A method, device and storage medium for discovering a task sequence

Abstract

The present invention discloses a method, device and storage medium for discovering a task sequence. The method comprises obtaining an original operation log data set for characterizing computer desktop operations; determining one or more target operation events from the original operation log data set according to a user's instructions; and searching for corresponding similar operation events within a certain time window according to the determined target operation events to form a task sequence. The present invention can effectively identify task sequences from a large number of operation logs based on the operation events selected by the user, thereby improving the efficiency of task mining.

Description

Technical Field

[0001] The present invention relates to a method, device and storage medium for discovering a task sequence, and belongs to the technical field of task mining. Background Art

[0002] Task mining often involves collecting a large number of operation logs related to different tasks, such as chatting and emailing. These logs not only increase the number of nodes in the flowchart, interfering with business analysis, but also implicate user privacy and increase the difficulty of analysis. Therefore, an effective method for determining task relevance and automatically identifying different task sequences is needed. Summary of the Invention

[0003] The purpose of the present invention is to overcome the deficiencies in the prior art and provide a method, device and storage medium for discovering task sequences, thereby improving the effectiveness of task mining raw data.

[0004] To achieve the above object, the present invention is implemented by adopting the following technical solutions:

[0005] In a first aspect, the present invention provides a method for discovering a task sequence, comprising:

[0006] Obtaining a raw operation log dataset for characterizing computer desktop operations;

[0007] According to the user's instructions, one or more target operation events are determined from the original operation log data set;

[0008] According to the determined target operation event, corresponding similar operation events are found within a certain time window to form a task sequence.

[0009] Furthermore, the original operation log data set stores multiple log records, each log record is used to represent an operation event in the computing desktop operation, and each log record or each operation event is configured with the following fields: location identifier, time identifier, event type, operation content, element label and element content.

[0010] Furthermore, the location identifier includes a window title, a page title, and an application name.

[0011] Furthermore, the step of searching for corresponding similar operation events within a certain time window based on the determined target operation event to form a task sequence includes:

[0012] In the time series represented by the original operation log dataset, starting from the target operation event, searching for similar operation events after or after the target operation event and within a preset time window, where the similar operation events are operation events similar to the target operation event;

[0013] If the stop condition of the search action is not met, searching, based on the target operation event or the similar operation event, for an operation event similar to the target operation event, or a similar operation event similar to the similar operation event, within a next time window after the time window; and repeating this step until the stop condition of the search action is met;

[0014] Based on the time tags of the target operation events and similar operation events, a task sequence is formed.

[0015] Furthermore, similar operation events are searched through the following methods, including:

[0016] Taking each field in the operation event as a feature, a feature vector of each field is generated;

[0017] Based on the feature vectors, comparing the distances between corresponding feature vectors of two operation events;

[0018] Based on the distance between the feature vectors, the similarity score between each feature is calculated, and the scores of all features are combined to obtain a similarity score to determine whether two operation events are similar;

[0019] Operation events that meet the similarity score threshold are determined as similar operation events.

[0020] Furthermore, the step of calculating the similarity scores between the features based on the distance between the feature vectors includes:

[0021] Based on the distance between feature vectors, calculate the similarity score between each feature;

[0022] For the features, distinguish between key features and non-key features; wherein key features are given high weights and non-key features are given low weights;

[0023] The corresponding similarity scores are weighted using the weights, and the weighted similarity scores are then used to calculate the similarity scores of the operation events.

[0024] Furthermore, the stopping condition of the search action includes at least one of the following conditions:

[0025] There are no operation events within one or several consecutive time windows;

[0026] There is no operation event similar to the target operation event in one or several consecutive time windows, or there is no operation event similar to a similar operation event in a time window before the one or several consecutive time windows.

[0027] In a second aspect, the present invention provides a task sequence discovery device, comprising:

[0028] An acquisition module, used to acquire an original operation log dataset for characterizing computer desktop operations;

[0029] A filtering module is used to determine one or more target operation events from the original operation log data set according to the user's instructions;

[0030] The task sequence forming module is used to search for corresponding similar operation events within a certain time window according to the determined target operation event to form a task sequence.

[0031] In a third aspect, the present invention provides an electronic device including a processor and a storage medium;

[0032] The storage medium is used to store instructions;

[0033] The processor is configured to operate according to the instructions to execute the steps of any of the aforementioned methods.

[0034] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, which implements the steps of any of the aforementioned methods when executed by a processor.

[0035] Compared with the prior art, the present invention has the following beneficial effects:

[0036] The present invention provides a method, device and storage medium for discovering a task sequence, which can effectively identify a task sequence from a large number of operation logs based on an operation event selected by a user, thereby improving the efficiency of task mining. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 This is a flowchart of a method for discovering a task sequence provided by an embodiment of the present invention.

[0038] Figure 2 This is a block diagram of a task sequence discovery device provided by an embodiment of the present invention.

[0039] Figure 3 A block diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0040] The present invention will be further described below in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solutions of the present invention and are not intended to limit the scope of protection of the present invention.

[0041] Example 1

[0042] like Figure 1 As shown, this embodiment introduces a method for discovering a task sequence, including:

[0043] Obtaining a raw operation log dataset for characterizing computer desktop operations;

[0044] According to the user's instructions, one or more target operation events are determined from the original operation log data set;

[0045] According to the determined target operation event, corresponding similar operation events are found within a certain time window to form a task sequence.

[0046] The application process of the task sequence discovery method provided in this embodiment specifically involves the following steps:

[0047] Step 100: obtaining an original operation log data set for characterizing computer desktop operations;

[0048] In this embodiment, the original operation log data set can be obtained by collecting the user's desktop operations through a data set collector deployed on the user terminal.

[0049] The original operation log dataset stores multiple log records, each of which is used to represent an operation event in the computing desktop operation;

[0050] Each log record or each operation event can be configured with the following fields:

[0051] The location identifier can include: window title, page title, and in some cases, the location identifier can also be the application name, but the granularity of the application name is too large to be convenient for task mining;

[0052] Time identifier, which can be a timestamp;

[0053] Event type, which can be mouse click, keyboard input, etc.

[0054] The element label can be the element label of the UI element clicked by the user when the event type is mouse click;

[0055] The element content can be the text content of the UI element when the event type is a mouse click. For example, if a user clicks a button displayed as "Submit", the element content in this operation event is "Submit".

[0056] The operation content can be represented as the field specifically input by the user when the event type is keyboard input.

[0057] In some implementations, each of the above-mentioned log records or each operation event may also be configured with fields such as UIL.

[0058] Step 200: According to the user's instruction, one or more target operation events are determined from the original operation log data set;

[0059] It can be understood as the user selecting one or more log records (operation events) from the original operation log dataset;

[0060] The user's choice of log recording is the prior information that the user gives to the method; the user can select one or more target operation events based on his or her understanding of the task. For example, the user directly determines the target operation event based on the location identifier, that is, the window title or page title. For example, it is necessary to automatically discover the task sequence about expense reimbursement, so the user can directly select the operation event with key fields such as expense reimbursement in the window title as the target operation event. Therefore, the above-mentioned user instruction is also the user's instruction to select the target operation event.

[0061] In this embodiment, the one or more target operation events may be one or more target operation events related to one task. When the user selects multiple target operation events, the multiple target operation events may be target operation events related to multiple tasks.

[0062] In the technical field of process mining, in a specific scenario, the concept of process is often clear. For example, in a reimbursement process, the nodes of each process are often set in advance. However, when the process is actually executed, one or more nodes may deviate from the process, but the ideal form of the process is clear.

[0063] However, in the technical field of task mining, there is no ideal task form. Users cannot predefine how a task should be executed. For example, for the process node "Initiate Reimbursement" in the reimbursement process, the specific execution process of this process node is also a task sequence. In reality, "Initiate Reimbursement" often has multiple task sequences, because users can first fill in their name in the CRM, then fill in the reimbursement amount and items, or fill in the items first, then fill in the name and amount, use only keyboard input, or use only mouse copy and paste, etc.

[0064] Therefore, when reflected in the original operation log dataset, two very different task sequences may very well represent the same task. Ultimately, for a specific task, different people often perform different operations, and these operations often involve operations on other tasks or operations completely unrelated to the task. Therefore, in this implementation, user instructions are introduced, allowing the user to select target operation events related to the task, and the corresponding task sequence is expanded based on the user-selected target operation events.

[0065] Step 300: Based on the determined target operation event, search for corresponding similar operation events within a certain time window to form a task sequence;

[0066] In this embodiment, step 300 may include:

[0067] Starting from the target operation event, searching for similar operation events after or after the target operation event and within a preset time window, where the similar operation events are operation events similar to the target operation event;

[0068] If the stop condition of the search action is not met, searching, based on the target operation event or the similar operation event, for an operation similar to the target operation event or a similar operation event similar to the similar operation event within a next time window after the time window; and repeating this step until the stop condition of the search action is met;

[0069] Based on the time stamps of the above target operation events and similar operation events, a task sequence is formed.

[0070] In this embodiment, the time window can be one minute, half a minute, or two or three minutes. In short, it is a small-scale time window. The reason for introducing this small-scale time window is because we believe that, or for most users, in a small-scale time window, operation events are often correlated.

[0071] Therefore, starting from the target operation event, in the time sequence (each operation event has a timestamp, and the whole has a time sequence), search for operation events related to the target operation event; the search can be forward or backward.

[0072] In this embodiment, a similarity algorithm can be used to search. During the search process, when the similarity between an operation event and a target operation event exceeds a preset threshold, the operation event can be added to the sequence. Therefore, whenever a similar operation event is found, the operation event can be added based on the time stamp of the operation event to form a task sequence.

[0073] In this embodiment, similar operation events may be searched by the following methods, including:

[0074] Extract the corresponding field content from each operation event (log record) and generate a corresponding text representation for each operation event, that is, describe an operation event through text;

[0075] Generate a vector representing the text corresponding to each operation event. You can use methods such as Bag-of-Words or Word Embedding to generate the corresponding text vector.

[0076] Use similarity measurement methods to calculate the similarity between operation events. For example, cosine similarity, Jaccard similarity, etc. can be used.

[0077] Add the action events that meet the similarity threshold to the task sequence.

[0078] In some implementations, similar operation events may also be searched for using the following methods, including:

[0079] Each field in the operation event is used as a feature to generate a feature vector for each field. For example, the location identifier, event type, element label, element content, and operation content can be used as features, and then the features are encoded to obtain the feature vector of each operation event. The above features can be encoded in a variety of ways to generate feature vectors. For example, the embedding method can be used. Also through the embedding encoding method, the features of each operation event are mapped into a low-dimensional dense vector through the word embedding model. The above features can also be encoded using the one-hot encoding method.

[0080] After obtaining the feature vectors for each feature of an operation event, we can compare the distances between the corresponding feature vectors of two operation events. It's easy to understand that the distance should be calculated between the feature vectors of the location identifier and the feature vector of the location identifier, rather than between the feature vector of the location identifier and the feature vector of the event type. Cosine distance can also be used for comparison.

[0081] Based on the distance between the feature vectors, the similarity scores between the features are calculated, and finally the scores of all the features are combined to obtain the similarity scores of whether the two operation events are similar.

[0082] It is worth noting that in the above-mentioned implementation, each field in the operation event can be used as a feature to generate a feature vector for each field. However, different features are actually of different importance for evaluating whether two operation events are similar. For example, the user performs multiple operations related to expense reimbursement (possibly interspersed with some operations unrelated to expense reimbursement). The location identifiers of the operations related to expense reimbursement, such as the window title, should be the same. For example, in a certain window about expense reimbursement in CRM, the user performs the expense reimbursement operation. In other words, the location identifier is an important feature for evaluating whether two operation events are similar. Correspondingly, some features are less important, such as the operation type. Using the above example, the user can perform mouse click operations in a certain window about expense reimbursement, such as selecting a specific department for reimbursement, and can also perform keyboard input operations, such as entering an amount.

[0083] In view of the above situation, the similarity comparison between operation events can also be performed by the following methods, including:

[0084] Select several fields from the operation event as features and generate corresponding feature vectors. For example, you can select location identifier, event type, element label, element content, etc. You do not need to select a time identifier because the target operation event is used to compare similarity with the operation events in a time window.

[0085] After obtaining the feature vectors of each feature of the operation event, the distances between the corresponding feature vectors of two operation events can be compared.

[0086] Based on the distance between feature vectors, the similarity score between each feature is calculated.

[0087] For the selected features, distinguish between key features and non-key features. For example, the location identifier can be used as the key feature and the rest of the features as non-key features. Then, high weights are assigned to key features, and correspondingly, low weights are assigned to non-key features. Then, the weights are used to weight the similarity scores between the features.

[0088] Finally, the weighted similarity scores are used to calculate the similarity scores of the operation events.

[0089] As mentioned above, in addition to using the target operation event as the comparison object, similar operation events can also be used as comparison objects. Specifically, for example, the user selects A as the target operation event, and the corresponding time identifier is a. Within a time window after (or before) the time identifier a, the time window can be defined as the first time window (the scale of the time window can be b). Several operation events are determined from the original operation log data set. These several operation events constitute a set, which can be the first operation event set.

[0090] Use the target operation event A to perform a similarity comparison with several operation events in the first operation event set. If there is an operation event that meets the similarity threshold, it is defined as a similar operation event B. According to the time mark of the operation event, the similar operation event B is connected after the target operation event A, and a task sequence is obtained, that is, a task sequence with two operation events. The task sequence can be extended to have more operation events. The extension method can be to extend a time window after the above-mentioned first time window, which can be defined as a second time window. Similarly, based on the second event window, several operation events are determined from the original operation log data set. These several operation events constitute the second operation event set. The target operation event A can be used to perform a similarity comparison with several operation events in the second operation event set to obtain a similar operation event C1. The similar operation event B can also be used to perform a similarity comparison with several operation events in the second operation event set to obtain a similar operation event C2. The advantage of this method is that, as mentioned above, in the technical field of task mining, there is no such a method. In an ideal task form, users will not predefine how to perform a task. In other words, the concept of a task is often not so clear. For example, in a certain task, the user may perform multiple operation events under a certain location identifier, and then switch to another or multiple location identifiers to perform related operation events. These operation events constitute a task sequence. The above-mentioned expense reimbursement scenario is also used to illustrate. For example, when reimbursing expenses, the user can first perform several operations under the location identifier of the reimbursement page in CRM, and then jump to the location identifier of the business trip page to check matters related to reimbursement (more than one operation may be performed). Therefore, for the task sequence of expense reimbursement, at least two or more location identifiers are involved. If the task sequence is expanded based solely on the user-selected target operation event A, some operation events may be missed. Therefore, similar operation events in the first time window can be expanded based on the target operation event A, and similar operation events in subsequent time windows can be determined based on similar operation events in the previous time window, that is, based on the target operation event or based on the similar operation event, the next time window after the time window is searched for operations similar to the target operation event, or similar operation events similar to the similar operation event. In the process of searching, similar operation events in a certain time window can be determined based on similar operation events in the time window before the time window. This method takes into account the continuity between operation events in a task sequence, and can better search for operation events related to the task.

[0091] In this embodiment, the stopping condition of the above-mentioned search action may be:

[0092] 1. There is no operation event (that is, no record) within one or several consecutive time windows. In this case, it can be considered that the user has completed the task and stopped the operation. Therefore, the search can be stopped and one or more task sequences can be obtained.

[0093] 2. If there is no operation event similar to the target operation event in one or several consecutive time windows, or no operation event similar to a similar operation event in a time window before the one or several consecutive time windows, in this case, it can be considered that the user is performing other tasks or performing operations unrelated to the task, so the search can be stopped to obtain one or more task sequences.

[0094] This method allows users to define business-related operational events, such as opening a specific CRM software or webpage. These business events are first matched against the original data set, which contains a large number of unrelated events, and marked as relevant events. A similarity algorithm is then used to search for strongly related events within a specified event window (e.g., one minute before and after), and these are merged into the task sequence until it cannot be expanded. This automatically generates multiple task sequences, each of which contains the user's desired business operations, improving the effectiveness of the original task mining data.

[0095] Example 2

[0096] like Figure 2 As shown, this embodiment provides a task sequence discovery device, including:

[0097] An acquisition module, used to acquire an original operation log dataset for characterizing computer desktop operations;

[0098] A filtering module is used to determine one or more target operation events from the original operation log data set according to the user's instructions;

[0099] The task sequence forming module is used to search for corresponding similar operation events within a certain time window according to the determined target operation event to form a task sequence.

[0100] Example 3

[0101] like Figure 3 As shown, this embodiment provides an electronic device, including a processor and a storage medium;

[0102] The storage medium is used to store instructions;

[0103] The processor is configured to operate according to the instructions to execute the steps of the method according to any one of the first embodiments.

[0104] Example 4

[0105] This embodiment provides a computer-readable storage medium having a computer program stored thereon. When the program is executed by a processor, the steps of any one of the methods described in Embodiment 1 are implemented.

[0106] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the technical principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for discovering a task sequence, characterized in that: include: Obtaining a raw operation log dataset for characterizing computer desktop operations; According to the user's instructions, one or more target operation events are determined from the original operation log data set; Based on the determined target operation event, corresponding similar operation events are searched within a certain time window to form a task sequence, including: In the time series represented by the original operation log dataset, starting from the target operation event, searching for similar operation events after or after the target operation event and within a preset time window, where the similar operation events are operation events similar to the target operation event; If the stop condition of the search action is not met, searching, based on the target operation event or the similar operation event, for an operation event similar to the target operation event, or a similar operation event similar to the similar operation event, within a next time window after the time window; and repeating this step until the stop condition of the search action is met; Based on the time tags of the target operation events and similar operation events, a task sequence is formed.

2. The method for discovering a task sequence according to claim 1, wherein: The original operation log data set stores multiple log records, each log record is used to represent an operation event in the computing desktop operation, and each log record or each operation event is configured with the following fields: location identifier, time identifier, event type, operation content, element tag and element content.

3. The method for discovering a task sequence according to claim 2, wherein: The location identifier includes a window title, a page title, and an application name.

4. The method for discovering a task sequence according to claim 1, wherein: Search for similar operation events using the following methods, including: Taking each field in the operation event as a feature, a feature vector of each field is generated; Based on the feature vectors, comparing the distances between corresponding feature vectors of two operation events; Based on the distance between the feature vectors, the similarity score between each feature is calculated, and the scores of all features are combined to obtain a similarity score to determine whether two operation events are similar; Operation events that meet the similarity score threshold are determined as similar operation events.

5. The method for discovering a task sequence according to claim 1, wherein: The steps of calculating the similarity scores between the features based on the distance between the feature vectors include: Based on the distance between feature vectors, calculate the similarity score between each feature; For the features, distinguish between key features and non-key features; wherein key features are given high weights and non-key features are given low weights; The corresponding similarity scores are weighted using the weights, and the weighted similarity scores are then used to calculate the similarity scores of the operation events.

6. The method for discovering a task sequence according to claim 1, wherein: The stopping condition of the search action includes at least one of the following conditions: There are no operation events within one or several consecutive time windows; There is no operation event similar to the target operation event in one or several consecutive time windows, or there is no operation event similar to a similar operation event in a time window before the one or several consecutive time windows.

7. A task sequence discovery device, using the task sequence discovery method according to claim 1, characterized in that: include: An acquisition module, used to acquire an original operation log dataset for characterizing computer desktop operations; A filtering module is used to determine one or more target operation events from the original operation log data set according to the user's instructions; The task sequence forming module is used to search for corresponding similar operation events within a certain time window according to the determined target operation event to form a task sequence.

8. An electronic device, characterized in that: including processor and storage medium; The storage medium is used to store instructions; The processor is configured to operate according to the instructions to execute the steps of the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images