A heterogeneous table population method

Abstract

The application provides a heterogeneous table filling method, comprising the following steps: S1, performing normalization processing on a source data table and a target table to be filled to obtain a normalized source table and a normalized target table; S2, when there is no candidate cell to be evaluated in the normalized source table, performing the operation of step S4; when there is a candidate cell to be evaluated in the normalized source table, performing the operation of step S3; S3, selecting an arbitrary candidate cell to be evaluated to perform simulation filling, and returning to the operation of step S2; S4, returning to the most suitable cell and filling the text data of the most suitable cell into the cell to be filled. The application converts table instances in different applications into a unified intermediate representation through normalization, so that the subsequent process of evaluating candidate cells does not need to pay attention to the specific implementation of the application, simulates the design and use habits of the user when filling the table, and makes the most suitable candidate cell consistent with the most suitable candidate cell selected by the user.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of data processing, and particularly relates to a heterogeneous table filling method. BACKGROUND

[0002] A table is a common data structure, which contains ordered organization of rows and columns. The intersection of rows and columns is called a cell. Tables are widely used in various fields, including education, medicine, finance, manufacturing, etc. due to their simplicity, clarity and ease of use. With the rapid development of the Internet, more and more applications supporting table form have emerged, such as spreadsheets (Excel, Google Sheet, Apple Numbers), relational databases (MySQL, Oracle), web forms, etc. These applications have different implementations of tables, but can all be regarded as instances of a table. Users complete table-related tasks through these applications. Table filling is one of the most common tasks, and users often need to manually input or copy and paste to fill in a large number of cell contents. This approach is not only prone to errors, but also takes a lot of time and effort from users. Improving the efficiency of table filling can save a lot of manpower costs.

[0003] In current table filling methods, one or more tables are used as data sources, and another table contains one or more cells that need to be filled. Users need to find the appropriate cells from the data sources and fill their contents into each cell that needs to be filled. One type of table automatic filling technology aims to help users automate this process. Current technologies mainly fill tables based on matching, such as the browser plug-in Autofill, which uses a table to record the first fill result locally, and then uses text and label matching to find the appropriate cell content for filling when filling the same table repeatedly.

[0004] However, existing technologies can handle part of the table filling task, but still have the disadvantages of inaccurate filling, poor scalability, and high development cost. The root cause of these shortcomings is the heterogeneity of tables, which manifests in two aspects: first, the heterogeneity of cell arrangement, users have complex and diverse designs and usage of tables, and different cell arrangements can express the same meaning. Second, the heterogeneity of table implementation, different applications have different implementations of tables, such as tables in Excel can have cell color, border thickness, etc. properties, while a MySQL data table does not have these properties. Therefore, there is an urgent need for a heterogeneous table filling method that is accurate, scalable, and low in development cost. SUMMARY

[0005] The present application aims at the deficiencies in the prior art, and provides a heterogeneous table filling method.

[0006] The present application provides a heterogeneous table filling method, comprising:

[0007] S1, performing normalization processing on a source data table and a target table to be filled, to obtain a normalized source table and a normalized target table respectively; wherein the source data table comprises at least one candidate cell, and the target table to be filled comprises at least one cell to be filled;

[0008] S2, judging whether there is an unevaluated candidate cell in the normalized source table; when there is no unevaluated candidate cell in the normalized source table, performing the operation of step S4; when there is an unevaluated candidate cell in the normalized source table, performing the operation of step S3;

[0009] S3, selecting an arbitrary unevaluated candidate cell for simulation filling, to evaluate the appropriateness of the selected candidate cell, and returning to perform the operation of step S2;

[0010] S4, returning the most appropriate cell and filling the text data of the most appropriate cell into the cell to be filled.

[0011] Further, the normalization processing on the source data table and the target table to be filled comprises converting the table instances of the source data table and the target table to be filled into a unified intermediate representation respectively.

[0012] Further, the selecting an arbitrary unevaluated candidate cell for simulation filling, to evaluate the appropriateness of the selected candidate cell, comprises:

[0013] S31, determining cell relationship features according to the cell features of the normalized cells in the normalized source table and the normalized target table;

[0014] S32, establishing a table filling appropriateness evidence graph G; the nodes in the graph G are pairs of normalized cells whose text data are equal or whose semantic similarity of text data exceeds a set threshold in the normalized source table and the normalized target table; the edges in the graph G are directed edges, and the graph G has multiple edges; for any two nodes A and B, there are multiple edges from A to B, and the multiple edges from A to B are spatial relationship features in the normalized source table and the normalized target table; the cell to be filled and the candidate cell of the simulation filling have the relationship of equal text data, and the composed nodes are taken as the nodes to be evaluated;

[0015] S33, enumerating all simple paths starting from an initial state in the graph G, and when there is no unevaluated simple path in the graph G, performing the operation of step S35;

[0016] S34, an arbitrary simple path is selected, the simple path is evaluated for reasonableness, and the operation of step S33 is returned to perform;

[0017] S35, the suitability of the candidate cell is obtained according to the reasonableness evaluation of the selected simple path.

[0018] Further, the cell features include spatial features and semantic features; the spatial features include the upper left corner coordinates and the lower right corner coordinates of the normalized cell; and the semantic features include the text data of the normalized cell.

[0019] Further, the cell relationship features include spatial relationship features and semantic relationship features; the spatial relationship features refer to the relationship between two normalized cells in the same normalized table; the spatial relationship features include the positional relationship, the rectangular area alignment relationship, and the cell distance relationship between the two normalized cells; the spatial relationship features are non-symmetrical, that is, the spatial relationship features from A to B are not equivalent to the spatial relationship features from B to A for two normalized cells A and B;

[0020] The semantic relationship features refer to the relationship between two normalized cells in different normalized tables; the semantic relationship features include the semantic similarity and the character matching degree of the text data of the two normalized cells;

[0021] The semantic relationship features are used to find the normalized cell pair with the semantic matching of the text data between the normalized source table and the normalized target table, so as to simulate the process of filling the table by the user; in the process of filling the table, the cell pairs with the same semantics in the source data table and the target table to be filled are matched, and the suitability of the candidate cell is evaluated according to the relationship between the different matched cell pairs.

[0022] Further, the arbitrary selection of a simple path, the evaluation of the reasonableness of the simple path, includes:

[0023] The path paradigm of the simple path is determined according to the spatial relationship features between the nodes in the simple path;

[0024] The reasonableness value of the target simple path is obtained according to the reasonableness value of the path paradigm to which the simple path belongs.

[0025] The application provides a heterogeneous table filling method, comprising the following steps: S1, performing normalization processing on a source data table and a target table to be filled to obtain a normalized source table and a normalized target table; wherein the source data table comprises at least one candidate cell, and the target table to be filled comprises at least one cell to be filled; S2, when there is no unevaluated candidate cell in the normalized source table, performing the operation of step S4; when there is an unevaluated candidate cell in the normalized source table, performing the operation of step S3; S3, selecting an arbitrary unevaluated candidate cell for simulation filling to evaluate the suitability of the selected candidate cell, and returning to perform the operation of step S2; and S4, returning the most suitable cell and filling the text data of the most suitable cell into the cell to be filled.

[0026] The application converts table instances in different applications into a unified intermediate representation through normalization, so that the subsequent process of evaluating candidate cells does not need to focus on the specific implementation of the application; based on the unified data structure, the cell features of the normalized cells are obtained, and then the cell relationship features are derived; the cell relationship features include spatial relationship features and semantic relationship features, which are used to simulate the design and use habits of the user when filling the table, so that the most suitable candidate cell is consistent with the most suitable candidate cell selected by the user. BRIEF DESCRIPTION OF DRAWINGS

[0027] In order to more clearly illustrate the technical solutions of the present application, the drawings needed in the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative labor.

[0028] Figure 1 A flowchart of a heterogeneous table filling method provided for the embodiments of the present application is shown in the figure.

[0029] Figure 2 A table filling task and a process of normalizing the table filling task provided for the embodiments of the present application are shown in the figure.

[0030] Figure 3 Cell features and cell relationship features provided for the embodiments of the present application are shown in the figure.

[0031] Figure 4 An effect diagram after simulation filling provided for the embodiments of the present application is shown in the figure.

[0032] Figure 5 A simple path diagram provided for the embodiments of the present application is shown in the figure. DETAILED DESCRIPTION

[0033] With reference to the accompanying drawings, the technical solutions in the embodiments of the present application will be clearly and completely described in the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative efforts belong to the scope of the present application.

[0034] The heterogeneous table provided by the embodiments of the present application can be a table instance of different applications, or a table instance of different cell arrangement modes in the same application. The table instance is a data structure composed of a cell set divided by ordered rows and columns, such as a Sheet table in Excel, a data table in a MySQL database, or a form in a webpage.

[0035] As shown in the figure, Figure 1 The heterogeneous table filling method provided by the embodiments of the present application comprises the following steps: S1, performing normalization processing on a source data table and a target table to be filled, to obtain a normalized source table and a normalized target table respectively; wherein the source data table comprises at least one candidate cell, and the target table to be filled comprises at least one cell to be filled.

[0036] As shown in the figure, Figure 2 The table filling task is to fill data from an Excel table into a webpage form. The source data table is 1011, and the target table to be filled is 1012. After normalization processing, the normalized source table 1013 and the normalized target table 1014 are obtained. The normalization processing of the Excel table mainly includes reading a file, parsing an xml data format, and processing a merged cell. The normalization processing of the webpage form mainly includes region segmentation and parsing an html file. The embodiments of the present application take the first candidate cell 1015, the second candidate cell 1016, and the cell to be filled 1017 as examples to illustrate the subsequent filling process, and the filling logic of other cells to be filled is consistent with that of the cell to be filled 1017. In the embodiments of the present application, for the cell to be filled 1017, the first candidate cell 1015 is the most suitable candidate cell selected by the user.

[0037] S2, judging whether there is an unevaluated candidate cell in the normalized source table; when there is no unevaluated candidate cell in the normalized source table, performing the operation of step S4; when there is an unevaluated candidate cell in the normalized source table, performing the operation of step S3.

[0038] S3, selecting an arbitrary unevaluated candidate cell for simulation filling to evaluate the suitability of the selected candidate cell, and returning to perform the operation of step S2.

[0039] The first time, candidate cell 1015 was selected for simulated filling. The second time, candidate cell 1016 was selected for simulated filling.

[0040] S31. Determine the cell relationship features based on the cell features of the normalized cells in the normalized source table and the normalized target table.

[0041] like Figure 3 As shown, cell feature 2011 includes spatial features and semantic features. The spatial features include the coordinates of the top-left corner (1,0) and the bottom-right corner (2,1). The semantic features include the text data "Name". Spatial relationship feature 2012, where the arrow direction indicates the direction of the spatial feature, includes, but is not limited to, the positional relationship between the two normalized cells, the alignment relationship of the rectangular area, and the cell distance relationship. Semantic relationship feature 2013 includes, but is not limited to, the semantic similarity and character matching degree of the text data in the two normalized cells. Each edge in the spatial and semantic relationship features represents a feature. The spatial relationship features are asymmetric; that is, for two normalized cells A and B, the spatial relationship feature from A to B is not equivalent to the spatial relationship feature from A to B. In this invention, the spatial relationship features are relative positional direction, whether they are fully aligned, and whether they are adjacent. Semantic relation features determine whether text data are equal. These features are used to find normalized cell pairs that semantically match the text data in the normalized source table and the normalized target table, simulating the user's table filling process. During the table filling process, cell pairs with the same semantics in the source data table and the target table to be filled are matched, and the suitability of candidate cells is evaluated based on the relationship between different matching cell pairs.

[0042] S32, establish a form-filling suitability evidence graph G; the nodes in graph G are normalized cell pairs where the text data in the normalized source table and the normalized target table are equal or the semantic similarity of the text data exceeds a set threshold; the edges in graph G are directed edges, and there are multiple edges in graph G; for any two nodes A and B, there are multiple edges from A to B, and the multiple edges from A to B are the spatial relationship features between the normalized source table and the normalized target table; the nodes formed by the simulated fill-in cells and candidate cells with equal text data are used as nodes to be evaluated.

[0043] After simulating filling and deriving cell relationship features, a form-filling suitability evidence graph is constructed based on the modified intermediate representation and features. The form-filling suitability evidence graph G is displayed using the simulated filling result of the first candidate cell 1015. The graph construction process for other candidate cells is the same and will not be repeated. Based on the semantic relationship features of whether the text data is equal, all matching normalized cell pairs in the normalized source table 1013 and the normalized target table 1014 can be found. For example... Figure 4As shown, when plotting, to ensure understandability, a plurality of spatial relationship features between nodes are represented by an edge.

[0044] S33, enumerate all simple paths from the initial state in the graph G, when there is no unassessed simple path in the graph G, perform the operation of step S35.

[0045] Enumerate the possible two simple paths and compare their path rationality, the rest of the path rationality evaluation and comparison are basically the same as the process, and will not be repeated. For example Figure 5 As shown, the solid line and the dashed line represent two different paths, Figure 5 The text in the edge represents the spatial relationship feature of each edge in the path.

[0046] S34, randomly select an unassessed simple path, evaluate the simple path rationality, and return to perform the operation of step S33; wherein, the simple path is a path without repeated nodes.

[0047] According to the user's design and use of table habits, design the path paradigm, each path paradigm has a specific rationality value calculation method.

[0048] The path paradigm can be an induction of any user's design and use of table habits, in the implementation of the present application, for the convenience of explanation, three possible path paradigms are listed.

[0049] Path paradigm 1: when the spatial feature relationship of all edges in the simple path is alignment and adjacency, the path score is 10.

[0050] Path paradigm 2: when the spatial feature relationship of all edges in the simple path is alignment, there is one or more non-adjacency, the path score is 5.

[0051] Path paradigm 3: when there is a spatial relationship feature of non-alignment in the simple path, the score is 0.

[0052] According to the spatial relationship feature between the nodes in the simple path, determine the path paradigm of the path.

[0053] For Figure 5 The two paths in the figure are judged, it is found that the simple path represented by the solid line conforms to the path paradigm 1, and the simple path represented by the dashed line conforms to the path paradigm 2.

[0054] According to the rationality value calculation method of the path paradigm to which the simple path belongs, the rationality value of the path is obtained.

[0055] According to the set path paradigm rationality calculation rule, the rationality of the simple path represented by the solid line is 10, and the rationality of the simple path represented by the dashed line is 0.

[0056] S35. Based on the rationality evaluation of the selected simple path, the suitability of the candidate cell is obtained.

[0057] S4 returns the optimal cell and fills the unfilled cell with the optimal cell's text data.

[0058] by Figure 5 Taking two paths as an example, the simple path represented by the solid line has a rationality score of 10, while the simple path represented by the dashed line has a rationality score of 0. In this invention, the candidate cell with the highest rationality score among all simple paths is selected as its suitability score. The higher the path rationality score, the more likely the simple path is to conform to a certain user's design and usage habits of the table. Therefore, the first candidate cell 1015 has the same rationality score as the simple path represented by the solid line, which is 10. In this embodiment of the invention, all candidate cells are sorted in descending order according to their corresponding suitability scores. The higher the score, the more likely it is to be selected by the user as the most suitable cell to fill in the unfilled cell; the lower the score, the less likely it is. The cell with the highest score is the most suitable candidate cell. The first candidate cell 1015 is selected as the most suitable candidate cell, a result consistent with the user's filling process.

[0059] The heterogeneous table filling method provided by this invention converts table instances from different applications into a unified intermediate representation through normalization, so that the subsequent evaluation of candidate cells does not need to concern itself with the specific implementation of the application; based on a unified data structure, the cell features of the normalized cells are obtained, and then the cell relationship features are derived; the cell relationship features include spatial relationship features and semantic relationship features, which can be used to simulate the user's design and usage habits when filling the table, so that the optimal candidate cell can be consistent with the optimal candidate cell selected by the user.

[0060] The present invention has been described in detail above with reference to specific embodiments and exemplary examples; however, these descriptions should not be construed as limiting the present invention. Those skilled in the art will understand that various equivalent substitutions, modifications, or improvements can be made to the technical solutions and embodiments of the present invention without departing from the spirit and scope of the invention, and all such modifications and improvements fall within the scope of the present invention. The scope of protection of the present invention is defined by the appended claims.

Claims

1. A method for filling heterogeneous tables, characterized in that, include: S1, normalize the source data table and the target table to be filled to obtain a normalized source table and a normalized target table respectively; wherein, the source data table includes at least one candidate cell and the target table to be filled includes at least one cell to be filled; S2, determine whether there are unevaluated candidate cells in the normalized source table; if there are no unevaluated candidate cells in the normalized source table, execute step S4; if there are unevaluated candidate cells in the normalized source table, execute step S3. S3, select any unevaluated candidate cells for simulated filling to evaluate the suitability of the selected candidate cells, and then return to execute step S2, specifically including: S31, determine the cell relationship characteristics based on the cell characteristics of the normalized cells in the normalized source table and the normalized target table; S32, establish a form-filling suitability evidence graph G; the nodes in graph G are normalized cell pairs where the text data in the normalized source table and the normalized target table are equal or the semantic similarity of the text data exceeds a set threshold; the edges in graph G are directed edges, and there are multiple edges in graph G; for any two nodes A and B, there are multiple edges from A to B, and the multiple edges from A to B are the spatial relationship features between the normalized source table and the normalized target table; the nodes formed by the simulated fill-in cells and candidate cells with equal text data are used as nodes to be evaluated; S33, enumerate all simple paths in graph G that start from the initial state. If there are no unevaluated simple paths in graph G, execute the operation of step S35. S34, arbitrarily select an unevaluated simple path, evaluate the rationality of the simple path, and return to execute the operation of step S33; where a simple path is a path without duplicate nodes; S35. Based on the rationality evaluation of the selected simple path, the suitability of the candidate cell is obtained. S4 returns the optimal cell and fills the unfilled cell with the optimal cell's text data.

2. The heterogeneous table filling method according to claim 1, characterized in that, The normalization process for the source data table and the target table to be filled includes converting the table instances of the source data table and the target table to be filled into a unified intermediate representation.

3. The heterogeneous table filling method according to claim 1, characterized in that, The cell features include spatial features and semantic features; the spatial features include the coordinates of the top left corner and the bottom right corner of the normalized cell; the semantic features include the text data of the normalized cell.

4. The heterogeneous table filling method according to claim 1, characterized in that, The cell relationship features include spatial relationship features and semantic relationship features; the spatial relationship features refer to the relationship between two normalized cells within the same normalized table; Spatial relationship features include the positional relationship between two normalized cells, the alignment relationship of rectangular regions, and the distance relationship between cells; Spatial relationship features are asymmetric, meaning that for two normalized cells A and B, the spatial relationship feature from A to B is not equivalent to the spatial relationship feature from B to A. The semantic relationship feature refers to the relationship between two normalized cells within different normalized tables; the semantic relationship feature includes the semantic similarity and character matching degree of the text data in the two normalized cells; The semantic relationship features are used to find normalized cell pairs that semantically match the text data in the normalized source table and the normalized target table, in order to simulate the process of a user filling in a table. During the table filling process, cell pairs with the same semantics in the source data table and the target table to be filled are matched, and the suitability of candidate cells is evaluated based on the relationship between different matching cell pairs.

5. The heterogeneous table filling method according to claim 1, characterized in that, The step of arbitrarily selecting an unevaluated simple path and evaluating its rationality includes: The path paradigm of a simple path is determined based on the spatial relationship characteristics between nodes in the simple path. Obtain the rationality value of the target simple path based on the rationality value of the path paradigm to which the simple path belongs.

Images