JP2019008615A - Document reconstruction device - Google Patents

Abstract

To provide a document reconstruction device capable of appropriately canceling the column setting of a document arranged in columns and generating a document easy to browse even with a small terminal.SOLUTION: A document reconstruction device includes a dividing unit 21 configured to divide an original document arranged in columns into a plurality of regions based on a predetermined region determination condition (for example, a boundary line or a blank), a sentence extraction unit 22 configured to extract sentences included in respective divided regions, a discrimination unit 23 configured to discriminate whether a sentence extracted from each region is a sentence linked to a sentence extracted from another region or an independent sentence, a sentence coupling unit 24 configured to connect sentences discriminated to be connected to sentences extracted from other regions into one sentence, and a reconstruction unit 25 configured to arrange the independent sentence and the sentences linked into one sentence by the sentence coupling units in a row and reconstruct the original document into a document whose column setting has been cancelled.SELECTED DRAWING: Figure 2

Description

  The present invention relates to a document reconstruction apparatus and program for reconfiguring a document by releasing the column of the document that has been arranged in columns.

  With the spread of small electronic terminals such as smartphones and tablet PCs, opportunities to browse documents on such small terminals are increasing. However, since smartphones and tablet PCs have small displays, when browsing a document such as a book or document, the document must be read while repeatedly enlarging / reducing and scrolling up and down. In particular, in order to view a columned document such as a newspaper or a magazine, the above-described repetitive operation is more required.

  For example, when a column-structured document as shown in FIG. 24 is viewed on a small electronic terminal, as shown in FIG. 25, the entire document is displayed in a full-screen display format on the screen by default. Since the characters are small and in columns as they are, the viewer needs to browse the document while enlarging / reducing the document and scrolling up and down. As a result, as indicated by the arrows in FIG. 26, it is necessary to scroll the screen horizontally or vertically, which makes the operation complicated and not convenient.

  As a technique for dealing with such a problem, Patent Document 1 below discloses a system that generates a second document by rearranging a plurality of text groups constituting the first document arranged in a line into one column. It is disclosed. In this system, according to the order in which the first document is read by humans, each text group of the first document in which a plurality of text groups composed of a plurality of vertical rows are arranged side by side is arranged. The second document is generated by arranging the text groups vertically in ascending order of ranking.

JP 2017-49865 A

  In Patent Document 1, a plurality of text groups constituting the first document are given a rank according to the order in which the first document is read by a person. There is no disclosure as to whether it is specified. For example, in newspapers and the like, the page is composed of complicated columns, so it is difficult to correctly find the order in which humans read and to arrange each text group in a line in an appropriate order. Therefore, in the first document, even if one text group and one other text group are originally one sentence, there are cases where they are not arranged consecutively and cannot be read correctly as sentences. Occurs.

  The present invention provides a document reconstruction device that can properly cancel a column of documents by connecting them even when a single sentence is originally divided and arranged in a jumped place in a columned document. And aims to provide a program.

  The gist of the present invention for achieving the object lies in the inventions of the following items.

[1] A dividing unit that divides a columned original document into a plurality of areas based on a predetermined area discrimination condition;
A text extraction unit that extracts text included in each divided area;
A discriminator for discriminating whether the text extracted from each area after division is a text connected to text extracted from other areas or an independent text;
A sentence combining unit that connects sentences determined to be connected to sentences extracted from other areas to one sentence;
A reconstruction unit that arranges the independent sentences and the sentences connected to one by the sentence combination unit in a row, and reconstructs the original document into a column-unreleased document;
A document reconstruction apparatus characterized by comprising:

  In the above invention, the columned document is divided into a plurality of regions based on region discrimination conditions such as column boundary lines and blanks, and whether the texts in each region are connected to the texts in other regions or not. The connected ones are combined into one sentence, and then the sentences are arranged in a line in the normal reading order to reconstruct the document. For example, even if it is complicatedly arranged like a newspaper, it is possible to reconstruct a document by appropriately connecting the sentences that are divided and arranged at the skipped positions by determining the connection of the sentences.

[2] The document reconstruction apparatus according to [1], wherein the determination unit performs the determination by digitizing the appropriateness of connection between sentences and comparing with a predetermined threshold.

[3] The document reconstruction device according to [2], wherein the threshold value can be set by a user.

[4] The discriminating unit discriminates based on the similarity of the content of the sentence and / or the continuity between the end of one sentence and the beginning of another sentence. [1] ] The document reconstruction device described in the above.

[5] The determination unit does not determine a connection between a sentence included in one area and a sentence included in an area at a position that is not likely to be continuous with the sentence included in the one area. The document reconstruction device according to [1].

  In the said invention, a processing burden is reduced by narrowing down the object which discriminate | determines a connection.

[6] The document reconstruction device according to [1], wherein the reconstruction unit arranges the sentences in a line according to a direction in which the sentence is read.

  In the above invention, sentences are arranged in the reading direction in line units. Thus, if the reconstructed document is scrolled in the reading direction of the sentence, the next sentence is naturally displayed.

[7] The document reconstruction device according to [1], wherein the reconstruction unit arranges the sentences in a line according to a direction designated by a user.

  In the above-described invention, the user can arbitrarily specify the arrangement direction of the sentences. Documents can be reconstructed according to user preferences.

[8] The reconstruction unit can select whether to perform the reconstruction by adjusting the character size while maintaining the layout of the sentence in each region, or to perform the reconstruction by reflow. The document reconstruction device according to [1].

  In the above invention, the document can be reconstructed in a form (layout) according to the user's preference.

[9] The document reconstruction device according to [8], wherein the user can specify a character size.

  In the above invention, the document can be reconstructed with a character size according to the user's preference.

[10] When an object such as an image or a graphic exists in one area in addition to a sentence, the reconstruction unit handles the object and the sentence included in the one area as a unit and arranges the objects. The document reconstruction device according to [1], wherein

  In the above invention, the correspondence between the sentence and the object is maintained.

[11] The document reconstruction device according to [1], wherein, when the original document is image data, the dividing unit divides the original document by determining an area by image processing.

[12] The document reconstruction according to [1], wherein, when the original document is a document described in a markup language, the dividing unit divides the document based on tag information indicating a column. Configuration equipment.

[13] The document reconstruction device according to [1], wherein when the original document is image data, the text extraction unit extracts text by character recognition.

[14] The document according to [1], wherein, when the original document is a document described in a markup language, the sentence extraction unit extracts a sentence based on tag information indicating a text area. Reconstruction device.

[15] A program for causing an information processing apparatus to function as any one of the document reconstruction apparatuses according to [1] to [14].

  According to the document reconstruction device and the program according to the present invention, even when a single sentence is originally divided and arranged at a place where it is skipped in a stacked document, the documents are appropriately connected to each other. Can be canceled appropriately.

It is a figure which shows an example of the document browsing system containing the document reconstruction apparatus which concerns on this invention. It is a block diagram which shows schematic structure of a document reconstruction apparatus. 10 is a flowchart illustrating an outline of processing in which a document reconstruction device releases a column of an original document and creates a reconstructed document. It is a figure which shows the state which divided and labeled the original document of Example 1, and this into several area | regions on the basis of a boundary line. It is a figure which shows the state which divided and labeled the original document of Example 2, and this into several area | regions on the basis of the blank. 4 is a flowchart showing a document reconstruction process (details of step 107 in FIG. 3). FIG. 5 is a diagram illustrating a reconstructed document obtained by reconstructing the original document of Example 1 shown in FIG. 4. 10 is a diagram illustrating an original document of Example 3. FIG. It is a figure which shows the state which divided | segmented the original document of Example 3 into the area | region 1-the area | region 4, and was labeled. FIG. 9 is a diagram illustrating a reconstructed document obtained by reconstructing the original document of Example 3 illustrated in FIG. 8. 10 is a diagram showing an original document of Example 4. FIG. It is a figure which shows the state which divided | segmented and labeled the original document of Example 4 into several area | regions. It is a figure which shows the reconstruction document of the result of having reconstructed the original document of Example 4. 10 is a diagram showing an original document of Example 5. FIG. It is a figure which shows the reconstruction document of the result of having reconstructed the original document of Example 5. 10 is a diagram showing an original document of Example 6. FIG. It is a figure which shows the reconstruction document of the result of having reconstructed the original document of Example 6. FIG. 10 is a diagram showing a reconstructed document of Example 7. FIG. 10 is a diagram showing a reconstructed document of Example 8. 10 is a diagram showing an original document of Example 9. FIG. It is a figure which shows the reconstruction document of the result of having reconstructed the original document of Example 9. 10 is a diagram showing an original document of Example 10. FIG. FIG. 10 is a diagram showing a reconstructed document of Example 10. It is a figure which shows the example of the document put in a column. It is a figure which shows a mode that the whole document of the column of FIG. 24 was displayed on the display part of a small portable terminal. It is a figure which shows a mode that the document shown in FIG. 24 is browsed with a small portable terminal.

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

  FIG. 1 is a diagram showing an example of a document browsing system 2 including a document reconstruction device according to the present invention. The document browsing system 2 includes a small portable terminal 5 such as a smartphone or a tablet used by a user, and a document reconstruction apparatus 10 that is a server (information processing apparatus) connected to the portable terminal 5 through a network. Configured.

  When the portable terminal 5 browses a columned document, the portable terminal 5 transmits data of the document (which is an original document) to the document reconstruction device 10 through the network, and requests cancellation of the column (P1). The document reconstruction device 10 cancels the column of the received original document, generates a reconstructed document that can be viewed only in a one-way school (P2), and transmits the reconstructed document to the mobile terminal 5. (P3). By browsing the reconstructed document on the portable terminal 5, it is possible to browse the document smoothly from the beginning to the end by simply performing a scrolling operation in one direction.

  Note that a program that performs the function of releasing the column by the document reconstruction device 10 may be installed in the mobile terminal 5 and the mobile terminal 5 may cancel the column setting of the stacked document (see FIG. 1 P4).

  FIG. 2 is a block diagram illustrating a schematic configuration of the document reconstruction device 10. The document reconstruction device 10 includes a CPU (Central Processing Unit) 11, a RAM (Random Access Memory) 12, a ROM (Read Only Memory), a storage unit 13 including a hard disk device, a network communication unit 14, an input I / O. The F unit 15 and the output I / F unit 16 are connected to each other.

  The CPU 11 has a microprocessor, and is based on an OS program, on which middleware and application programs are executed. The storage unit 13 stores various programs and data. The function is realized as the document reconstruction device 10 by the CPU 11 executing processing according to these programs. The RAM 12 is used as a work memory that temporarily stores various data when the CPU 11 executes processing.

  The network communication unit 14 has a function of communicating with the mobile terminal 5 and various external devices through the network. The network communication unit 14 receives the original document data and column release request from the mobile terminal 5. In addition, the network communication unit 14 transmits the reconstructed document to the mobile terminal 5.

  The input device 15 is a device for inputting user operations such as a keyboard and a mouse. The output device 16 is a display device such as a liquid crystal monitor. Various settings relating to column cancellation (for example, layout selection and character size selection described later) can be received from the input device 15 of the document reconstruction device 10 or from the mobile terminal 5.

  The CPU 11 functions as a dividing unit 21, a text extracting unit 22, a determining unit 23, a text combining unit 24, and a reconstruction unit 25 by executing a program.

  The dividing unit 21 divides the original document into a plurality of areas based on predetermined area determination conditions.

  The sentence extraction unit 22 extracts sentences (text group) included in each divided area.

  The determination unit 23 determines whether the sentence extracted from each area is a sentence connected to a sentence extracted from another area or an independent sentence.

  The sentence combining unit 24 connects sentences determined to be connected to sentences extracted from other areas to one sentence.

  The reconstruction unit 25 arranges the independent sentences and the sentences joined together by the sentence combination unit 24 in a line, and reconstructs the original document into a column-removed document.

  FIG. 3 is a flowchart showing an outline of processing in which the document reconstruction device 10 cancels the column of the original document and creates a reconstructed document. The document reconstruction apparatus 10 first inputs an original document (step S101). The input original document may be a document represented by a character code, or may be represented by image data such as bitmap data. In this example, it is assumed that the original document is image data obtained by reading a paper document with a scanner or the like.

  Next, the CPU 11 (dividing unit 21) of the document reconstruction device 10 divides the original document into a plurality of areas based on a predetermined area discrimination condition, and performs labeling on each divided area (step S102). .

  The area discrimination condition is a condition for discriminating a range in which characters exist together as one area, such as the presence of a column boundary or a blank area of a certain level or more. As shown in FIG. 4, when a column boundary is drawn in the columned original document, the document is divided into a plurality of areas based on the line. FIG. 4B shows a state where the original document of Example 1 is divided and labeled. A broken line in the figure indicates a region. The original document in FIG.

  When there is no boundary line, as shown in FIG. 5, it divides | segments into a some area | region on the basis of the blank area | region provided between sentences and between each stage of a column. More specifically, the line spacing of each character string is confirmed, and if there is a certain difference in spacing between before and after, it is determined that the space is blank. For example, if the interval becomes 30 pt after the interval continues to 12 pt, 12 pt,..., It is determined that there is a blank area delimited at the position where the interval becomes 30 pt. FIG. 5B shows a state where the original sentence of Example 2 is divided and labeled. A broken line in the figure indicates a region. The original document in FIG.

  Next, the CPU 11 (sentence extraction unit 22) of the document reconstruction device 10 recognizes characters included in each divided area by optical character recognition or the like, converts them into character codes, and is included in each area. A sentence (character group) is extracted in the form of a character code (step S103). If the original document is described in character code, the character code may be extracted as it is.

  The CPU 11 (sentence extraction unit 22) of the document reconstruction device 10 analyzes the extracted sentence (character code group) and determines its language, line direction (vertical writing, horizontal writing), reading direction in line units, and the like. Specify (step S104). For example, in the case of the original document of Example 1 shown in FIG. 4, it is recognized that the sentence is Japanese, vertical writing, and a sentence that is read from right to left. Then, specify the standard reading order (from which area to read in order) when reading the text in each area of this document from the language, reading direction, etc., and set the standard reading order for the text in each area. The order according to this is given as the initial value.

  In the case of Example 1 shown in FIG. 4, it is determined that each stage proceeds from right to left and is read sequentially from the top to the bottom, and the standard reading order is region 1 → region 2 → region 3 → Region 4 is entered. Accordingly, rank 1 is assigned to the text in region 1, rank 2 is assigned to the text in region 2, rank 3 is assigned to the text in region 3, and rank 4 is assigned to the text in region 4. Note that a sentence extracted from the region N is a sentence N. Further, the order given to the sentence N is appended with () as in the document N (1).

  Next, the CPU 11 (determination unit 23) of the document reconstruction device 10 determines whether the sentence extracted from each area is a sentence connected to another sentence or an independent sentence (step S105). The sentences extracted from the areas 1 to 4 are compared with each other to determine whether or not the sentences in the two areas are continuous sentences. Here, a predetermined index value is calculated for the determination. Examples of the comparison method based on index values include text similarity and context matching calculation.

As the text similarity, for example, the similarity between sentences is calculated by TF (Term Frequency) -IDF (Inverse Document Frequency) or Cos similarity. Here, the similarity is calculated based on the Cos similarity. An example of calculating the similarity between the sentence 1 extracted from the area 1 and the sentence 2 extracted from the area 2 is shown. First, when the frequency of words in each sentence is expressed as a vector,
Sentence 1: (This year, economy, ...) = (3, 10 ...),
Document 2: (Recent technology trend, ...) = (15, 3 ...)
And Cos similarity is
Cos θ = vector of sentence 1 · vector of sentence 2 / | document 1 || document 2 |, where Cos θ = 0.2 (upper limit is 1.0).

  With regard to the degree of coincidence of context, when the end of sentence 1 is “...” and the beginning of sentence 2 is “in recent technological trends ...”, the end of sentence 1 is punctuation ( Therefore, it is determined that there is no high possibility that another sentence continues, and the degree of coincidence of the context is calculated as 0.3 (the upper limit is 1.0), for example.

  Assume that the above is comprehensively determined and the final sentence continuity of sentence 1 and sentence 2 is calculated to be 0.25.

  The same calculation is performed between adjacent sentences. That is, the sentence continuity was calculated for each of sentence 1 and sentence 2, sentence 1 and sentence 3, sentence 1 and sentence 4, sentence 2 and sentence 3, sentence 2 and sentence 4, sentence 3 and sentence 4, and found. It is determined whether the value of the sentence continuity exceeds a threshold value. For example, if the sentence continuity value between sentence 1 and sentence 2 is 0.25 and the default or user-set threshold is 0.8 (upper limit 1.0), sentence 1 and sentence 2 are not consecutive sentences. Judge. In Example 1 shown in FIG. 4, all sentences are not continuous with other sentences, and it is determined that each sentence is an independent sentence.

  Next, when there are sentences determined to be continuous sentences in step S105, the CPU 11 (sentence combining unit 24) of the document reconstruction device 10 combines these sentences into one sentence (step). S106). For example, if it is determined in step S105 that the sentence 2 and the sentence 4 are continuous documents, the sentence 2 and the sentence 4 are combined into one sentence (this is described as a document 2 + 4). The reading order of the sentence 2 + 4 after combination is the smaller of the order of the sentence 2 and the order of the document 4.

  After combining successive sentences in step S106 based on the determination result of step S105, the CPU 11 (reconstruction unit 25) of the document reconstruction device 10 is combined with each sentence (originally independent sentences and one). ) Are arranged in a line, a column-removed document (reconstructed document) is generated (step S107), and the process ends.

  FIG. 6 is a flowchart showing details of step S107 in FIG. First, the document layout is determined (step S201). Here, as the layout, it is possible to select whether to adjust the character size while maintaining the layout of the sentence or to reflow. This selection is received from the user of the portable terminal 5, for example. For the original document in Example 1 of FIG. 4, the document is reconfigured so as to adjust the character size while maintaining the text layout.

  Next, the CPU 11 (reconstruction unit 25) of the document reconstruction device 10 determines an arrangement direction for arranging sentences in a line based on the reading direction specified in step S104 (step S202). In the case of the original document of Example 1 shown in FIG. 4, since each line is vertically read and the reading direction is line by line, the text extracted from each area is arranged in a line from right to left. When browsing this text, the user adjusts the viewing position in the document by horizontal scrolling.

  Finally, the CPU 11 (reconstruction unit 25) of the document reconstruction device 10 uses the layout determined in step S201 and the layout determined in step S202 for the independent text extracted from each area and the text combined in step S106. According to the direction, the documents are reconstructed by arranging them in ascending order (step S203).

  FIG. 7 shows a reconstructed document obtained by reconstructing the original document of Example 1 shown in FIG. FIG. 5B shows a scroll state when the reconstructed document is browsed with the portable terminal 5. This reconstructed document can be browsed only by scrolling right. Thereby, compared with the original document of FIG. 4, it is possible to reduce the number of times of character enlargement and the number of scrolls of vertical movement, and it becomes a document that is easy to browse suitable for the terminal that the user is browsing.

  Next, various examples when generating a reconstructed document from an original document will be described.

<Example 3>
FIG. 8 shows the original document of Example 3. This original document is arranged in two columns, an upper level and a lower level, and is a document that is read from right to left in Japanese and vertical writing. The area is divided into four parts: one upper sentence on the right side of the upper section, one independent sentence on the lower right area, and a lower left sentence connected to the upper left area. ing.

  The CPU 11 (dividing unit 21) of the document reconstruction apparatus 10 inputs the original document (step S101), divides the document into four areas based on the boundary line, and labels each area (step S102). FIG. 9 shows a state where the regions 1 to 4 are labeled. A broken line in the figure indicates a region.

  The CPU 11 (sentence extraction unit 22) of the document reconstruction device 10 performs optical character recognition for each of the labeled areas 1 to 4, and extracts text (sentence) (step S103). In this example, “Regarding this year's economy ○ ... ○” from the region 1, “Regarding recent technological trends × ... ×” from the region 2, and “Regarding yesterday's sports Δ ...” from the region 3. The text “×... X” is extracted from the region 4.

  Then, the language, the line direction, and the reading direction in line units are specified (step S104). In the case of the original document in Example 3, the document is specified as Japanese, vertical writing, and a document that is read from right to left. Then, in accordance with the standard reading order under these conditions, a rank is assigned to the document in each area. In the case of the original document of Example 3, the standard reading order is region 1 → region 2 → region 3 → region 4, the initial order given to sentence 1 is 1, the initial order given to sentence 2 is 2, and sentence 3 The initial ranking given to the sentence 4 is 3, and the initial ranking given to the sentence 4 is 4.

  Next, the CPU 11 (determination unit 23) of the document reconstruction device 10 determines whether the sentence extracted from each area is a sentence connected to another sentence or an independent sentence (step S105). Here, it is assumed that the similarity between the sentence 2 and the sentence 4 is 0.8 (upper limit 1.0) as a result of comparison using the same method as that of the original document in Example 1. Further, the degree of coincidence of the context is assumed that the end of the sentence 2 is “..., general”, and the beginning of the sentence 4 is “highly coming”. When “general” and “coming” are combined, it becomes the character string “future”, and the character string “future” is recognized as one word by matching with the word database (dictionary). The degree of coincidence is calculated as 1.0 (upper limit 1.0). Assume that the final sentence continuity is calculated as 0.9 as a result of comprehensive judgment.

  Similarly, adjacent sentences are compared and all combinations (sentence 1 and sentence 2, sentence 1 and sentence 3, sentence 1 and sentence 4, sentence 2 and sentence 3, sentence 2 and sentence 4, sentence 3 and sentence 4) ) Is calculated for the sentence continuity value.

  Here, as a result of comparing the value of the sentence continuity with a threshold value (0.8), sentence 1 and sentence 2, sentence 1 and sentence 3, sentence 1 and sentence 4, sentence 2 and sentence 3, sentence 3 and sentence 4 The value for the interval does not exceed the user-set threshold (0.8) and is not a continuous sentence. The value of the sentence continuity calculated for sentences 2 and 4 exceeds the threshold and is determined to be a continuous sentence. Suppose. Accordingly, sentences 1 and 3 are independent sentences, and sentences 2 and 4 are combined into one sentence (step S106). Since the initial rank of document 2 is (2) and the initial rank of document 4 is (4), the rank of document 2 + 4 is (2).

  In document reconstruction, a method of adjusting the character size while maintaining the layout is determined (step S201), and the arrangement direction when arranging the sentences in a line based on the reading direction specified in step S104 is: , “Arrange sentences in a line from right to left” is determined (step S202). Then, according to the layout and arrangement direction, the reconstructed document is generated by arranging the sentences 1 (1), sentences 2 + 4 (2), and sentences 3 (3) in ascending order of ranks in () (step S203). .

  FIG. 10 shows a reconstructed document obtained by reconstructing the original document of Example 3 shown in FIG. FIG. 5B shows a scroll state when the reconstructed document is browsed with the portable terminal 5. This reconstructed document can be browsed only by scrolling right. As a result, even if the text is divided into separate areas in columns, it is possible to browse from the original document by reducing the number of character enlargements and the number of scrolls of vertical movement, which is suitable for the user's terminal It becomes easy to do.

<Example 4>
The original document of Example 4 is composed of a plurality of pages (2 pages) (see FIG. 11). The first page is composed of two sentences at the top and two sentences at the bottom, and the second page is composed of two sentences at the top and one sentence at the bottom. Both are Japanese, vertical reading, and sentences that are read from right to left. The lower left sentence on the first page continues to the upper right sentence on the second page, and the upper left sentence on the second page continues to the lower sentence. Therefore, the two-page document includes five independent sentences.

  FIG. 12 shows the result of labeling by step S101 and step S102 of FIG. Steps S103 and S104 are performed in the same manner as in the first and third examples.

  In the connection determination in step S105, the sentences 1 to 7 are compared, and the value of the sentence continuity, which is an index for determining whether or not each is a continuous sentence, is calculated. Here, for example, the sentence 5, the sentence 6, and the sentence 7 are not adjacent to the sentence 1, and the pages are different. Therefore, the calculation of the similarity between the sentences and the coincidence of the contexts is omitted.

  That is, in this example, when all combinations are calculated, the sentence continuity is calculated for 21 combinations, but the calculations for combinations that are not likely to be continuous are omitted. In this case, sentence 2 and sentence 3 for sentence 1, sentence 3 and sentence 4 and sentence 5 for sentence 2, sentence 4 for sentence 3, sentence 5 for sentence 4, sentence 6 and sentence 7 for sentence 5, For the sentence 6, a combination with the sentence 7 may be considered, and a total of 10 patterns may be calculated.

  In the example of FIG. 12, as a result of determining the connection, it is determined that sentence 1, sentence 2, and sentence 3 are independent sentences, sentence 4 and sentence 5 are continuous, sentence 6 and sentence 7 are continuous. The FIG. 13 shows the result of restructuring the document by adjusting the character size while maintaining the sentence layout and arranging the sentences in a line from right to left so that they can be read by horizontal scrolling. This makes it possible to reduce the number of character enlargements and the number of vertical scrolls compared to the original document even if the text straddles the page, and is an easy-to-view document suitable for the terminal being browsed by the user. It becomes. In addition, although the case where it straddled 2 pages was illustrated in Example 4, more than that may be used.

<Example 5>
In the original document of Example 5, objects such as figures and images exist in the area. FIG. 14 shows an original document of Example 5. The upper part includes two sentences (four sentences in total), and the sentence on the right side of the upper part includes an image object. This document is a Japanese document in which each line is read vertically, and read line by line from right to left.

  As in the other examples, the CPU 11 of the document reconstruction device 10 performs the process of FIG. 3 to divide the document into regions and label the regions. At this time, when an object such as a sentence and an image exists in one area, the object is treated as being associated (linked) with the sentence. FIG. 14B shows a state where the original document of Example 5 is divided and labeled. A broken line in the figure indicates a region.

  The CPU 11 of the document reconstruction device 10 executes each step of FIG. 3 to extract sentence 1, sentence 2, sentence 3, and sentence 4, and object 1 is linked to document 1, and each sentence is independent. It is discriminated from the sentence. Then, the text size is adjusted while maintaining the text layout, and the text is arranged in a line from the right to the left according to the order so as to be read by horizontal scrolling.

  FIG. 15 shows the generated reconstructed document. An object A is arranged in the sentence 1 in the same manner as the layout in the original document. FIG. 5B shows a scroll state when the reconstructed document is browsed with the portable terminal 5.

  In this way, even if an object such as an image is included in the text, it is possible to reduce the number of times of enlargement and scrolling of the up and down movement, and it is easy to view a document suitable for the terminal that the user is viewing. Become. In addition, although the case where one image exists is illustrated as an example of the object, a plurality of images may exist. Moreover, although the case of the image was shown as an example of an object, objects, such as a graph and a table | surface, may be sufficient.

<Example 6>
The original document of Example 6 (see FIG. 16) is a horizontally written Japanese document, and each line is read from left to right, and read line by line from top to bottom. In this example, there are two sentences at the top and two sentences at the bottom. FIG. 16B shows a state where the original document of Example 6 is divided and labeled. A broken line in the figure indicates a region.

  The CPU 11 of the document reconstruction device 10 executes each step of FIG. 3 to determine that each is an independent Japanese sentence, adjust the character size while maintaining the sentence layout, and from left to right Because it is read in the direction, the document is reconstructed so that it can be read by vertical scrolling. FIG. 17 shows a reconstructed document as a result of reconstructing. Thereby, it is possible to reduce the number of character enlargements and the number of scrolls for vertical movement from the original document, and the document is easy to browse and suitable for the terminal being browsed by the user.

<Example 7>
In Example 7, when a sentence is arranged and reconstructed, a document corresponding to reflow display is used. An example of reflow display of the document is shown in FIG.

  In some cases, reflow display can reduce the number of times the character is enlarged and scrolled up and down rather than adjusting the character size while maintaining the document layout. It is an easy-to-view document suitable for the terminal that is being used. For example, the document reconstruction apparatus 10 receives a selection for reflow from the portable terminal 5 together with a column release instruction.

<Example 8>
In Example 8, reflow display is performed with the character size and character font specified by the user. FIG. 19 shows an example when reflow display is performed with the character size and character font specified by the user. The original document is the same as in Example 7. The document reconstruction device 10 receives the designation of the character size for reflow from, for example, the portable terminal 5 together with the instruction for releasing the column. Note that the character size displayed on the portable terminal 5 may be changed by receiving an instruction on the character size in the portable terminal 5 when browsing.

  Even if reflow display is performed according to the device, the user may feel that the characters are small, so the number of times of enlargement can be reduced by setting and adjusting the character size in advance, and the user prefers it. By changing the font, the document can be easily viewed by the user.

<Example 9>
FIG. 20 shows the original document of Example 9. The original document of Example 9 is a document in which two sentences are arranged in the upper part and two sentences are arranged in the lower part, and read from the upper left to the lower right. Here, it is a Mongolian sentence. FIG. 20B shows a state in which the original document of Example 9 is divided into regions and labeled. A broken line in the figure indicates a region. The order of labeling is the same as in Japanese.

  In this example, it is assumed that the original document of Example 9 is determined to be a Mongolian sentence and each sentence is an independent document through the steps of FIG. Then, the text size is adjusted while maintaining the text layout and read from the left to the right, so that a reconstructed document in which the texts are arranged in a line from the left to the right as if read by horizontal scrolling is generated. Since it is Mongolian, the reading order of the sentences is sentence 2 (1), sentence 1 (2), sentence 4 (3), and sentence 3 (4).

  FIG. 21 shows a reconstructed document obtained by reconstructing the original document of Example 9. FIG. 5B shows a scroll state when the reconstructed document is browsed with the portable terminal 5. The top of this document is on the left side. The scroll direction during browsing is from left to right.

  In this way, depending on the language, even if it is the same vertical writing as Japanese, the reading direction in line units is different, so by arranging the sentences in the direction suitable for the language, the user can browse more. It can be reconstructed into an easy document.

<Example 10>
FIG. 22 shows the original document of Example 10. The original document of Example 10 is an English document, and sentences are arranged in a total of four areas, two at the top and two at the bottom. Each line is a document that is read from left to right and read line by line from top to bottom. FIG. 22B shows a state in which the original document of Example 10 is divided into regions and labeled. A broken line in the figure indicates a region.

  With the processing in Fig. 3, the text size is adjusted while maintaining the layout of the text, English is translated into Japanese, and read from left to right. It shows in FIG.

<Example 11>
Example 11 is a case where an XML data document having the same layout as Example 1 in FIG. 4 is an original document. Using an XML parser, processing is performed by acquiring layout tags and text tags.

  For example, tags with columned lines are extracted and divided into each area of the document, and sentences included in each area are acquired by extracting text tags.

<Others>
The threshold used when determining whether the document is an independent document or a continuous document compared to the sentence continuity may be arbitrarily set by the user, or a value determined in advance by the device It is good. In addition, although the scroll direction when reading a document has been shown as an example in which the device automatically determines the scroll direction based on the language and line-by-line reading direction, the user can specify the scroll direction and the same direction as the specified scroll direction A reconstructed document may be generated by arranging sentences.

  When restructuring a document, the user can accept the selection of whether to adjust the character size while maintaining the layout or to support reflow display, and reconfigure the document using the layout method selected by the user. May be.

  As described above, according to the present invention, even when a single sentence is originally divided and arranged in a jumped place in a stacked document, the columns of the document are appropriately released by connecting them, It is possible to generate a document that is easy to view even on a small terminal.

  The embodiment of the present invention has been described with reference to the drawings. However, the specific configuration is not limited to that shown in the embodiment, and there are changes and additions within the scope of the present invention. Are also included in the present invention.

  The area determination conditions for dividing a document into a plurality of areas are not limited to boundary lines and blanks. For example, a difference in character size (a large character is used for the headline), a difference in background color, or the like may be used. In addition, the determination method for determining whether the sentence extracted from each divided area is a sentence connected to a sentence extracted from another area or an independent sentence is not limited to the one exemplified in the embodiment. For example, semantic analysis may be used together.

  In the embodiment, a document that is vertically arranged as an original document is exemplified. However, the present invention is applied to a document that is horizontally arranged.

2 ... Document browsing system 10 ... Document reconstruction device 11 ... CPU
12 ... RAM
DESCRIPTION OF SYMBOLS 13 ... Memory | storage part 14 ... Network communication part 15 ... Input device 16 ... Output device 21 ... Dividing part 22 ... Text extraction part 23 ... Discrimination part 24 ... Text coupling | bond part 25 ... Reconstruction part

Claims (15)

A dividing unit that divides the columned original document into a plurality of areas based on a predetermined area discrimination condition;
A text extraction unit that extracts text included in each divided area;
A discriminator for discriminating whether the text extracted from each area after division is a text connected to text extracted from other areas or an independent text;
A sentence combining unit that connects sentences determined to be connected to sentences extracted from other areas to one sentence;
A reconstruction unit that arranges the independent sentences and the sentences connected to one by the sentence combination unit in a row, and reconstructs the original document into a column-unreleased document;
A document reconstruction apparatus characterized by comprising: The document reconstructing apparatus according to claim 1, wherein the discriminating unit digitizes the appropriateness of connection between sentences and compares the text with a predetermined threshold value to perform the discrimination. The document reconstruction apparatus according to claim 2, wherein the threshold value can be set by a user. The said discrimination | determination part performs the said discrimination based on the similarity of the content of a sentence, or the continuity of the end of one sentence, and the head of another one sentence. Document reconstruction device. The determination unit does not determine a connection between a sentence included in one area and a sentence included in an area at a position that is not likely to be continuous with the sentence included in the one area. 2. The document reconstruction device according to 1. The document reconstruction apparatus according to claim 1, wherein the reconstruction unit arranges the sentences in a line according to a direction in which the sentence is read. The document reconstruction device according to claim 1, wherein the reconstruction unit arranges the sentences in a line according to a direction designated by a user. The reconstructing unit can select whether the reconfiguration is performed by adjusting a character size while maintaining a layout of a sentence in each region, or the reconfiguration is performed by reflow. Document reconstruction device described in 1. The document reconstruction apparatus according to claim 8, wherein a user can specify a character size. When an object such as an image or a graphic exists in one area in addition to a sentence, the reconstruction unit treats the object and the sentence included in the one area together and arranges them. The document reconstruction apparatus according to claim 1. The document reconstruction apparatus according to claim 1, wherein, when the original document is image data, the dividing unit divides the original document by determining an area by image processing. The document reconstruction device according to claim 1, wherein when the original document is a document described in a markup language, the dividing unit divides the document based on tag information indicating a column. The document reconstruction apparatus according to claim 1, wherein when the original document is image data, the sentence extraction unit extracts a sentence by character recognition. The document reconstruction device according to claim 1, wherein when the original document is a document described in a markup language, the sentence extraction unit extracts a sentence based on tag information indicating a text area. .   A program causing an information processing apparatus to function as the document reconstruction apparatus according to any one of claims 1 to 14.