Article
To read the full-text of this research, you can request a copy directly from the authors.

Abstract

Chatbots enable organizations in the business-to-customer domain to respond to repetitive requests efficiently. Extant approaches in Natural Language Processing (NLP) already address the essential requirement of understanding user input and synthesizing a response as close as possible to a response a human interlocutor would give. However, we argue that the organizational adoption of chatbots further depends on the underlying model’s capability to learn and comply with organizations’ business processes, for example, authenticating a customer before providing sensitive details. To address this issue, we develop an approach that quantifies chatbots’ ability to learn business processes using standardized process mining metrics. We demonstrate our approach by training chatbots on a dataset of more than 500,000 customer service conversations from three companies on Twitter and show how our approach supports the quantification of a chatbot’s overall ability to learn business processes from the training data. Furthermore, we quantify a chatbot’s ability to learn a particular variant of the underlying process and we show how to compare the chatbot’s executed steps against a given normative process model. Our approach that seamlessly integrates with existing approaches to evaluate NLP-based chatbots mitigates the current hurdles that practitioners face and, therefore, strives to foster the adoption of chatbots in practice.

No full-text available

Request Full-text Paper PDF

To read the full-text of this research,
you can request a copy directly from the authors.

... Surprisingly, none of these studies effectively harnessed the potential of chatbots for generating process models from natural language text, nor for the conversion of process models into natural language text. Studies S4 [15] and S7 [18] were the only two studies we found with our search string including a chatbot. Study S4, 'Comparing Generative Chatbots Based on Process Requirements: A Case Study' [15], evaluates chatbots to determine their effectiveness in executing and managing predefined business processes. ...
... Study S4 assesses how well these chatbots can follow BPMN standards, manage task sequences, and handle decision gateways, providing insights into their practical applicability in supporting business process execution and management. Study S7, 'Quantifying Chatbots' Ability to Learn Business Processes' [18], focuses on evaluating the capability of chatbots to learn and execute business processes. Study S7 develops a method to measure this proficiency using process mining metrics. ...
Article
Full-text available
This paper presents a comprehensive literature review, focusing on the emerging intersection of chatbot technology and the generation of process models. As an evolving field of study, the integration of interactive chatbots into process model generation represents a promising approach, blending advancements in artificial intelligence in general, and natural language processing in particular, with process management methods. This review systematically examines the existing literature across multiple disciplines, identifying and analyzing studies that touch upon the individual components of this nascent topic: chatbot technology, process model generation, and their synergistic potential. Despite the scarcity of direct research aimed at using chatbots for process model generation, this review synthesizes relevant findings from related domains, such as natural language processing applications in process modeling, and the broader impact of chatbot interfaces in various domains. Through this analysis, we aim to map the current landscape of research, highlight significant gaps, and suggest potential pathways for future investigations. This paper not only aggregates existing knowledge, but also assesses the applicability and implications of current technologies and theories when generating process models with the assistance of interactive chatbots. The outcome is a foundational compendium for researchers and practitioners interested in exploring this innovative intersection, providing a springboard for future research and development in this promising area.
... Additionally, they may not be able to handle sensitive or confidential information securely, which can pose a risk to customers' privacy and security. Finally, they may lack the ability to learn and improve over time, limiting their effectiveness in meeting evolving customer needs and expectations (Kecht et al., 2023). ...
... The limited studies on ChatGPT found a few valid claims, such as constant learning features, knowledge base features, and customization (Alawi, 2023;Marchi & Sampieri, 2023). A study confirms that AI-based chatbots can overcome the hurdles faced by practitioners if they are trained on a large set of data and advanced NLP algorithms (Kecht et al., 2023). ...
Article
Full-text available
The launch of the open AI chatbot, ChatGPT, in November 2022 has generated widespread excitement around Generative Artificial Intelligence (AI). While researchers have explored ChatGPT's ability to produce content and respond to input, our study takes a different approach and examines its use cases in the financial industry. We aim to understand what ChatGPT offers the financial industry and how it differs from existing banking and financial chatbots. Financial institutions can use ChatGPT for a variety of purposes, including customer engagement, personalization, up-selling and cross-selling, stock forecasting, product development, and financial education. By focusing on the potential of ChatGPT in finance, we hope to spark discussions about its applications in other domains and explore the possibilities of a larger revolution in the future. Finally, this study identifies the challenges associated with the use of Generative Open AI and LLMs-based chatbots in the financial industry and provides recommendations for addressing these challenges.
... In the realm of business process management and optimization, GenAI is a game-changer, streamlining operations and sparking a wave of innovative methodologies. Recent research by Kecht et al. (2023) and van Dun et al. (2023) underscores GenAI's prowess in enhancing process documentation and inspiring novel process development. In ideation, GenAI leverages extensive online textual data to substantially broaden the knowledge repositories of project teams, thus democratizing and economizing innovation (Bouschery et al. 2023). ...
... With the continuous evolution of GenAI algorithms, there has been a notable rise in chatbot research Jeon et al., 2023). Traditionally, chatbots relied on Natural Language Processing (NLP) to interpret user queries and match them to the most suitable response sets within the system (Kecht et al., 2023). However, chatbots have further advanced by integrating language models and deep learning techniques to offer users instantaneous responses, enhancing their ability to handle NLP challenges in real-time while engaging with customers (Fitria et al., 2023). ...
Article
Full-text available
This study examines key reasons (for and against) that influence B2B managers' intention to adopt Generative AI (GenAI). We also investigate how GenAI adoption intention influences firm performance, along with the moderating effect of ethical leadership. Study 1 undertakes a series of in-depth interviews, yielding a set of hypotheses that are tested in study 2. A total of 277 responses was collected from respondents in the USA, UK, Canada, India, Australia, Malaysia, and Japan to test the proposed model using structural equation modeling. The findings highlight that need for uniqueness, information completeness, convenience, and deceptiveness significantly impact GenAI adoption intention. The results also highlight that GenAI adoption intention boosts firm performance. Finally, ethical leadership was found to moderate the effect of GenAI adoption intention on firm performance. This study enriches the GenAI, technology adoption, and behavioral reasoning theory literatures, while also providing pertinent insight for firms intending to adopt GenAI.
... Training chatbots using machine learning-based approaches typically requires a vast amount of training data to synthesize suitable responses [15]. ...
... This can be accomplished by analyzing documentation or communication logs, or by utilizing chatbots as interview partners for domain expert interviews. Similarly, other opportunities for applying LLMs in process discovery have been identified, such as producing process descriptions in various formats [6,16]. In the process analysis phase, LLMs can summarize massive amounts of data to identify patterns and issues in the process. ...
Conference Paper
Full-text available
Large language models (LLMs) have disrupted knowledge work in many application areas. Accordingly, the Business Process Management (BPM) community has started to explore how LLMs can be leveraged, resulting in a variety of promising research directions across the BPM lifecycle. Despite rapid adoption in practice and strong research interest, however, little is known about the actual design of BPM systems that leverage LLMs in organizational contexts. In this paper, we report on design science-based research in collaboration with a large multinational company to design a BPM system that leverages LLMs for process knowledge extraction from diverse enterprise content. Based on the development of our prototype, we observe that LLMs provide the means to organize and generate process knowledge independent of specific forms of representation. We present a conceptual framework that describes the role of LLMs in generating process knowledge from diverse input formats and, in turn, making it available in diverse output formats via prompting, resulting in representation-agnostic process knowledge. We also highlight implications of our study for BPM research and practice.
... Interfaces are filtering and matching processes, that is, how external attributes (SCALE) are transformed into internal control frameworks (IDEAS); interfaces are geared toward filtering and shaping (Kecht et al., 2023). ...
Article
Full-text available
The present research seeks to measure the relationship between the culture of experimentation and autonomy and the relationship between both variables when moderated by the organization's massive transformative purpose (MTP). This quantitative and exploratory research was conducted through a questionnaire for decision-makers in 43 medium and large IT companies in Jalisco. The results reveal a significant positive relationship between the culture of experimentation and autonomy, corroborating the first hypothesis of this research. However, according to the analyses to calculate the degree of moderation of the massive transformation purpose variable (moderator variable) to adjust the relationship between the culture of experimentation and autonomy, the results showed that there are no significant relationships, which rejects the second hypothesis of this research.
... The lack of regular updates and improvements emphasizes the necessity to maintain and upgrade the CRM system to meet company and customer needs. Without enough resources for these changes, the CRM system may become outdated and ineffective [3]. When considering the integration of an AI Chatbot into secretarial services, it is important to evaluate it from different perspectives. ...
... They suggest that future research should employ more robust methods and provide clearer descriptions of AI tasks to ensure that health chatbots are effective and can be deployed in as many contexts as possible. Moreover, chatbot's capacity to learn and comply with corporate processes by examining over 500,000 customer support interactions [17]. Their technique measured chatbots capacity to follow normative process models. ...
Preprint
Full-text available
In today's contemporary digital landscape, chatbots have become indispensable tools across various sectors, streamlining customer service, providing personal assistance, automating routine tasks, and offering health advice. However, their potential remains underexplored in the realm of network security, particularly for intrusion detection. To bridge this gap, we propose an architecture chatbot specifically designed to enhance security within edge networks specifically for intrusion detection. Leveraging advanced machine learning algorithms, this chatbot will monitor network traffic to identify and mitigate potential intrusions. By securing the network environment using an edge network managed by a Raspberry Pi module and ensuring ethical user consent promoting transparency and trust, this innovative solution aims to safeguard sensitive data and maintain a secure workplace, thereby addressing the growing need for robust network security measures in the digital age.
... Integrating chatbots into e-business platforms offers several advantages [6,7]. Firstly, they enhance customer satisfaction by offering immediate and accurate responses to frequently asked questions, such as inquiries about product details, shipping information, and return policies. ...
Article
Full-text available
E-businesses often face challenges related to customer service and communication, leading to increased dissatisfaction among customers and potential damage to the brand. To address these challenges, data-driven and AI-based approaches have emerged, including predictive analytics for optimizing customer interactions and chatbots powered by AI and NLP technologies. This study focuses on developing a hybrid rule-based and extractive-based chatbot for e-business, which can handle both routine and complex inquiries, ensuring quick and accurate responses to improve communication problems. The rule-based QA method used in the chatbot demonstrated high precision and accuracy in providing answers to user queries. The rule-based approach achieved impressive 98% accuracy and 97% precision rates among 1684 queries. The extractive-based approach received positive feedback, with 91% of users rating it as “good” or “excellent” and an average user satisfaction score of 4.38. General user satisfaction was notably high, with an average Likert score of 4.29, and 54% of participants gave the highest score of 5. Communication time was significantly improved, as the chatbot reduced average response times to 41 s, compared to the previous 20-min average for inquiries.
... As chatbots gather valuable data through interactions, they can offer insights into customer behavior and preferences, which can be used to inform business strategies [8,9]. Chatbots like IBM Watson Assistant have transformed customer support and engagement. ...
Article
Full-text available
This study explores the progress of chatbot technology, focusing on the aspect of error correction to enhance these smart conversational tools. Chatbots, powered by artificial intelligence (AI), are increasingly prevalent across industries such as customer service, healthcare, e-commerce, and education. Despite their use and increasing complexity, chatbots are prone to errors like misunderstandings, inappropriate responses, and factual inaccuracies. These issues can have an impact on user satisfaction and trust. This research provides an overview of chatbots, conducts an analysis of errors they encounter, and examines different approaches to rectifying these errors. These approaches include using data-driven feedback loops, involving humans in the learning process, and adjusting through learning methods like reinforcement learning, supervised learning, unsupervised learning, semi-supervised learning, and meta-learning. Through real life examples and case studies in different fields, we explore how these strategies are implemented. Looking ahead, we explore the different challenges faced by AI-powered chatbots, including ethical considerations and biases during implementation. Furthermore, we explore the transformative potential of new technological advancements, such as explainable AI models, autonomous content generation algorithms (e.g., generative adversarial networks), and quantum computing to enhance chatbot training. Our research provides information for developers and researchers looking to improve chatbot capabilities, which can be applied in service and support industries to effectively address user requirements.
... As noted by Selamat et al. (Selamat, M. A., 2021), chatbots can gather insights on customer preferences and behaviors, providing businesses with critical data to inform decision-making and strategy development. The work of Kecht et al. (Kecht, C., 2023) further illustrates how this data can be used to refine business processes, product offerings, and overall customer service strategies. The strategic use of chatbot-collected data has become a key component in understanding and responding to evolving customer needs and market trends. ...
Chapter
This chapter delves into the transformative impact of chatbots in customer service, highlighting their evolution from basic automated responders to advanced AI-driven conversational agents. Utilizing technologies like AI, ML, and NLP, these bots are reshaping customer interactions by offering round-the-clock service and handling complex inquiries with increasing sophistication. The chapter explores their development, operational mechanics, and various types while addressing the challenges in implementation and the balance between automation and human touch. Ethical considerations, particularly in data privacy, are critically examined. Real-world case studies across different industries illustrate the practical impact and efficiency gains from these bots. Future advancements are discussed, focusing on enhanced personalization and empathetic interactions through improved AI and NLP, underscoring the significant yet evolving role of chatbots in modern customer service.
... In adopting AI chatbots successfully, understanding specific user needs and aligning them with the chatbot's functionalities is crucial (Kecht et al., 2023). Continual review and optimization, guided by user feedback and performance monitoring, ensure chatbots remain effective in evolving user demographics and preferences. ...
... Thus, AI is becoming an essential concept in customer service in multiple ways. An example is using chatbots to handle simple requests, freeing up customer service agents to focus on more complex issues [45]. Furthermore, the use of natural language processing has enabled improvements in the chatbots' capacity to handle more sophisticated interactions [46]. ...
Article
Full-text available
The present study discusses the impact of Human-Robot Collaboration (HRC) powered by Artificial Intelligence (AI) on customer service. It is based on the four types of intelligence-mechanical, analytical, intuitive, and empathetic-and how they are integrated into HRC to provide customers with more efficient and personalized services. The benefits of AI-enabled HRC are highlighted, including reduced operational costs, increased productivity, improved decision-making, and enhanced customer experience. However, the article also addresses the challenges of implementing this approach, such as the potential loss of jobs due to automation, and emphasizes the importance of ethical and responsible implementation. The study has significant practical and academic implications, warning that continuous research is needed to understand the potential and limitations of AI-enabled HRC on customer service. Overall, through a literature review, the article aims to appeal to the reader's critical spirit and explore topics on the transformative power of AI in customer service. Abstract The present study discusses the impact of Human-Robot Collaboration (HRC) powered by Artificial Intelligence (AI) on customer service. It is based on the four types of intelligence-mechanical, analytical, intuitive, and empathetic-and how they are integrated into HRC to provide customers with more efficient and personalized services. The benefits of AI-enabled HRC are highlighted, including reduced operational costs, increased productivity, improved decision-making, and enhanced customer experience. However, the article also addresses the challenges of implementing this approach, such as the potential loss of jobs due to automation, and emphasizes the importance of ethical and responsible implementation. The study has significant practical and academic implications, warning that continuous research is needed to understand the potential and limitations of AI-enabled HRC on customer service. Overall, through a literature review, the article aims to appeal to the reader's critical spirit and explore topics on the transformative power of AI in customer service.
... Concrete implications and research directions can be connected to various phases of the BPM lifecycle model (Vidgof et al. 2023). For example, in the context of process discovery, generative AI models could be used to generate process descriptions, which can help businesses identify and understand the different stages of a process (Kecht et al. 2023). From the perspective of business process improvement, generative process models could be used for idea generation and to support innovative process (re-)design initiatives (van Dun et al. 2023). ...
... One example of such a system is PACA [38]. Automatically learning from user interactions cannot only be achieved for neural networks (e.g., reinforcement learning), but also by encoding interaction automatically into rules, such as in [33,4]. ...
Chapter
Chatbots such as ChatGPT have caused tremendous hype lately. For BPM applications, it is often not clear how to apply chatbots to generate business value. Hence, this work aims at the systematic analysis of existing chatbots for their support of conversational process modelling as a process-oriented capability. Application scenarios are identified along the process life cycle. Then a systematic literature review on conversational process modelling is performed. The resulting taxonomy serves as input for the identification of application scenarios for conversational process modelling, including paraphrasing and improvement of process descriptions. The application scenarios are evaluated for existing chatbots based on a real-world test set from the higher education domain. It contains process descriptions as well as corresponding process models, together with an assessment of the model quality. Based on the literature and application scenario analyses, recommendations for the usage (practical implications) and further development (research directions) of conversational process modelling are derived. KeywordsConversational process modellingChatbotsProcess DescriptionsProcess Models
... In the realm of chatbot technology, Kecht et al (2023) highlighted the importance of the capability of chatbots to learn and adhere to organizations' business processes. They developed an approach that quantifies chatbots' ability to learn business processes using standardized process mining metrics. ...
Preprint
Full-text available
This technical report describes the intersection of process mining and large language models (LLMs), specifically focusing on the abstraction of traditional and object-centric process mining artifacts into textual format. We introduce and explore various prompting strategies: direct answering, where the large language model directly addresses user queries; multi-prompt answering, which allows the model to incrementally build on the knowledge obtained through a series of prompts; and the generation of database queries, facilitating the validation of hypotheses against the original event log. Our assessment considers two large language models, GPT-4 and Google's Bard, under various contextual scenarios across all prompting strategies. Results indicate that these models exhibit a robust understanding of key process mining abstractions, with notable proficiency in interpreting both declarative and procedural process models. In addition, we find that both models demonstrate strong performance in the object-centric setting, which could significantly propel the advancement of the object-centric process mining discipline. Additionally, these models display a noteworthy capacity to evaluate various concepts of fairness in process mining. This opens the door to more rapid and efficient assessments of the fairness of process mining event logs, which has significant implications for the field. The integration of these large language models into process mining applications may open new avenues for exploration, innovation, and insight generation in the field.
... Meanwhile, [25] focuses on identifying constraints for business process execution from natural language documents. In [12], chatbots are trained on customer service conversations to learn the underlying business process, showing the effectiveness of such an approach, though the generalization capabilities remain unclear. ...
Preprint
Full-text available
Large Language Models (LLMs) are capable of answering questions in natural language for various purposes. With recent advancements (such as GPT-4), LLMs perform at a level comparable to humans for many proficient tasks. The analysis of business processes could benefit from a natural process querying language and using the domain knowledge on which LLMs have been trained. However, it is impossible to provide a complete database or event log as an input prompt due to size constraints. In this paper, we apply LLMs in the context of process mining by i) abstracting the information of standard process mining artifacts and ii) describing the prompting strategies. We implement the proposed abstraction techniques into pm4py, an open-source process mining library. We present a case study using available event logs. Starting from different abstractions and analysis questions, we formulate prompts and evaluate the quality of the answers.
... Chatbots are shown to be highly effective to automate frequently requested tasks. Using NLP, they adapt to tasks that are slightly different from their programmed base task (Kecht et al., 2023). Another case (Shaikh et al., 2022) presents a framework with various stages that offers guidelines with enough flexibility to be applicable in complex and heterogeneous corporate environments as well as for small and medium sized companies. ...
... Concrete implications and research directions can be connected to various phases of the BPM lifecycle model (Vidgof et al., 2023). For example, in the context of process discovery, generative AI models could be used to generate process descriptions, which can help businesses identify and understand the various stages of a process (Kecht et al., 2023). From the perspective of business process improvement, generative process models could be used for idea generation and to support innovative process (re-)design initiatives (van Dun et al., 2023). ...
Article
Full-text available
Published version: Feuerriegel, S., Hartmann, J., Janiesch, C. et al. Generative AI. Bus Inf Syst Eng (2023). https://doi.org/10.1007/s12599-023-00834-7
... Concrete implications and research directions can be connected to various phases of the BPM lifecycle model (Vidgof et al., 2023). For example, in the context of process discovery, generative AI models could be used to generate process descriptions, which can help businesses identify and understand the various stages of a process (Kecht et al., 2023). From the perspective of business process improvement, generative process models could be used for idea generation and to support innovative process (re-)design initiatives (van Dun et al., 2023). ...
Preprint
Full-text available
The term "generative AI'' refers to computational techniques that are capable of generating seemingly new, meaningful content such as text, images, or audio from training data. The widespread diffusion of this technology with examples such as Dall-E 2, GPT-4, and Copilot is currently revolutionizing the way we work and communicate with each other. In this article, we provide a conceptualization of generative AI as an entity in socio-technical systems and provide examples of models, systems, and applications. Based on that, we introduce limitations of current generative AI and provide an agenda for Business & Information Systems Engineering (BISE) research. Different from previous works, we focus on generative AI in the context of information systems, and, to this end, we discuss several opportunities and challenges that are unique to the BISE community and make suggestions for impactful directions for BISE research.
... One example of such a system is PACA [38]. Automatically learning from user interactions cannot only be achieved for neural networks (e.g., reinforcement learning), but also by encoding interaction automatically into rules, such as in [33,4]. ...
Preprint
Chatbots such as ChatGPT have caused a tremendous hype lately. For BPM applications, it is often not clear how to apply chatbots to generate business value. Hence, this work aims at the systematic analysis of existing chatbots for their support of conversational process modelling as process-oriented capability. Application scenarios are identified along the process life cycle. Then a systematic literature review on conversational process modelling is performed. The resulting taxonomy serves as input for the identification of application scenarios for conversational process modelling, including paraphrasing and improvement of process descriptions. The application scenarios are evaluated for existing chatbots based on a real-world test set from the higher education domain. It contains process descriptions as well as corresponding process models, together with an assessment of the model quality. Based on the literature and application scenario analyses, recommendations for the usage (practical implications) and further development (research directions) of conversational process modelling are derived.
... Throughout their experience, clients can communicate with chatbots (Alyahya et al., 2023a;Kecht et al., 2023). By utilizing learning algorithms and predictive modelling, chatbots may instantaneously match a customer's enquiry with accessible products that fulfil their needs (Alyahya et al., 2023b;Brachten et al., 2021). ...
Chapter
New paradigms of machine processing are made possible by recent advancements in AI, which allowed for a transition from data-driven, discriminative AI jobs to complex, creative tasks through generative AI. Generative AI is an exciting & quickly developing field with enormous potential for problem-solving, data synthesis, & creative expression. GANs have emerged as an AI research hotspot. GANs are based on theory of adversarial learning and consist of a discriminator & generator. Estimating possible distribution of actual data samples & creating new samples are objectives of GANs. GANs are extensively researched because of their numerous potential applications (voice & language processing, image & vision computing, etc). The chapter provides clarification on the notion of generative AI & emphasizes its significance in the context of AI. It offers an in-depth exploration of different methods employed in generative AI. It analyzes current state of the art in GANs & provides an outlook for future. It draws attention to the field's present shortcomings & possible future advancements.
Article
Industry 4.0 has led to a huge increase in data coming from machine maintenance. At the same time, advances in Natural Language Processing (NLP) and Large Language Models provide new ways to analyse this data. In our research, we use NLP to analyse maintenance work orders, and specifically the descriptions of failures and the corresponding repair actions. Many NLP studies have focused on failure descriptions for categorising them, extracting specific information about failure, or supporting failure analysis methodologies (such as FMEA). Whereas, the analysis of repair actions and its relationship with failure remains underexplored. Addressing this gap, our study makes three significant contributions. Firstly, we focused on the Italian language, which presents additional challenges due to the dominance of NLP systems that are mainly designed for English. Secondly, it proposes a method for automatically subdividing a repair action into a set of sub-tasks. Lastly, it introduces an approach that employs association rule mining to recommend sub-tasks to maintainers when addressing failures. We tested our approach with a case study from an automotive company in Italy. The case study provides insights into the current barriers faced by NLP applications in maintenance, offering a glimpse into the future opportunities for smart maintenance systems.
Article
This manuscript introduces a multimedia business process dataset provided by a German research institute. The dataset was systematically collected in a laboratory environment that reflects the workspace of IT staff managing IT Asset Management (ITAM) processes. It encompasses data from 121 process instances across six basic processes, captured using 37 video recordings from two camera perspectives, motion tracking, environmental sensors, an ITAM system trace, and event log data from user interactions. The data is made available in its raw state and processed form. The object-centric event log format (OCEL) provides discrete business process events from system activities. Event data from reality is supplied as raw video files and logs from environmental sensors. The video files were also manually labelled with identifiable business process activities and their associated entities. This multimedia dataset has been designed as a resource for developing, training, and evaluating process mining techniques based on unstructured data. Consequently, the dataset design emphasizes the traceability of activities and entities across the multimedia data sources.
Chapter
The inception of large language models has helped advance the state-of-the-art on numerous natural language tasks. This has also opened the door for the development of foundation models for other domains and data modalities (e.g., images and code). In this paper, we argue that business process data has unique characteristics that warrant the creation of a new class of foundation models to handle tasks like activity prediction, process optimization, and decision making. These models should also tackle the challenges of applying AI to business processes which include data scarcity, multi-modal representations, domain specific terminology, and privacy concerns. To support our claim, we show the effectiveness of few-shot learning and transfer learning in next activity prediction, crucial properties for the success of foundation models.
Chapter
Large Language Models (LLMs) are capable of answering questions in natural language for various purposes. With recent advancements (such as GPT-4), LLMs perform at a level comparable to humans for many proficient tasks. The analysis of business processes could benefit from a natural process querying language and using the domain knowledge on which LLMs have been trained. However, it is impossible to provide a complete database or event log as an input prompt due to size constraints. In this paper, we apply LLMs in the context of process mining by i) abstracting the information of standard process mining artifacts and ii) describing the prompting strategies. We implement the proposed abstraction techniques into pm4py, an open-source process mining library. We present a case study using available event logs. Starting from different abstractions and analysis questions, we formulate prompts and evaluate the quality of the answers.
Article
Cet article analyse les impacts de l’intelligence artificielle sur les services clients, leurs employés et leurs clients. L’analyse de la littérature a permis aux auteurs de synthétiser six sujets principaux. En parallèle, une approche qualitative a été choisie en réalisant 15 entretiens semi-directifs avec des responsables de service clients de différents secteurs utilisant des solutions d’IA. Cette recherche comporte trois volets : 1) une analyse documentaire et une synthèse des principaux résultats ; 2) une description de la réalité des solutions d’intelligence artificielle et leur utilisation par les praticiens ; 3) les limites identifiées afin de proposer des orientations pour les recherches futures. Les solutions d’intelligence artificielle offrent des fonctionnalités fragmentées et les praticiens disposent de peu d’indications sur la meilleure façon de les mettre en œuvre.
Article
Full-text available
Evaluating binary classifications is a pivotal task in statistics and machine learning, because it can influence decisions in multiple areas, including for example prognosis or therapies of patients in critical conditions. The scientific community has not agreed on a general-purpose statistical indicator for evaluating two-class confusion matrices (having true positives, true negatives, false positives, and false negatives) yet, even if advantages of the Matthews correlation coefficient (MCC) over accuracy and F 1 score have already been shown.In this manuscript, we reaffirm that MCC is a robust metric that summarizes the classifier performance in a single value, if positive and negative cases are of equal importance. We compare MCC to other metrics which value positive and negative cases equally: balanced accuracy (BA), bookmaker informedness (BM), and markedness (MK). We explain the mathematical relationships between MCC and these indicators, then show some use cases and a bioinformatics scenario where these metrics disagree and where MCC generates a more informative response.Additionally, we describe three exceptions where BM can be more appropriate: analyzing classifications where dataset prevalence is unrepresentative, comparing classifiers on different datasets, and assessing the random guessing level of a classifier. Except in these cases, we believe that MCC is the most informative among the single metrics discussed, and suggest it as standard measure for scientists of all fields. A Matthews correlation coefficient close to +1, in fact, means having high values for all the other confusion matrix metrics. The same cannot be said for balanced accuracy, markedness, bookmaker informedness, accuracy and F 1 score.
Article
Full-text available
This literature review presents the History, Technology, and Applications of Natural Dialog Systems or simply chatbots. It aims to organize critical information that is a necessary background for further research activity in the field of chatbots. More specifically, while giving the historical evolution, from the generative idea to the present day, we point out possible weaknesses of each stage. After we present a complete categorization system, we analyze the two essential implementation technologies, namely, the pattern matching approach and machine learning. Moreover, we compose a general architectural design that gathers critical details, and we highlight crucial issues to take into account before system design. Furthermore, we present chatbots applications and industrial use cases while we point out the risks of using chatbots and suggest ways to mitigate them. Finally, we conclude by stating our view regarding the direction of technology so that chatbots will become really smart.
Conference Paper
Full-text available
Process mining techniques are able to extract knowledge from event logs commonly available in today’s information systems. These techniques provide new means to discover, monitor, and improve processes in a variety of application domains. There are two main drivers for the growing interest in process mining. On the one hand, more and more events are being recorded, thus, providing detailed information about the history of processes. On the other hand, there is a need to improve and support business processes in competitive and rapidly changing environments. This manifesto is created by the IEEE Task Force on Process Mining and aims to promote the topic of process mining. Moreover, by defining a set of guiding principles and listing important challenges, this manifesto hopes to serve as a guide for software developers, scientists, consultants, business managers, and end-users. The goal is to increase the maturity of process mining as a new tool to improve the (re)design, control, and support of operational business processes.
Article
Full-text available
A conversation is based on internal knowledge that the participants already know or external knowledge that they have gained during the conversation. A chatbot that communicates with humans by using its internal and external knowledge is called a knowledge-grounded chatbot. Although previous studies on knowledge-grounded chatbots have achieved reasonable performance, they may still generate unsuitable responses that are not associated with the given knowledge. To address this problem, we propose a knowledge-grounded chatbot model that effectively reflects the dialogue context and given knowledge by using well-designed attention mechanisms. The proposed model uses three kinds of attention: Query-context attention, query-knowledge attention, and context-knowledge attention. In our experiments with the Wizard-of-Wikipedia dataset, the proposed model showed better performances than the state-of-the-art model in a variety of measures.
Article
Full-text available
Business process management (BPM) is a mature discipline that drives corporate success through effective and efficient business processes. BPM is commonly structured via capability frameworks, which describe and bundle capability areas relevant for implementing process orientation in organizations. Despite their comprehensive use, existing BPM capability frameworks are being challenged by socio-technical changes such as those brought about by digitalization. In line with the uptake of novel technologies, digitalization transforms existing and enables new processes due to its impact on individual behavior and needs, intra- and inter-company collaboration, and new forms of automation. This development led the authors to presume that digitalization calls for new capability areas and that existing frameworks need to be updated. Hence, this study explored which BPM capability areas will become relevant in view of digitalization through a Delphi study with international experts from industry and academia. The study resulted in an updated BPM capability framework, accompanied by insights into challenges and opportunities of BPM. The results show that, while there is a strong link between current and future capability areas, a number of entirely new and enhanced capabilities are required for BPM to drive corporate success in view of digitalization.
Article
Full-text available
Background: To evaluate binary classifications and their confusion matrices, scientific researchers can employ several statistical rates, accordingly to the goal of the experiment they are investigating. Despite being a crucial issue in machine learning, no widespread consensus has been reached on a unified elective chosen measure yet. Accuracy and F1 score computed on confusion matrices have been (and still are) among the most popular adopted metrics in binary classification tasks. However, these statistical measures can dangerously show overoptimistic inflated results, especially on imbalanced datasets. Results: The Matthews correlation coefficient (MCC), instead, is a more reliable statistical rate which produces a high score only if the prediction obtained good results in all of the four confusion matrix categories (true positives, false negatives, true negatives, and false positives), proportionally both to the size of positive elements and the size of negative elements in the dataset. Conclusions: In this article, we show how MCC produces a more informative and truthful score in evaluating binary classifications than accuracy and F1 score, by first explaining the mathematical properties, and then the asset of MCC in six synthetic use cases and in a real genomics scenario. We believe that the Matthews correlation coefficient should be preferred to accuracy and F1 score in evaluating binary classification tasks by all scientific communities.
Chapter
Full-text available
A chatbot can be defined as a computer program, designed to interact with users using natural language or text in a way that the user thinks he is having dialogue with a human. Most of the chatbots utilise the algorithms of artificial intelligence (AI) in order to generate required response. Earlier chatbots merely created an illusion of intelligence by employing much simpler pattern matching and string processing design techniques for their interaction with users using rule-based and generative-based models. However, with the emergence of new technologies more intelligent systems have emerged using complex knowledge-based models. This paper aims to discuss chatbots classification, their design techniques used in earlier and modern chatbots and how the two main categories of chatbots handle conversation context.
Conference Paper
Full-text available
There is a growing interest in chatbots, which are machine agents serving as natural language user interfaces for data and service providers. However, no studies have empirically investigated people’s motivations for using chatbots. In this study, an online questionnaire asked chatbot users (N = 146, aged 16–55 years) from the US to report their reasons for using chatbots. The study identifies key motivational factors driving chatbot use. The most frequently reported motivational factor is “productivity”; chatbots help users to obtain timely and efficient assistance or information. Chatbot users also reported motivations pertaining to entertainment, social and relational factors, and curiosity about what they view as a novel phenomenon. The findings are discussed in terms of the uses and gratifications theory, and they provide insight into why people choose to interact with automated agents online. The findings can help developers facilitate better human–chatbot interaction experiences in the future. Possible design guidelines are suggested, reflecting different chatbot user motivations.
Conference Paper
Full-text available
Over recent years, the world has seen multiple uses for conversational agents. Chatbots has been implemented into ecommerce systems, such as Amazon Echo's Alexa [1]. Businesses and organizations like Facebook are also implementing bots into their applications. While a number of amazing chatbot platform exists, there are still difficulties in creating data-driven-systems as they large amount of data is needed for development and training. This paper we describe an advanced platform for evaluating and annotating human-chatbot interactions, its main features and goals, as well as the future plans we have for it.
Conference Paper
Full-text available
Given an event log describing observed behaviour, process discovery aims to find a process model that ‘best’ describes this behaviour. A large variety of process discovery algorithms has been proposed. However, no existing algorithm returns a sound model in all cases (free of deadlocks and other anomalies), handles infrequent behaviour well and finishes quickly. We present a technique able to cope with infrequent behaviour and large event logs, while ensuring soundness. The technique has been implemented in ProM and we compare the technique with existing approaches in terms of quality and performance.
Article
Full-text available
In this paper, we present our solution for argumentative analysis of call center conversations in order to provide useful insights for enhancing Customer Interaction Analytics to a level that will enable more qualitative metrics and key performance indicators (KPIs) beyond the standard approach used in Customer Interaction Analytics. These metrics rely on understanding the dynamics of conversations by highlighting the way participants discuss about topics. By doing that we can detect relevant situations such as social behaviors, controversial topics, customer oriented behaviors, and also predict customer satisfaction.
Article
Full-text available
Artificial Intelligence in machines is a very challenging discussion. It involves the creation of machines which can simulate intelligence. This paper discusses some of the current trends and practices in AI and subsequently offers alternative theory for improvement in some of today’s prominent and widely accepted postulates. For this, focus on the structuring and functioning of a simple A.I. system - chatbots (or chatter bots) is made. The paper shows how current approach towards A.I. is not adequate and offers a new theory that discusses machine intelligence, throwing light to the future of intelligent systems.
Article
Full-text available
The paper motivates, presents, demonstrates in use, and evaluates a methodology for conducting design science (DS) research in information systems (IS). DS is of importance in a discipline oriented to the creation of successful artifacts. Several researchers have pioneered DS research in IS, yet over the past 15 years, little DS research has been done within the discipline. The lack of a methodology to serve as a commonly accepted framework for DS research and of a template for its presentation may have contributed to its slow adoption. The design science research methodology (DSRM) presented here incorporates principles, practices, and procedures required to carry out such research and meets three objectives: it is consistent with prior literature, it provides a nominal process model for doing DS research, and it provides a mental model for presenting and evaluating DS research in IS. The DS process includes six steps: problem identification and motivation, definition of the objectives for a solution, design and development, demonstration, evaluation, and communication. We demonstrate and evaluate the methodology by presenting four case studies in terms of the DSRM, including cases that present the design of a database to support health assessment methods, a software reuse measure, an Internet video telephony application, and an IS planning method. The designed methodology effectively satisfies the three objectives and has the potential to help aid the acceptance of DS research in the IS discipline.
Article
Full-text available
Contemporary workflow management systems are driven by explicit process models, i.e., a completely specified workflow design is required in order to enact a given workflow process. Creating a workflow design is a complicated time-consuming process and typically, there are discrepancies between the actual workflow processes and the processes as perceived by the management. Therefore, we propose a technique for rediscovering workflow models. This technique uses workflow logs to discover the workflow process as it is actually being executed. The workflow log contains information about events taking place. We assume that these events are totally ordered and each event refers to one task being executed for a single case. This information can easily be extracted from transactional information systems (e.g., Enterprise Resource Planning systems such as SAP and Baan). The rediscovering technique proposed in this paper can deal with noise and can also be used to validate workflow processes by uncovering and measuring the discrepancies between prescriptive models and actual process executions.
Article
Full-text available
Deep learning allows computational models that are composed of multiple processing layers to learn representations of data with multiple levels of abstraction. These methods have dramatically improved the state-of-the-art in speech recognition, visual object recognition, object detection and many other domains such as drug discovery and genomics. Deep learning discovers intricate structure in large data sets by using the backpropagation algorithm to indicate how a machine should change its internal parameters that are used to compute the representation in each layer from the representation in the previous layer. Deep convolutional nets have brought about breakthroughs in processing images, video, speech and audio, whereas recurrent nets have shone light on sequential data such as text and speech.
Article
Full-text available
Several variants of the Long Short-Term Memory (LSTM) architecture for recurrent neural networks have been proposed since its inception in 1995. In recent years, these networks have become the state-of-the-art models for a variety of machine learning problems. This has led to a renewed interest in understanding the role and utility of various computational components of typical LSTM variants. In this paper, we present the first large-scale analysis of eight LSTM variants on three representative tasks: speech recognition, handwriting recognition, and polyphonic music modeling. The hyperparameters of all LSTM variants for each task were optimized separately using random search and their importance was assessed using the powerful fANOVA framework. In total, we summarize the results of 5400 experimental runs (about 15 years of CPU time), which makes our study the largest of its kind on LSTM networks. Our results show that none of the variants can improve upon the standard LSTM architecture significantly, and demonstrate the forget gate and the output activation function to be its most critical components. We further observe that the studied hyperparameters are virtually independent and derive guidelines for their efficient adjustment.
Article
Robotic Process Automation (RPA) is an emerging technology in the field of Business Process Management (BPM) that enables the automation of intensive repetitive tasks (or simply routines). RPA solutions access the user interface (UI) layer of software (SW) applications and provide a virtual workforce of SW robots that are able to mimic human keyboard and mouse interactions with a UI as if a real person was doing them. To take full advantage of this technology, organizations leverage the support of skilled human experts that preliminarily observe how routines are executed on the UI of the involved SW applications, and then implement the executable RPA scripts required to automate the routines enactment by SW robots on a target computer system. However, the current practice is time-consuming and error-prone, as it strongly relies on the ability of the human experts to correctly interpret the routines (and their variants) to automate. In this paper, to tackle this issue, we use a design science research method to develop an approach, called SmartRPA, which is able to interpret the UI logs keeping track of many routine executions, and to automatically synthesize SW robots that emulate the most suitable routine variant for any specific intermediate user input that is required during the routine execution. The approach is implemented as an open-source tool and evaluated with four non-functional requirements employing both syntectic and real-world data.
Article
Process mining is one of the most active research streams in business process management. In recent years, numerous methods have been proposed for analyzing structured process data. In many cases, however, only the digitized parts of processes are directly captured by process-aware information systems, whereas manual activities often leave blind spots in the process analysis. While video data can contain valuable process-related information that is not captured in information systems, a standardized approach to extracting event logs from unstructured video data remains lacking. To solve this problem and facilitate the systematic usage of video data in process mining, we have designed the ViProMiRA, a reference architecture that bridges the gap between computer vision and process mining. The various evaluation activities in our design science research process ensure that the proposed ViProMiRA allows flexible, use case-driven, and context-specific instantiations. Our results also show that a prototypical implementation of the ViProMiRA is capable of automatically extracting more than 70% of the process-relevant events from a real-world video dataset in a supervised learning scenario.
Conference Paper
A fundamental requirement for the successful application of process mining are event logs of high data quality that can be constructed from structured data stored in organizations' core information systems. However, a substantial amount of data is processed outside these core systems, particularly in organizations doing consumer business with many customer interactions per day, which generate high amounts of unstructured text data. Although Natural Language Processing (NLP) and machine learning enable the exploitation of text data, these approaches remain challenging due to the required high amount of labeled training data. Recent advances in NLP mitigate this issue by providing pre-trained and ready-to-use language models for various tasks such as Natural Language Inference (NLI). In this paper, we develop an approach that utilizes NLI to derive topics and process activities from customer service conversations and that represents them in a standardized XES event log. To this end, we compute the probability that a sentence describing the topic or the process activity can be inferred from the customer's inquiry or the agent's response using NLI. We evaluate our approach utilizing an existing corpus of more than 500,000 customer service conversations of three companies on Twitter. The results show that NLI helps construct event logs of high accuracy for process mining purposes, as our successful application of three different process discovery algorithms confirms.
Chapter
Robotic Process Automation (RPA) is an emerging technology for automating tasks using bots that can mimic human actions on computer systems. Most existing research focuses on the earlier phases of RPA implementations, e.g. the discovery of tasks that are suitable for automation. To detect exceptions and explore opportunities for bot and process redesign, historical data from RPA-enabled processes in the form of bot logs or process logs can be utilized. However, the isolated use of bot logs or process logs provides only limited insights and not a good understanding of an overall process. Therefore, we develop an approach that merges bot logs with process logs for process mining. A merged log enables an integrated view on the role and effects of bots in an RPA-enabled process. We first develop an integrated data model describing the structure and relation of bots and business processes. We then specify and instantiate a ‘bot log parser’ translating bot logs of three leading RPA vendors into the XES format. Further, we develop the ‘log merger’ functionality that merges bot logs with logs of the underlying business processes. We further introduce process mining measures allowing the analysis of a merged log.
Article
Business process management (BPM) is a mature discipline that drives corporate success through effective and efficient business processes. BPM is commonly structured via capability frameworks, which describe and bundle capability areas relevant for implementing process orientation in organizations. Despite their comprehensive use, existing BPM capability frameworks are being challenged by socio-technical changes such as those brought about by digitalization. In line with the uptake of novel technologies, digitalization transforms existing and enables new processes due to its impact on individual behavior and needs, intra- and inter-company collaboration, and new forms of automation. This development led the authors to presume that digitalization calls for new capability areas and that existing frameworks need to be updated. Hence, this study explored which BPM capability areas will become relevant in view of digitalization through a Delphi study with international experts from industry and academia. The study resulted in an updated BPM capability framework, accompanied by insights into challenges and opportunities of BPM. The results show that, while there is a strong link between current and future capability areas, a number of entirely new and enhanced capabilities are required for BPM to drive corporate success in view of digitalization.
Article
Today, customers can conveniently compare products and decide how to interact with companies. With customer centricity becoming an important success factor, companies must drive customer satisfaction not only through excellent products but also through customer-centric processes. As many companies face an abundance of action possibilities, fast-changing customer needs, and scarce resources, guidance regarding the customer-centric prioritization of process improvement projects is in high need. As existing approaches predominantly focus on process efficiency, we propose a decision model that accounts for the effects of process improvement on customer centricity in line with justificatory knowledge on value-based process decision-making, project portfolio selection, and the measurement of customer satisfaction. When building the decision model, we adopted the design science paradigm and used multi-criteria decision analysis as well as normative analytical modeling as research methods. We evaluated the model by discussing it with practitioners, by building a software prototype, and by applying it at a German insurance company. Overall, our research extends the prescriptive knowledge on process prioritization and customer process management.
Article
Purpose This paper aims to clarify the notion of artificial intelligence (AI), reviewing the present scope of the phenomenon through its main applications. It aims at describing the various applications while assessing the markets, highlighting some of the leading industrial sectors in the field. Therefore, it identifies pioneering companies and the geographical distribution of AI companies. Design/methodology/approach The paper builds upon an in-depth investigation of public initiatives focusing mostly on the EU. It is based on desk research, a comprehensive review of the main grey and scientific literature in this field. Findings The paper notes that there is no real consensus on any definition for this umbrella term, that the definition does fluctuate over time but highlights some of the main changes and advances that took place over the past 60 years. It stresses that, in spite of the hype, on both the business and consumer sides, the demand appears uncertain. The scope of the announced disruptions is not easy to assess, technological innovation associated with AI may be modest or take some time to be fully deployed. However, some companies and regions are leading already in the field. Research limitations/implications The paper, based on desk research, does not consider any expert opinions. Besides, the scientific literature on the phenomenon is still scarce (but not the technical one in the specific research sectors of AI). Most of the data come from consultancies or government publications which may introduce some bias, although the paper gathered various, often conflicting viewpoints. Originality/value The paper gives a thorough review of the available literature (consultancies, governments) stressing the limitations of the available research on economic and social aspects. It aims at providing a comprehensive overview of the major trends in the field. It gives a global overview of companies and regions.
Article
Driven by ‘success stories’ reported by private sector firms, government agencies have also started adopting various Artificial Intelligence (AI) technologies in diverse domains (e.g. health, taxation, and education); however, extensive research is required in order to exploit the full potential of AI in the public sector, and leverage various AI technologies to address important problems/needs. This paper makes a contribution in this direction: it presents a novel approach, as well as the architecture of an ICT platform supporting it, for the advanced exploitation of a specific AI technology, namely chatbots, in the public sector in order to address a crucial issue: the improvement of communication between government and citizens (which has for long time been problematic). The proposed approach builds on natural language processing, machine learning and data mining technologies, and leverages existing data of various forms (such as documents containing legislation and directives, structured data from government agencies’ operational systems, social media data, etc.), in order to develop a new digital channel of communication between citizens and government. Making use of appropriately structured and semantically annotated data, this channel enables ‘richer’ and more expressive interaction of citizens with government in everyday language, facilitating and advancing both information seeking and conducting of transactions. Compared to existing digital channels, the proposed approach is appropriate for a wider range of citizens’ interactions, with higher levels of complexity, ambiguity and uncertainty. In close co-operation with three Greek government agencies (the Ministry of Finance, a social security organization, and a big local government organization), this approach has been validated through a series of application scenarios.
Article
In the digital age, customers want to define on their own how to interact with organizations during their customer journeys. Thus, many organizations struggle to implement an omni-channel strategy (OCS) that meets their customers' channel preferences and can be operated efficiently. Despite this high practical need, research on omni-channel management predominantly takes a descriptive perspective. What is missing is prescriptive knowledge that guides organizations in the valuation and selection of an appropriate OCS. Most existing studies investigate single facets of omni-channel management in detail while neglecting the big picture. They also require customer journeys to follow sequential and organization-defined purchase decision processes. To address this research gap, we propose an economic decision model that considers online and offline channels, the opening and closing of channels, non-sequential customer journeys, and customers' channel preferences. Drawing from the principles of value-based management, the decision model recommends choosing the OCS with the highest contribution to an organization's long-term firm value. We applied and validated the decision model based on real-world data from a German bank.
Conference Paper
A conversational agent also referred to as chatbot is a computer program which tries to generate human like responses during a conversation. Earlier chatbots employed much simpler retrieval based pattern matching design techniques. However, with time a number of new chatbots evolved with an aim to make it more human like and hence to pass the Turing test. Now, most of the chatbots employ generative knowledge based techniques. This paper will discuss about various chatbot design techniques, classification of chatbot and discussion on how the modern chatbots have evolved from simple pattern matching, retrieval based model to modern complex knowledge based models. A table of major conversational agents in chronological order along with their design techniques is also provided at the end of the paper.
Conference Paper
Deep Neural Networks (DNNs) are powerful models that have achieved excellent performance on difficult learning tasks. Although DNNs work well whenever large labeled training sets are available, they cannot be used to map sequences to sequences. In this paper, we present a general end-to-end approach to sequence learning that makes minimal assumptions on the sequence structure. Our method uses a multilayered Long Short-Term Memory (LSTM) to map the input sequence to a vector of a fixed dimensionality, and then another deep LSTM to decode the target sequence from the vector. Our main result is that on an English to French translation task from the WMT-14 dataset, the translations produced by the LSTM achieve a BLEU score of 34.7 on the entire test set, where the LSTM's BLEU score was penalized on out-of-vocabulary words. Additionally, the LSTM did not have difficulty on long sentences. For comparison, a strong phrase-based SMT system achieves a BLEU score of 33.3 on the same dataset. When we used the LSTM to rerank the 1000 hypotheses produced by the aforementioned SMT system, its BLEU score increases to 36.5, which beats the previous state of the art. The LSTM also learned sensible phrase and sentence representations that are sensitive to word order and are relatively invariant to the active and the passive voice. Finally, we found that reversing the order of the words in all source sentences (but not target sentences) improved the LSTM's performance markedly, because doing so introduced many short term dependencies between the source and the target sentence which made the optimization problem easier.
Conference Paper
We propose two novel model architectures for computing continuous vector representations of words from very large data sets. The quality of these representations is measured in a word similarity task, and the results are compared to the previously best performing techniques based on different types of neural networks. We observe large improvements in accuracy at much lower computational cost, i.e. it takes less than a day to learn high quality word vectors from a 1.6 billion words data set. Furthermore, we show that these vectors provide state-of-the-art performance on our test set for measuring syntactic and semantic word similarities.
Article
This paper proposes a simple and efficient approach for text classification and representation learning. Our experiments show that our fast text classifier fastText is often on par with deep learning classifiers in terms of accuracy, and many orders of magnitude faster for training and evaluation. We can train fastText on more than one billion words in less than ten minutes using a standard multicore CPU, and classify half a million sentences among 312K classes in less than a minute.
Article
The amount of data that is generated during the execution of a business process is growing. As a consequence it is increasingly hard to extract useful information from the large amount of data that is produced. Linguistic summarization helps to point business analysts in the direction of useful information, by verbalizing interesting patterns that exist in the data. In previous work we showed how linguistic summarization can be used to automatically generate diagnostic statements about event logs, such as ‘for most cases that contained the sequence ABC, the throughput time was long’. However, we also showed that our technique produced too many of these statements to be useful in a practical setting. Therefore this paper presents a novel technique for linguistic summarization of event logs, which generates linguistic summaries that are concise enough to be used in a practical setting, while at the same time enriching the summaries that are produced by also enabling conjunctive statements. The improved technique is based on pruning and clustering of linguistic summaries. We show that it can be used to reduce the number of summary statements 80–100% compared to previous work. In a survey among 51 practitioners, we found that practitioners consider linguistic summarization useful and easy to use and intend to use it if it were commercially available.
Conference Paper
While models and event logs are readily available in modern organizations, their quality can seldom be trusted. Raw event recordings are often noisy, incomplete, and contain erroneous recordings. The quality of process models, both conceptual and data-driven, heavily depends on the inputs and parameters that shape these models, such as domain expertise of the modelers and the quality of execution data. The mentioned quality issues are specifically a challenge for conformance checking. Conformance checking is the process mining task that aims at coping with low model or log quality by comparing the model against the corresponding log, or vice versa. The prevalent assumption in the literature is that at least one of the two can be fully trusted. In this work, we propose a generalized conformance checking framework that caters for the common case, when one does neither fully trust the log nor the model. In our experiments we show that our proposed framework balances the trust in model and log as a generalization of state-of-the-art conformance checking techniques.
Conference Paper
Considerable amounts of data, including process event data, are collected and stored by organisations nowadays. Discovering a process model from recorded process event data is the aim of process discovery algorithms. Many techniques have been proposed, but none combines scalability with quality guarantees, e.g. can handle billions of events or thousands of activities, and produces sound models (without deadlocks and other anomalies), and guarantees to rediscover the underlying process in some cases. In this paper, we introduce a framework for process discovery that computes a directly-follows graph by passing over the log once, and applying a divide-and-conquer strategy. Moreover, we introduce three algorithms using the framework.We experimentally show that it sacrifices little compared to algorithms that use the full event log, while it gains the ability to cope with event logs of 100,000,000 traces and processes of 10,000 activities.
Article
Machine learning addresses the question of how to build computers that improve automatically through experience. It is one of today’s most rapidly growing technical fields, lying at the intersection of computer science and statistics, and at the core of artificial intelligence and data science. Recent progress in machine learning has been driven both by the development of new learning algorithms and theory and by the ongoing explosion in the availability of online data and low-cost computation. The adoption of data-intensive machine-learning methods can be found throughout science, technology and commerce, leading to more evidence-based decision-making across many walks of life, including health care, manufacturing, education, financial modeling, policing, and marketing.
Article
An attentional mechanism has lately been used to improve neural machine translation (NMT) by selectively focusing on parts of the source sentence during translation. However, there has been little work exploring useful architectures for attention-based NMT. This paper examines two simple and effective classes of attentional mechanism: a global approach which always attends to all source words and a local one that only looks at a subset of source words at a time. We demonstrate the effectiveness of both approaches over the WMT translation tasks between English and German in both directions. With local attention, we achieve a significant gain of 5.0 BLEU points over non-attentional systems which already incorporate known techniques such as dropout. Our ensemble model using different attention architectures has established a new state-of-the-art result in the WMT'15 English to German translation task with 25.9 BLEU points, an improvement of 1.0 BLEU points over the existing best system backed by NMT and an n-gram reranker.
Article
Process discovery techniques automatically extract the real workflow of a process by analyzing the events that are collected and stored in log files. Although in the last years several process discovery algorithms have been presented, none of them guarantees to find complete, precise and simple models for all the given logs. In this paper we address the problem of process discovery through a genetic algorithm with a new fitness function that takes into account both completeness, precision and simplicity. ProDiGen (Process Discovery through a Genetic algorithm) includes new definitions for precision and simplicity, and specific crossover and mutation operators. The proposal has been validated with 39 process models and several noise levels, giving a total of 111 different logs. We have compared our approach with the state of the art algorithms; non-parametric statistical tests show that our algorithm outperforms the other approaches, and that the difference is statistically significant.