list only when the number of reading characters is a predetermined number of 2 or more and the reading character string is in units of words or phrases.

  According to a fourth aspect of the present invention, in the input character string prediction device according to the third aspect, the prediction candidate display control means is estimated by the input style estimation means that the user input style is the first input style. When the prediction candidate list is displayed in the case where the user input style is estimated to be the second input style by the means for displaying the word-by-word prediction candidate and the input style estimation means, the prediction candidate When the list is displayed, a means for displaying prediction candidates in compound word units and / or sentence units is provided.

  According to a fifth aspect of the present invention, in the input character string prediction device according to the fourth aspect, the prediction candidate display control means is estimated by the input style estimation means that the user input style is the first input style. The number of prediction candidates displayed as the prediction candidate list is less than the number of prediction candidates displayed as the prediction candidate list when the user input style is estimated to be the second input style. And a means for controlling the number of prediction candidates displayed as a prediction candidate list according to the input style.

  According to a sixth aspect of the present invention, in the first to fifth aspects of the present invention, the input style estimating means includes the key input speed by the user, the frequency of use of the backspace key, and the length of reading when a prediction candidate is selected. In addition, the user's input style is estimated based on one or any combination arbitrarily selected from the frequency of reading in units of words or phrases when a prediction candidate is selected. And

  The invention according to claim 7 is an input character string for generating and displaying a prediction candidate list by predicting a character string to be input by the user based on a character or character string key-input by the user. A prediction program comprising: a computer, an input style estimation unit that estimates a user input style based on a user key input history; and a prediction candidate based on a user input style estimated by the input style estimation unit It is a program for making it function as a prediction candidate display control means which controls the display method of a list, It is characterized by the above-mentioned.

  According to an eighth aspect of the present invention, the electronic medical record system includes a medical record DB that stores a patient name and medical record contents in association with each other, and an input character string predicting device according to any one of the first to seventh aspects. And

  According to the present invention, the prediction candidate list presentation method can be changed according to the input style, and the input efficiency can be improved.

  DESCRIPTION OF THE PREFERRED EMBODIMENTS Embodiments of the present invention applied to an electronic medical chart system will be described below with reference to the drawings.

[1] Configuration of Input Character String Prediction Device FIG. 1 shows a configuration of an input character string prediction device.

  The input character string prediction device 10 includes an input character string prediction unit 11, a key log storage unit 12, an input style estimation unit 13, a prediction candidate display control unit 14, and a medical record database (medical record DB) 15.

[1] Input character string prediction unit 11
The input character string prediction unit 11 includes a prediction dictionary, a learning dictionary, and a learning history storage memory. The prediction dictionary stores words, compound words, and sentences that are candidates for prediction. Each prediction candidate object stored in the prediction dictionary is added with a score representing the degree to which the prediction candidate object is correct (degree of correctness). This score is determined based on, for example, a word appearance probability (word occurrence probability) determined by the language model.

  In the learning dictionary, a prediction candidate selected by the user from the prediction candidate list, a character string that has been confirmed to be input, and the like are stored as a learning history. When the input character string prediction apparatus is powered on, the learning history stored in the learning dictionary is read into the learning history storage memory. Of the learning histories read into the learning history storage memory, learning histories for which a predetermined period (for example, six months) has elapsed since registration are deleted.

  After power-on, every time an input is confirmed, the notation and reading of the character string whose input is confirmed are stored in the learning history storage memory in time series. Note that the number of character strings that can be stored in the learning history storage memory is finite. When the number of character strings to be stored exceeds a predetermined number, the old character string is deleted and a new character string is stored.

  The input character string predicting unit 11 selects a prediction candidate target (word, compound word, sentence) starting with the input character string (reading) every time a character is input with the keyboard 1, a prediction dictionary, and a learning history storage memory. Search from the learning history, and based on the score added to the searched prediction candidate target and the appearance frequency of the prediction candidate target in the learning history, the priority order of each searched prediction candidate target is determined and given priority A prediction candidate list is generated in descending order. The generated prediction candidate list is given to the prediction candidate display control unit 14 and displayed on the display 2 by the prediction candidate display control unit 14.

  Further, when a prediction candidate is selected from the prediction candidate list displayed on the display 2, the input character string prediction unit 11 performs connection prediction and generates a prediction candidate list. In other words, the same character string as the selected prediction candidate is searched from the learning history in the learning history storage memory, and the character string registered next (next in chronological order) after the searched character string is predicted. Search as a candidate. Then, a priority order is determined for each obtained prediction candidate based on the frequency searched this time, and a prediction candidate list is generated in descending order of priority. The generated prediction candidate list is given to the prediction candidate display control unit 14 and displayed on the display 2 by the prediction candidate display control unit 14.

  The input character string prediction unit 11 also outputs the selected prediction candidate to the input style estimation unit 13 when a prediction candidate is selected from the prediction candidate list displayed on the display 2.

[2] Key log storage unit 12
The key log storage unit 12 calculates key log information based on the key input history of the keyboard 1. The key log information includes a key input speed a, a backspace key usage frequency b, and a reading character count c when a prediction candidate is selected.

  The key input speed a is an average time interval of key input within a predetermined period. The use frequency b of the backspace key is a ratio (BN / TN) of the use count BN of the backspack key to the use count (total key input count) TN of all keys within a predetermined period. The number of reading characters c when selecting a prediction candidate is an average value of the number of characters stored as a reading when a prediction candidate is selected within a predetermined period. These key log information a, b, and c are updated every predetermined time.

[3] Input style estimation unit 13
The input style estimation unit 13 estimates the input style based on the key log information calculated by the key log storage unit 12 and the selected prediction candidate given from the input character string prediction unit 11.

  The input style estimation method will be described more specifically.

The input style estimation unit 13 first determines the following determination elements x ₁ to x ₁ based on the key log information calculated by the key log storage unit 12 and the selected prediction candidate given from the input character string prediction unit 11. to calculate the x _6.

x ₁ : Key input speed (average time interval of key input) a
x _2: use of the backspace key frequency b
x ₃ : Reciprocal 1 / c of the average value c within a predetermined period of the number of characters stored as a reading when the prediction candidate is selected
x _4: reciprocal number of input characters difference during the prediction candidate selection; the absolute value of the difference between the number of reading of characters and the selected prediction candidates are memory as read when the predicted candidates are selected, a predetermined time period value the number of characters memory divided by the number of characters read prediction candidates corresponding selected as reading when the predicted candidates are selected,; number of input characters ratio during the prediction candidate selection: the reciprocal of the average value in the inner x ₅ Average value within a predetermined period x ₆ : Reciprocal of frequency that reading at prediction candidate selection is word or phrase unit; d is the total number of prediction candidate selections within predetermined period, and reading at prediction candidate selection is word or If the number of times in the phrase unit is e, x ₆ = d / e.

When calculating x ₄ and x ₅ , the number of characters stored as a reading when the prediction candidate is selected is used as the number of characters stored as a reading, and the reading of the selected prediction candidate is used. The number of characters after conversion of the reading of the prediction candidate may be used as the number of characters. For example, a case is assumed in which the reading when a prediction candidate is selected is “Koten” and the reading of the selected prediction candidate is “Kotensenki Fuzen”. When the prediction candidate is selected, the number of characters of the reading “Koten” is 4, but since the character string after conversion is “acquired”, the number of characters after conversion is 2. The number of characters of the selected prediction candidate reading “Kotensei Ekifuzen” is 13, but since the converted character string is “acquired immune deficiency”, the number of characters after conversion is 7.

x ₁ is assumed to be slow for style A people (long average time interval) and fast for style B people (short average time interval).
It is assumed that x ₂ is large for style A people and small for style B people.
x ₃ is assumed to be large for style A people and small for style B people. The reason for this is c is the reciprocal of x ₃ short in human style A, since long by the human style B.
x ₄ is assumed to be large for style A people and small for style B people. This is because it is assumed that the “difference in the number of input characters when selecting a prediction candidate”, which is the reciprocal of x ₄ , is small for a style A person and large for a style B person.
x ₅ is greater in people of style A, is assumed to be small in the person of style B.
x ₆ is assumed to be large for style A people and small for style B people. This is because it is assumed that the reciprocal of x ₆ “the frequency with which the reading at the time of selecting a prediction candidate is in units of words or phrases” is small for the person of style A and large for the person of style B.

When the determination elements x _{1 to} x ₆ are calculated, the input style estimation unit 13 calculates a determination index Y from these determination elements x _{1 to} x ₆ . As the determination index Y, for example, a weighted average of the determination elements x _{1 to} x ₆ is used as shown in the following equation (1).

Y = Σw _i · x _i
= W ₁ x ₁ + w ₂ x ₂ + w ₃ x ₃ + w ₄ x ₄ + w ₅ x ₅ + w ₆ x ₆
... (1)

In the above formula (1), w _i is a predetermined weight for x _i .

As the determination index Y, a logarithmic weighted addition value of the determination elements x _{1 to} x ₆ may be used as shown in the following expression (2), or as shown in the following expression (3), it may be used a weighted multiplication value of x ₁ ~x _6.

Y = Σw _i · log x _i (2)

Y = Πw _i · x _i
= W ₁ x ₁ x w ₂ x ₂ x w ₃ x ₃ x w ₄ x ₄ x w ₅ x ₅ x w ₆ x ₆
... (3)

Note that the determination index Y is generally represented by a function F having determination elements x _{1 to} x ₆ as variables, as represented by the following equation (4).

_{Y = F (x 1, x} 2, x 3, x 4, x 5, x 6) ... (4)

  When the determination index Y is calculated, the input style estimation unit 13 compares the determination index Y with a predetermined threshold Th. If Y ≧ Th, the input style estimation unit 13 estimates that the input style is style A, and if Y <Th. , It is estimated that the input style is style B.

[4] Prediction candidate display control unit 14
The prediction candidate display control unit 14 controls the display method of the prediction candidate list generated by the input character string prediction unit 11 based on the input style estimation result by the input style estimation unit 13.

  FIG. 2 shows a prediction candidate display control processing procedure by the prediction candidate display control unit 14 when the input style is estimated to be style A.

  When the prediction candidate list is generated by the input character string prediction unit 11 (step S1), the top N pieces (for example, N = 3) of the prediction candidate list generated by the input character string prediction unit 11 are displayed on the display 2. (Step S2). Then, the process returns to step S1.

  When N prediction candidates are displayed in step S2, it is preferable to select and display the top N prediction candidates in word units from the prediction candidates included in the prediction candidate list.

  FIG. 3 shows a display example of prediction candidates when the display control process of FIG. 2 is performed. As shown in FIG. 3, every time a character is read, a prediction candidate is displayed. The number of prediction candidates displayed is as small as three.

  FIG. 4 shows a prediction candidate display control processing procedure by the prediction candidate display control unit 14 when the input style is estimated to be style B.

  When the prediction candidate list is generated by the input character string prediction unit 11 (step S11), whether or not the display condition that the number of reading characters is 3 or more and the reading character string is in units of words or phrases is satisfied. Is determined (step S12). This determination is performed as follows, for example. That is, word information and part-of-speech information for each word are stored in the prediction dictionary. Whether or not the character string of the reading is a word unit is determined based on whether or not the character string is registered as a word in the prediction dictionary. Whether the reading character string is a phrase unit is whether the character string is a combination of a word registered in the prediction dictionary and a specific particle registered in the prediction dictionary (word + specific particle). Judgment based on. If it is determined in step S12 that the display condition is not satisfied, the process returns to step S11 without displaying the prediction candidate.

  If it is determined in step S12 that the display condition is satisfied, the top M prediction candidate lists generated by the input character string prediction unit 11 (M is greater than N, eg, M = 20) are displayed on the display 2. (Step S13). Then, the process returns to step S11.

  When M prediction candidates are displayed in step S13, it is preferable to select and display the top M prediction candidates in compound words or sentence units from the prediction candidates included in the prediction candidate list.

  FIG. 5 shows a display example of prediction candidates when the display control process of FIG. 3 is performed. As shown in FIG. 5, many prediction candidates are displayed only when the number of reading characters is 3 or more and the reading character string is in units of words or phrases.

  Note that a prediction dictionary for style A and a prediction dictionary for style B are prepared in advance as prediction dictionaries used by the input character string prediction unit 11 to generate a prediction candidate list, and are estimated by the input style estimation unit 13. The prediction dictionary used by the input character string prediction unit 11 may be switched according to the style set. In this case, as a prediction dictionary for style A, a dictionary storing words that are candidates for prediction candidates is prepared, and as a prediction dictionary for style B, compound words and sentences that are candidates for prediction candidates are stored. Prepare a dictionary. In each prediction dictionary, word information and part-of-speech information of each word are stored in order to determine whether or not the reading character string is in units of words or phrases.

  Also in this case, when the input style estimation unit 13 estimates that the input style is style A, the prediction candidate display control unit 14 performs the display control process of FIG. 2 and the input style estimation unit 13 performs the input style. Is estimated to be style B, the display control process of FIG. In this way, when the input style is estimated to be style A, prediction candidates are displayed in units of words, and when the input style is estimated to be style B, prediction is performed in units of compound words or sentences. Candidates will be displayed.

[5] Medical record DB15
In the medical record DB 15, patient names and medical record contents are recorded in association with each other. The patient name is data for specifying a patient, and the name of the patient may be used, or an identification number or an identification symbol that can identify the patient may be used. The medical record DB 15 stores a character string whose input is confirmed every time input is confirmed, such as when a prediction candidate is selected by the user from the prediction candidate list. In addition, when the input of the entire electronic medical record for a certain patient is completed, the contents of the medical record may be stored in the medical record DB 15.

It is a block diagram which shows the structure of an input character string prediction apparatus. It is a flowchart which shows the prediction candidate display control processing procedure by the prediction candidate display control part 14, when it is estimated that the input style is style A. It is a schematic diagram which shows the example of a display of a prediction candidate at the time of performing the display control process of FIG. It is a flowchart which shows the prediction candidate display control processing procedure by the prediction candidate display control part 14, when it is estimated that the input style is the style B. It is a schematic diagram which shows the example of a display of a prediction candidate at the time of performing the display control process of FIG.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 Input character string prediction apparatus 11 Input character string prediction part 12 Key log storage part 13 Input style estimation part 14 Prediction candidate display control part 15 Medical record DB

Claims (8)

In an input character string prediction device for generating and displaying a prediction candidate list by predicting a character string that a user intends to input based on a character or character string keyed by the user,
An input style estimation unit that estimates a user's input style based on a user's key input history, and a prediction candidate that controls a display method of a prediction candidate list based on the user's input style estimated by the input style estimation unit Display control means,
An input character string predicting device comprising: The input style estimation means is a first input style for selecting a relatively short prediction candidate when a relatively small number of characters are input based on the user's key input history. The input character string prediction apparatus according to claim 1, wherein the input character string prediction device estimates whether the second input style selects a relatively long prediction candidate when a large number of characters are input. The prediction candidate display control means is a means for displaying a prediction candidate list every time a reading is input when the input style of the user is estimated to be the first input style by the input style estimation means. When the input style of the user is estimated to be the second input style by the input style estimating means, the number of reading characters is a predetermined number of 2 or more and the reading character string is in units of words or phrases The input character string prediction apparatus according to claim 2, further comprising means for displaying a prediction candidate list only when The prediction candidate display control means displays the word-by-word prediction candidates when displaying the prediction candidate list when the input style estimation means estimates that the user input style is the first input style. When the prediction style list is displayed when the input style of the user is estimated to be the second input style by the input style estimation means, the prediction candidates in compound word units and / or sentence units are displayed. The input character string prediction apparatus according to claim 3, further comprising: means. The prediction candidate display control means is configured so that the number of prediction candidates displayed as the prediction candidate list when the input style of the user is estimated to be the first input style by the input style estimation means, Means for controlling the number of prediction candidates displayed as a prediction candidate list according to the input style so that the number of prediction candidates displayed as the prediction candidate list is less than the number of prediction candidates displayed when the input style is estimated to be the second input style The input character string prediction apparatus according to claim 4, comprising: The input style estimation means determines the speed of key input by the user, the frequency of use of the backspace key, the length of reading when a prediction candidate is selected, and the frequency at which the reading is in units of words or phrases when a prediction candidate is selected. 6. The input character string prediction apparatus according to claim 1, wherein the input style of the user is estimated based on one or any combination arbitrarily selected from among them. An input character string prediction program for generating and displaying a prediction candidate list by predicting a character string that the user intends to input based on a character or character string keyed by the user,
Computer
An input style estimation unit that estimates a user's input style based on a user's key input history, and a prediction candidate that controls a display method of a prediction candidate list based on the user's input style estimated by the input style estimation unit Display control means,
Input string prediction program to function as An electronic medical record system comprising a medical record DB that stores a patient name and medical record contents in association with each other, and an input character string predicting device according to claim 1.

Images