DrugExBERT for Pharmacovigilance – A Novel Approach for Detecting Drug Experiences from User-Generated Content

2023 | International Conference on Information Systems | Citations: 0

Authors: Bohnen, Eva; Erlebach, Stefanie; Zimmermann, Steffen

Abstract: Pharmaceutical companies have to maintain drug safety through pharmacovigilance ... Expand

Abstract: Pharmaceutical companies have to maintain drug safety through pharmacovigilance systems by monitoring various sources of information about adverse drug experiences. Recently, user-generated content (UGC) has emerged as a valuable source of real-world drug experiences, posing new challenges due to its high volume and variety. We present DrugExBERT, a novel approach to extract adverse drug experiences (adverse reaction, lack of effect) and supportive drug experiences (effectiveness, intervention, indication, and off-label use) from UGC. To be able to verify the extracted drug experiences, DrugExBERT additionally provides explications in the form of UGC phrases that were critical for the extraction. In our evaluation, we demonstrate that DrugExBERT outperforms state-of-the-art pharmacovigilance approaches as well as ChatGPT on several performance measures and that DrugExBERT is data-and drug-agnostic. Thus, our novel approach can help pharmaceutical companies meet their legal obligations and ethical responsibility while ensuring patient safety and monitoring drug effectiveness. Collapse

Topics: health information system application programming interface generative AI database system

Methods: chatGPT natural language processing RoBERTa BERT machine learning

From semantics to pragmatics: where IS can lead in Natural Language Processing (NLP) research

2021 | European Journal of Information Systems | Citations: 0

Authors: Li, Yan; Thomas, Manoj a; Liu, Dapeng

Abstract: Natural Language Processing (NLP) is now widely integrated into web and mobile a ... Expand

Abstract: Natural Language Processing (NLP) is now widely integrated into web and mobile applications, enabling natural interactions between humans and computers. Although there is a large body of NLP studies published in Information Systems (IS), a comprehensive review of how NLP research is conceptualised and realised in the context of IS has not been conducted. To assess the current state of NLP research in IS, we use a variety of techniques to analyse a literature corpus comprising 356 NLP research articles published in IS journals between 2004 and 2018. Our analysis indicates the need to move from semantics to pragmatics. More importantly, our findings unpack the challenges and assumptions underlying current research trends in NLP. We argue that overcoming these challenges will require a renewed disciplinary IS focus. By proposing a roadmap of NLP research in IS, we draw attention to three NLP research perspectives and present future directions that IS researchers are uniquely positioned to address. Collapse

Topics: database system knowledge representation social media analytics social media information retrieval

Methods: natural language processing qualitative content analysis literature study computational algorithm literature sample

A Weakly Supervised WordNet-Guided Deep Learning Approach to Extracting Aspect Terms from Online Reviews

2020 | ACM Transactions on Management Information Systems | Citations: 0

Authors: Tao, Jie; Zhou, Lina

Abstract: The unstructured nature of online reviews makes it inefficient and inconvenient ... Expand

Abstract: The unstructured nature of online reviews makes it inefficient and inconvenient for prospective consumers to research and use in support of purchase decision making. The aspects of products provide a fine-grained meaningful perspective for understanding and organizing review texts. Traditional aspect term extraction approaches rely on discrete language models that treat words in isolation. Despite that continuous-space language models have demonstrated promise in addressing a wide range of problems, their application in aspect term extraction faces significant challenges. For instance, existing continuous-space language models typically require large collections of labeled data, which remain difficult to obtain in many domains. More importantly, previous methods are largely data driven but overlook the role of human knowledge in guiding model development. To address these limitations, this study designs and develops weakly supervised WordNet-guided deep learning to aspect term extraction. The approach draws on deep-level semantic information from WordNet to guide not only the selection representative seed terms but also the pruning of aspect candidate terms. The weak supervision is provided by a very small set of labeled data. We conduct a comprehensive evaluation of the proposed method using both direct and indirect methods. The evaluation results with Yelp restaurant reviews demonstrate that our proposed method consistently outperforms all baseline methods including discrete models and the state-of-the-art continuous-space language models for aspect term extraction across both direct and indirect evaluations. The research findings have broad research, technical, and practical implications for various stakeholders of online reviews. Collapse

Topics: knowledge representation online review organizational productivity user-generated content decision making

Methods: language model WordNet structural equation modeling word embedding deep learning

Don’t Mention It? Analyzing User-Generated Content Signals for Early Adverse Event Warnings

2019 | Information Systems Research | Citations: 0

Authors: Abbasi, Ahmed; Li, Jingjing; Adjeroh, Donald; Abate, Marie; Zheng, Wanhong

Abstract: With greater impetus on broad postmarket surveillance, the Voice of the Customer ... Expand

Abstract: With greater impetus on broad postmarket surveillance, the Voice of the Customer (VoC) has emerged as an important source of information for understanding consumer experiences and identifying potential issues. In organizations, risk management groups are increasingly interested in working with their information technology teams to develop robust VoC listening platforms. Two key challenges have impeded success. First, prior work has leveraged diverse sets of channels, adverse event types, and modeling methods, resulting in diverging conclusions regarding the viability and efficacy of various user-generated channels and accompanying modeling methods. Second, many existing detection methods rely on “mention models” that have low detection rates, have high false positives, and lack timeliness. Following the information systems design science approach, in this research note we propose a framework for examining key design elements for VoC listening platforms. As part of our framework, we also develop a novel heuristic-based method for detecting adverse events. We evaluate our framework and method on two large test beds each encompassing millions of tweets, forums postings, and search query logs pertaining to hundreds of adverse events related to the pharmaceutical and automotive industries. The results shed light on the interplay between user-generated channels and event types, as well as the potential for more robust event modeling methods that go beyond basic mention models. Our analysis framework reveals that user-generated content channels can facilitate timelier detection of adverse events: on average, two to three years or earlier than commonly used databases. The inclusion of negative sentiment polarity in the models can further reduce false-positive rates. Additionally, we find social media channels provide higher detection rates but lower precision than do search-based signals. The search and web forum channels are timelier than Twitter. The proposed heuristic-based method attains markedly better results than do existing methods—with earlier detection rates of 50%–80% and far fewer false positives across an array of VoC channels and event types. The heuristic method is also well suited for signal fusion across channels. Our note makes several contributions to research. The results also have important implications for various practitioner groups, including regulatory agencies and risk management teams at product manufacturing firms. Collapse

Topics: Twitter social media user-generated content database system digital platform

Methods: machine learning time series analysis longitudinal research design science parametric test

Developing insights from social media using semantic lexical chains to mine short text structures

2019 | Decision Support Systems | Citations: 0

Authors: Chua, Cecil Eng Huang; Storey, Veda C.; Li, Xiaolin; Kaul, Mala

Abstract: Social media is increasingly being used for communication by individuals and org ... Expand

Abstract: Social media is increasingly being used for communication by individuals and organizations. Social media stores vast amounts of publicly available data that provides a rich source of information and insights. Often, social media users can easily infer meaning from short text such as microblogs and Facebook posts because they understand the context and terminology used. Although automated data-mining can be effective for gaining insights from text data, a significant challenge is to accurately infer meaning from social media text derived from a single social media account. This is difficult because social media communication uses very short, or sparse, text, which yields a relatively small sample of usable words for analysis. Furthermore, interpreting the contextual meaning from a relatively small set of words is challenging. This research proposes a methodology for extracting semantic lexical chains from frequently occurring words in a single social media account and using these chains to mine short text structures to infer the overall themes of the user. The methodology is based on a proposed clustering algorithm and illustrated with examples from Facebook posts. The algorithm is tested and illustrated by comparing it to existing work and further applying it to a variety of news posts. This methodology could be useful for gaining decision-making insights from social media, or other online forms with short or sparse text. Collapse

Topics: Facebook social media knowledge representation decision support system

Methods: computational algorithm cluster analysis WordNet word sense disambiguation machine learning

Discovering composable web services using functional semantics and service dependencies based on natural language requests

2019 | Information Systems Frontiers | Citations: 0

Authors: S, Sowmya Kamath; V. S., Ananthanarayana

Abstract: The processes of service discovery, selection and composition are crucial tasks ... Expand

Abstract: The processes of service discovery, selection and composition are crucial tasks in web service based application development. Most web service-driven applications are complex and are composed of more than one service, so, it becomes important for application designers to identify the best service to perform the next task in the intended application’s workflow. In this paper, a framework for discovering composable service sets as per user’s complex requirements is proposed. The proposed approach uses natural language processing and semantics based techniques to extract the functional semantics of the service dataset and also to understand user context. In case of simple queries, basic services may be enough to satisfy the user request, however, in case of complex queries, several basic services may have to be identified to serve all the requirements, in the correct sequence. For this, the service dependencies of all the services are used for constructing a service interface graph for finding suitable composable services. Experiments showed that the proposed approach was effective towards finding relevant services for simple & complex queries and achieved an average accuracy rate of 75.09 % in finding correct composable service templates. Collapse

Topics: web service knowledge representation user requirement hard disk drive workstation

Methods: experiment computational algorithm machine learning WordNet term frequency–inverse document frequency

Unsupervised tip-mining from customer reviews

2018 | Decision Support Systems | Citations: 0

Authors: Zhu, Di; Lappas, Theodoros; Zhang, Juheng

Abstract: In recent years, large review-hosting platforms have extended their functionalit ... Expand

Abstract: In recent years, large review-hosting platforms have extended their functionality to allow their users to submit tips: short pieces of text that deliver valuable insight on a specific aspect of the reviewed business. These tips are meant to serve as a concise source of information that complements the often overwhelming number of customer reviews. Recent work has tackled the problem of automatically generating tips by mining review text. The motivation for this effort is to obtain tips for businesses or business aspects that have been overlooked by users. Another motivating factor is the quality of the user-submitted tips, which often provide trivial or redundant information. Existing tip-mining methods are limited by a reliance on training data, which is unlikely to be available and is also very costly to create for different domains. In this work, we present TipSelector, a completely unsupervised algorithm that delivers high quality-tips without the need for annotated training data. We verify the efficacy of TipSelector via an evaluation that includes real data from the hospitality industry and comparisons with the state-of-the-art. A secondary contribution of our work is a method for automatically evaluating tip-mining algorithms without humans in the loop. As we demonstrate in our experiments, this method can be used to enable large-scale evaluations and complement the user studies that are typically used for this purpose. Collapse

Topics: knowledge representation website electronic commerce business-to-business

Methods: computational algorithm experiment WordNet descriptive statistic scale validation

LEVERAGING TEXT MINING FOR THE DESIGN OF A LEGAL KNOWLEDGE MANAGEMENT SYSTEM

2017 | European Conference On Information Systems | Citations: 0

Authors: Hanke, Jannis; Thiesse, Frédéric

Abstract: In today’s globalized world, companies are faced with numerous and continuously ... Expand

Abstract: In today’s globalized world, companies are faced with numerous and continuously changing legal requirements. To ensure that these companies are compliant with legal regulations, law and consulting firms use open legal data published by governments worldwide. With this data pool growing rapidly, the complexity of legal research is strongly increasing. Despite this fact, only few research papers consider the application of information systems in the legal domain. Against this backdrop, we pro-pose a knowledge management (KM) system that aims at supporting legal research processes. To this end, we leverage the potentials of text mining techniques to extract valuable information from legal documents. This information is stored in a graph database, which enables us to capture the relation-ships between these documents and users of the system. These relationships and the information from the documents are then fed into a recommendation system which aims at facilitating knowledge trans-fer within companies. The prototypical implementation of the proposed KM system is based on 20,000 legal documents and is currently evaluated in cooperation with a Big 4 accounting company. Collapse

Topics: knowledge management system graph database website knowledge management user interface

Methods: natural language processing computational algorithm artificial neural network design science spaCy

A Roadmap for Natural Language Processing Research in Information Systems

2017 | HICSS | Citations: 2

Authors: Liu, Dapeng; Li, Yan; Thomas, Manoj A.

Abstract: Natural Language Processing (NLP) is now widely integrated into web and mobile a ... Expand

Abstract: Natural Language Processing (NLP) is now widely integrated into web and mobile applications, enabling natural interactions between human and computers. Although many NLP studies have been published, none have comprehensively reviewed or synthesized tasks most commonly addressed in NLP research. We conduct a thorough review of IS literature to assess the current state of NLP research, and identify 12 prototypical tasks that are widely researched. Our analysis of 238 articles in Information Systems (IS) journals between 2004 and 2015 shows an increasing trend in NLP research, especially since 2011. Based on our analysis, we propose a roadmap for NLP research, and detail how it may be useful to guide future NLP research in IS. In addition, we employ Association Rules (AR) mining for data analysis to investigate co-occurrence of prototypical tasks and discuss insights from the findings. Collapse

Topics: information retrieval database system

Methods: natural language processing design science literature study design artifact qualitative content analysis

From valence to emotions: Exploring the distribution of emotions in online product reviews

2016 | Decision Support Systems | Citations: 26

Authors: Ullah, Rahat; Amblee, Naveen; Kim, Wonjoon; Lee, Hyunjong

Abstract: Word-of-mouth (WOM) in the form of online customer reviews has received consider ... Expand

Abstract: Word-of-mouth (WOM) in the form of online customer reviews has received considerable attention by practitioners and academics. Prior literature has focused more on the understanding of the phenomenon using the frequency or overall rating/valence information of WOM, while questions on how firms can potentially use or design online WOM platforms and benefit from it based on the content of WOM are still open, and need more attention from researchers. In addition, an important antecedent for the generation of word-of-mouth is a strong emotional imbalance known as schema discrepancy, which is considered to trigger the consumer to post a customer review online. However, only a limited number of studies to date have actually examined the emotional content of reviews to validate this line of reasoning. To fill this gap, we analyzed the emotional content of a large number of online product reviews using Natural Language Processing (NLP) methods. We find that there is a difference in the emotional content of reviews across search and experience goods in the early stages of product launch. However, interestingly, these differences disappear over time as the addition of reviews reduces the information asymmetry gap. This suggests that traditional experience goods are evaluated more like search goods in online environments, because consumers can easily evaluate attributes of products prior to purchase based on the reviews accumulated. In addition, we find that more extreme reviews have a greater proportion of emotional content than less extreme reviews, revealing a bimodal distribution of emotional content, thereby empirically validating a key assumption that underpins much of the extant literature on online WOM. Furthermore, reviews have a greater proportion of positive emotional content within positive extreme ratings as compared to negative emotional content within negative extreme ratings which is a major factor in online WOM generation, and helps explain the commonly observed J-shaped distribution of reviews. Our findings suggest important managerial implications regarding product development, advertisement, and platform design using WOM content. Collapse

Topics: experience good online word of mouth online review marketing management new product development

Methods: natural language processing computational algorithm product review parametric test ANOVA

Theories: lemon market theory discrepancy theory