Within the process mining eld we are witnessing a tremendous growth of applications and development frameworks available to perform data analyses. Such growth, which is very positive and desirable, comes with the cost of learning each new tool and di culties in integrating di erent systems in order to complement the analyses. In addition, we are noticing the lack of tools enabling collaboration among the users involved in a project. Finally, we think it would be highly recommended to enable ubiquitous processing of data. This paper proposes a solution to all these issues by presenting a chat bot which can be included in discussions to enable the execution of process mining directly from the chat.
Process mining [
With a multitude of tools available, both as standalone, web applications and
libraries, process mining projects can now be completed taking advantage of a
broad range of possibilities. In particular, just by focusing on the open source
tools, one of the leading software is ProM [
Another important issue, that only recently is receiving attention [
To address all these challenges, the tool described in this paper has been designed with the following functional requirements in mind: { To propose a uniform interface to interact with any number of process mining applications in a seamless way and without the need of confronting every time a steep learning curve; { To enable collaboration on process mining projects by design: not just sharing intermediate or nal results, but discussing and performing each individual step of the analysis in real-time and allow all persons involved to contribute to it; { To enable ubiquitous and truly platform independent execution of process mining projects (e.g., start a mining session at the o ce PC, continue on the smartphone while commuting, and nish the analysis at home, on a tablet).
The rest of the paper is structured as follows: Sec. 2 reports some technical details about the tool, Sec. 3 describes the main implementation details, and nally Sec. 4 concludes the paper. 2
The strategy we decide to adopt in order ful ll our requirements is to develop
a chat bot [
The general idea behind a chat bot is to exploit the APIs o ered by di erent instant messaging systems and de ne a software, i.e. the bot, that consumes each message sent by the participants in a discussion. Whenever a message matches a prede ned pattern (also known as command ) which is recognized by the bot, the software performs some actions and sends the response back into the discussion so that all participants can read it. The software that runs the bot has to be executed in a machine, i.e. a server, which has access to the APIs of the instant messaging. In the context of this paper, we decided to use the instant messaging platform Telegram1 due to its explicit support of chat bots, as well as its portability, which is manifested in a number of implementations available for a large number of platforms, including open source applications2.
The chat bot we devised, which is called Process Mining Bot (or pmbot), assumes that each chat is analyzing a single log at the time. This log, called the
ProM bupaR PM4Py
pmbot pyTelegramBotAPI
Chat cj Chat ck
Chat ci current log, is provided to the bot by sharing it within the chat. When a le is shared, the pmbot automatically is noti ed and, if such le contains an event log, the server running the pmbot will download it and it will be used for all following analyses. By uploading a new le, the current log is overwritten with the new one. It is important to note that if a le containing the log is uploaded in a chat, it is not necessary for all clients to download it (in fact, only the pmbot has to download it). This is of critical importance as it enables the analysis also in devices where memory is a precious resource, such as smart-phones.
As depicted on the right-hand side of Figure 1, the Process Mining Bot can
be included in as many chats as needed: it will keep track of each of them
individually and therefore each chat will have its own current log. The left-hand
side of Figure 1 depicts the three di erent process mining frameworks, written
in three di erent programming languages that the current implementation of
the pmbot interacts with ProM [
This command uses the PM4Py library;
nhm [DEP THR] This command runs the Heuristics Miner algorithm [
These commands use the bupaR library; nresources Command to plot how often each resource is involved in an event, by using the bupaR library; nkeepactivities This command changes the current log by ltering it in order to keep only some activities. The user can enter the name of all activities to keep until a special termination command is given; nremovefilters This command can be used to reset the current log to its original form (i.e., the same status as it has been uploaded). It is useful to reset the current log whenever the users want to the applied lters.
The current set of commands already allows several interesting analyses. Actually, in all data and process mining projects it is arguable that the most important element of the analysis are the users who have to be smart in (i) understanding which algorithm is the most suitable in the given setting and (ii) interpreting the results of the mining algorithm. On top of this, the ability to recursively apply lters allows the investigation to go beyond the raw log.
Whenever treating data, one of the most delicate elements is the con dentiality of the data. Whenever implementing a bot it is important to understand that it is actually just an application running on someone's machine and the owner of such machine has access to the log. Therefore, in case the data to be mined is con dential it might be wise to have an in house and trusted machine allocated to the purpose. In addition to the machine running the pmbot, when the log is uploaded it has to transit over the Telegram's servers. To impede Telegram from being able to read the log, the current implementation supports encrypted zip les for transmitting the le. Of course, more advanced techniques (e.g., GPG) might be easily implemented as well.
A second element to be considered is the accountability of the activities performed by the bot: if the log le is very large, the machine running the pmbot might require a vast amount of resources. For this reason, in the current implementation of the bot we introduced a limit on the size of the event log (easily adjustable in the con guration le) and, in order to send any command, it is necessary to enter a \license code". This code enables the access to speci c set of functionalities and time-based coded might also be implemented to grant access only for limited periods.
The Process Mining Bot has been implemented as a standalone Python software, exploiting the Telegram API3 and, in particular, the corresponding Python
wrapping implemented in the pyTelegramBotAPI4. The current version of the
software embeds the PM4Py libraries. In order to execute bupaR and ProM,
external calls are done using the subprocess Python module. The complete source
code of the bot is available in GitHub at https://github.com/delas/pmbot.
A screencast showing all functionalities in a mock-up session is available at
https://youtu.be/eg8jJ3bB0NI. Finally, a detailed tutorial presenting a walk
through of the basic steps in order to interact with the bot is available at [
In this paper we presented an innovative technique of completing process mining projects in a collaborative, ubiquitous, platform independent and integrated manner. The technique exploits the idea of chat bots and, by including them into existing instant messaging discussions, it is possible to have direct access to process mining functionalities. The prototype implementation presented in this paper is available for Telegram and the bot is capable of executing commands on ProM, bupaR, and PM4Py.