StreamTag: A Platform for Flexible Tagged Data Management

David Díaz-Jiménez

1 a

, Francisco Mata-Mata

1 b

, José L. López

1 c

, Luis G. Pérez-Cordón

1 d

José-María Serrano

1 e

, Carmen Martínez-Cruz

2 f

, Juana M. Morcillo-Martínez

3 g

Ángeles Verdejo-Espinosa

4 h

, Juan C. Cuevas-Martínez

5 i

, Raquel Viciana-Abad

5 j

Pedro J. Reche-López

5 k

, José M. Pérez-Lorenzo

5 l

, Juan F. Gaitán-Guerrero

1 m

and

Macarena Espinilla

1 n

Department of Computer Science, University of Jaén, 23071, Jaén, Spain

Department of Languages and Computer Systems, University of Granada, 18071, Granada, Spain

Psycology Department, Faculty of Social Work, University of Jaén, 23071 Jaén, Spain

Electrical Engineering Department, University of Jaén, 23071 Jaén, Spain

Telecommunication Engineering Department, University of Jaén, 23071 Jaén, Spain

Keywords:

Human Activity Recognition, Data Labeling, Healthcare Monitoring, Internet of Things, Artiﬁcial

Intelligence, Personalized Care, Nursing Home Monitoring.

Abstract:

This paper presents StreamTag, a platform designed for the efﬁcient management of labeled data in healthcare

environments, particularly for activity recognition systems in residential and nursing home settings. Human

Activity Recognition (HAR) is crucial for monitoring patient behaviors and supporting personalized care, and

this ﬁeld has evolved signiﬁcantly with advances in IoT and AI. StreamTag integrates a ﬂexible data labeling

structure and a modular architecture, enabling data collection, labeling, and secure management of activity

data. The system leverages non-relational databases for scalable data handling, along with secure protocols

to ensure data integrity and privacy. This work examines existing approaches in HAR, including data-driven,

knowledge-based, and hybrid models, and situates StreamTag as a versatile solution that combines ﬂexible

user-controlled labeling with high adaptability for diverse healthcare contexts. Future directions are suggested

for enhancing system functionality and integration with more advanced analytical tools.

1 INTRODUCTION

Healthcare stands as a pivotal area within artiﬁcial in-

telligence (AI), where the potential for innovation and

https://orcid.org/0000-0003-1791-4258

https://orcid.org/0000-0001-6099-0016

https://orcid.org/0000-0003-2583-8638

https://orcid.org/0000-0002-0753-6460

https://orcid.org/0000-0001-5046-0724

https://orcid.org/0000-0002-8117-0647

https://orcid.org/0000-0002-5271-6145

https://orcid.org/0000-0002-7998-553X

https://orcid.org/0000-0003-3749-5986

https://orcid.org/0000-0003-2545-7229

https://orcid.org/0000-0002-5417-3551

https://orcid.org/0000-0002-5286-8026

https://orcid.org/0009-0007-6872-1401

https://orcid.org/0000-0003-1118-7782

impact is immense (Yu et al., 2018; Davenport and

Kalakota, 2019; Alowais et al., 2023). As AI appli-

cations continue to emerge, they address a range of

objectives in health—from patient monitoring and di-

agnostics (Kumar et al., 2023) to personalized treat-

ment plans—transforming the landscape of healthcare

practices (Johnson et al., 2021). However, the ef-

fectiveness of these AI applications heavily relies on

the availability and quality of data, particularly those

datasets where speciﬁc types of events are labeled.

Accurate event labeling is necessary for training AI

models to recognize patterns, predict outcomes, and

ultimately support clinical decision-making (Miller

and Brown, 2017; Duan et al., 2019).

In many healthcare environments, event labeling

has traditionally been managed through manual pro-

cesses, such as written logs or entries in physical note-

books. While these conventional methods have served

Díaz-Jiménez, D., Mata-Mata, F., López, J. L., Pérez-Cordón, L. G., Serrano, J.-M., Mar tínez-Cruz, C., Morcillo-Martínez, J. M., Verdejo-Espinosa, Á., Cuevas-Martínez, J. C., Viciana-Abad,

R., Reche-López, P. J., Pérez-Lorenzo, J. M., Gaitán-Guerrero, J. F. and Espinilla, M.

StreamTag: A Platform for Flexible Tagged Data Management.

DOI: 10.5220/0013278600003938

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 11th International Conference on Information and Communication Technologies for Ageing Well and e-Health (ICT4AWE 2025), pages 233-239

ISBN: 978-989-758-743-6; ISSN: 2184-4984

233

a purpose, they fall short in terms of efﬁciency, accu-

racy, and scalability, particularly when handling large

volumes of data in real-time. For applications that

require prompt and continuous data input, such as

patient activity monitoring or health event tracking,

manual systems often introduce delays and inconsis-

tencies. These limitations underscore the need for

tools capable of automating data labeling and central-

izing information storage, ensuring instant access and

seamless data ﬂow within healthcare systems.

In response to these challenges, Stream Tag was

developed as an mobile application designed to meet

the speciﬁc data management needs of healthcare

providers. This tool offers a solution for real-time

event labeling, addressing the inefﬁciencies and in-

accuracies of traditional data entry methods. Stream

Tag’s core functionality revolves around its ability

to instantly label and transmit events to a central-

ized server, thereby enabling a real-time ﬂow of crit-

ical information. This feature is particularly advanta-

geous in dynamic healthcare settings, such as hospi-

tals and nursing homes, where up-to-date information

can make a signiﬁcant difference in patient care and

clinical decision-making.

The application provides a interface that allows

healthcare staff to quickly label events such as patient

activities, medication administration, or health inci-

dents, eliminating the delays typically associated with

manual data entry. It is designed to work with existing

healthcare systems, providing a user-friendly experi-

ence that requires minimal training. Each event, once

labeled, is securely transmitted to a dedicated server

which makes it readily available to authorized person-

nel across the network.

In addition, the system’s ﬂexibility allows users

to deﬁne a set of predeﬁned speciﬁc labels and activ-

ities relevant to their context, creating a customized

data environment that reﬂects the unique needs of

each facility. In elderly care settings, for instance,

the staff can track residents’ daily activities or health

events with ease, supporting continuous monitoring.

By replacing manual records with a digital, central-

ized platform, Stream Tag reduces the administrative

burden on healthcare providers, allowing them to fo-

cus more on direct patient care.

The structure of the paper is as follows: in Sec-

tion 1, we present the general context, motivation, and

objectives of the research on labeled data manage-

ment in healthcare environments. Section 2 provides

a perspective of current artiﬁcial intelligence use, la-

beling solutions and relevant technologies in the ﬁeld

of data labeling and management. Section 3 describes

the architecture of the StreamTag platform, detailing

its main components and their interactions for secure

and efﬁcient data handling. In Section 4, we explain

the functionality of each view and the conﬁguration

options available to users. Finally, Section 5 presents

the main contributions of StreamTag and outlines fu-

ture challenges and potential improvements.

2 RELATED WORKS

In the ﬁeld of healthcare and well-being, Data Acqui-

sition and Management has become essential in order

to monitor and understand user behavior patterns, es-

pecially in residential and home care contexts. Re-

search in human activity recognition (HAR) (Joban-

putra et al., 2019) has advanced signiﬁcantly in recent

years, enabled by technologies such as the Internet

of Things (IoT) (Laghari et al., 2021; Mouha et al.,

2021; Bhuiyan et al., 2021) and artiﬁcial intelligence

(AI) (Haug and Drazen, 2023; Zhang and Lu, 2021).

HAR focuses on identifying and classifying human

activities using data collected from sensors, such as

accelerometers and gyroscopes, found in portable or

ﬁxed devices (Ramanujam et al., 2021; Dang et al.,

2020), while another approaches make use of RGB

cameras, thermal cameras an so on (Ke et al., 2013;

Shaikh and Chai, 2021; Dang et al., 2020). These

type of systems allows movements and behaviors to

be analyzed, generating data that can be labeled and

subsequently used to build models to detect activities

in real time. This is especially useful in the health-

care context, where accurate activity monitoring can

help prevent incidents, assess patient condition and

support personalized care (Johnson et al., 2021).

The application of AI in HAR has improved the

accuracy and efﬁciency in the identiﬁcation of activ-

ities. However, these methods often require labeled

data in order to train the models properly. The quality

of this data is essential, as the correct functioning of

the models to perform accurate classiﬁcation depends

heavily on properly deﬁned labels. In this sense, plat-

forms that allow ﬂexible and controlled labeling are

relevant, as they ensure the system’s adaptability for

different contexts and users.

The ability to personalize and adapt HAR plat-

forms is a crucial component in the usability of these

technologies in healthcare environments. On the one

hand, there are solutions for tagging data based on

collaborative effort (Chang et al., 2017; Wang et al.,

2012; Huang and Zhao, 2024), although this type of

solution achieves its goal, it requires a large number

of people depending on the data to be tagged. On

the other hand are platforms that allow automatic data

labeling (Ratner et al., 2017; Wu et al., 2021; Dong

et al., 2014), but rely on the use of machine learning

ICT4AWE 2025 - 11th International Conference on Information and Communication Technologies for Ageing Well and e-Health

234

models to label the data. Although depending on the

data to be labeled there may be greater or lesser accu-

racy in the ﬁnal results. In the case of StreamTag,

an interface has been developed that allows select-

ing predetermined activities or creating and tagging

customized activities according to the context, a func-

tionality that facilitates the adaptation of the system to

various situations and allows activities to be recorded

in a semiautomatic-way. The need to store data in

an efﬁcient and scalable manner has led to the adop-

tion of non-relational databases, such as MongoDB,

which allow data to be stored without a rigid struc-

ture, adapting to changes in storage requirements and

the volume of activity data that is continuously gen-

erated. This type of database is for this reason ideal

for systems such as StreamTag, which need to han-

dle heterogeneous information and fast queries with-

out compromising storage ﬂexibility.

3 ARCHITECTURE

This section presents the architecture of StreamTag,

detailing its various components.

Firstly, as illustrated in Figure 1, there are differ-

ent layers that interact with each other, speciﬁcally

the application layer, a reverse proxy, an API, and the

database. The foundational component of this sys-

tem is the application. Both the development of the

application and the other essential elements within

the system are carefully guided by a set of structured

guidelines and a set of considerations. One of these is

the decision to proceed with native development for

the Android platform. This choice primarily stems

from Android’s signiﬁcant market share, which sur-

passes that of other mobile operating systems, further

motivating this selection are practical factors such as

the need for accelerated development cycles suited to

internal projects and the straightforward deployment

capabilities that Android provides. The application

can be distributed directly to devices, circumventing

the lengthy approval processes often required by other

platforms. The activities detected by the system are

securely stored locally on the device. This data re-

mains on the device until the user manually activates

the transmission process, at which point the informa-

tion is sent to the server for further processing or anal-

ysis.

The system’s next component is the reverse proxy,

which is implemented using Nginx. Nginx is a web

server conﬁgured to handle the redirection of appli-

cation requests to the backend API, optimizing re-

quest management. Beyond its redirecting function,

the reverse proxy increase the security by enforcing

TLS encryption standards, speciﬁcally versions 1.2

and 1.3, which facilitate secure data exchange and

ensure conﬁdentiality. Moreover, Nginx is enhanced

with protective rules against potential saturation at-

tacks and unsupported methods in the API that could

introduce vulnerabilities, such as DELETE requests.

The API layer, constructed using the FastAPI

framework in Python, serves as the interface through

which the application interacts with the system’s

database. FastAPI is selected here for its efﬁ-

ciency in handling professional-grade API develop-

ment. Within this layer, a series of endpoints has been

designed to enable data access for the application. To

maintain secure access, the API relies on JSON Web

Tokens (JWT) for authentication. Each time a user

logs in, a JWT is generated and must accompany each

request’s headers, ensuring that access to API end-

points is authenticated and secure.

The data storage layer leverages MongoDB, a

non-relational database. This choice is justiﬁed by

the system’s lack of complex relational data require-

ments, thereby eliminating the need for cross-queries

and making MongoDB an optimal solution. The data

storage setup is centralized, housing user details and

default application data. Should alternative databases

be required, other MongoDB-based instances can be

seamlessly integrated by updating connection creden-

tials. However, it is essential to note that user access

credentials will continue to be centralized in the main

database to maintain consistency and control.

Listing 1: Sample Document Actitity Structure.

1 {

2 " _id ": {

3 " $oid ": "660

cf794bdc4ea b 8 6 2 5 d f d 4 e "

4 } ,

5 " user ": " llop ez ",

6 " da t e _in i t " : {

7 " $n u m b e rLon g " :

"1 7 1 2 1 2 4 6 4 1 4 4 8 "

8 } ,

9 " da t e_e n d " : {

10 " $n u m b e rLon g " :

"1 7 1 2 1 2 5 8 3 9 6 6 5 "

11 } ,

12 " tag ": " Eat i ng " ,

13 " add i t i o n a l _ i n f o " : " Cof fe

Br e a kfa s t "

14 }

• _id: A unique identiﬁer for each document, rep-

resented by an ObjectId. The $oid key designates

the format used by MongoDB.

StreamTag: A Platform for Flexible Tagged Data Management

235

Device

Server

Reverse

Proxy

API DB

HTTPS

TLS 1.3

Figure 1: System components.

Figure 2: API endpoints.

• user: Stores the username associated with the ac-

tivity. In this example, the user is "llopez", link-

ing the document to a speciﬁc individual.

• date_init and date_end: Timestamps indicating

the start and end times of the activity. These are

stored as 64-bit integers ($numberLong) for pre-

cision. These ﬁelds allow for calculating the du-

ration of the activity.

• tag: Represents the type of activity. In this exam-

ple, "Eating" classiﬁes the document as related

to food intake, enabling easy categorization and

ﬁltering.

• additional_info: Provides additional context

for the activity. Here, it contains "Coffe

Breakfast", indicating the speciﬁc meal.

Additionally store user authentication data, in-

cluding personal and security information.

Listing 2: Sample Document User Structure.

1 {

2 " _id ": " 65 0

d55ec652c99 8 8 0 a 8 c 5 4 f 9 " ,

3 " us e rna m e " : " llo pez " ,

4 " fu l l _na m e " : " Jo se Lui s L ope z

Rui z " ,

5 " ema il ": " llo p e z @ u j a en . es ",

6 " has h e d _ p a s s w o r d " : "

$a r g o n 2 id$ v =1 9 $m = 65 5 36 , t =3

, p =4 $ S M n 5 v x e i 9 B 4 D I M S 4 t 3 Z O

... " ,

7 " di s abl e d " : f als e

8 }

It should be noted that the information is en-

crypted at the disk level, rather than at the database

level, as this functionality is reserved for premium

ICT4AWE 2025 - 11th International Conference on Information and Communication Technologies for Ageing Well and e-Health

236

versions of mongoDB.

4 SYSTEM

In this section, information on the use of the Stream

Tag platform is provided. The functionality of the

StreamTag platform is divided into several views,

each designed to allow for easy user interaction and

efﬁcient management of labeled data. These views

provide both default options and advanced conﬁgura-

tions, allowing for a personalized and optimized ex-

perience for the healthcare environment in nursing

homes or any environment requiring activity monitor-

ing and labeling. The main sections of the application

and their functionalities are described below.

4.1 Login

Access to the system starts with the login screen, Fig-

ure 3, where the user is presented with the Stream-

Tag logo and two ﬁelds to enter his credentials (user-

name and password). This interface also includes a

of advanced parameters and loading the last server

conﬁguration. These options allow customizing the

startup process, ensuring that the system adapts to

the speciﬁc conditions of use, such as the security re-

quirements of the data server.

Figure 3: Login with additional options.

The user can access the system by entering their

name and password and pressing the login button.

To optimize the experience, StreamTag automatically

remembers the credentials entered at the last login,

which streamlines future logins. If the user selects the

“Conﬁgure Server” option, a set of advanced options

are displayed that allow the user to deﬁne the database

URL for storing the labels, along with authentication

settings and user-speciﬁc access data. This modular

and conﬁgurable system allows data storage to oper-

ate in an isolated and secure manner, guaranteeing the

integrity and privacy of the information generated and

complying with security standards in the handling of

sensitive data.

4.2 Main View

After logging in, the user accesses the main view Fig-

ure 4, an interface divided into three functional sec-

tions.

Figure 4: Main view.

4.2.1 Tag Upload

The ﬁrst section contains a button that allows upload-

ing the generated labels to the server, along with a

text ﬁeld where the user can specify the name of the

collection in which the data will be saved. This name

can be deﬁned manually or, if the user leaves the ﬁeld

empty, it will be generated automatically based on the

user’s name and the current date. This functionality

ensures that data is stored in an orderly manner and

can be easily retrieved for further analysis, optimiz-

ing data management on the server.

StreamTag: A Platform for Flexible Tagged Data Management

237

4.2.2 Activity Management

In the second section, the user can select and incor-

porate speciﬁc activities from a predeﬁned list. This

list, deﬁned on the server, is adapted according to the

speciﬁc conﬁguration of the database, showing only

the authorized activities for each user or context. The

user can add an individual activity using the “Add”

button or load a set of preconﬁgured activities using

the “Add full day” button. This conﬁguration capabil-

ity is particularly useful in contexts where recurring

or full-day activities are required, thus simplifying in-

teraction and reducing the time required to deﬁne ac-

tivities.

4.2.3 Viewing and Managing Activities

The third section displays the activities that the user

has added in the current session. Each activity appears

with its name and speciﬁc buttons to start, stop or

delete it, allowing precise control over the recording

of each activity. Additionally, each activity includes a

text ﬁeld where the user can add complementary de-

tails, such as the type of medication administered or

any relevant observations. The list of activities can be

reordered with a long press and a drag-and-drop mo-

tion, which facilitates organization and management

according to the user’s priorities or context of use.

4.3 Custom View

The custom view, Figure 5, offers an alternative to

the main view, providing greater ﬂexibility in activity

management. This view differs in two fundamental

aspects:

4.3.1 Dynamic Activity Input Field

Instead of a predeﬁned list, this view features a text

entry ﬁeld that allows the user to enter speciﬁc ac-

tivities not included in the predeﬁned set. This func-

tionality is ideal for users who need to record unusual

or unique activities, expanding the application’s cus-

tomization and adaptability possibilities of the appli-

cation.

4.3.2 Detailed Activity Conﬁguration

Unlike the main view, the custom view does not in-

clude an “Add Full Day” button, allowing the user to

design and manage activities in a more detailed and

individualized way. This is useful in scenarios where

a high level of speciﬁcity in activity recording is re-

quired, as each entry can be manually adjusted to the

user’s needs.

Figure 5: Custom view.

5 CONCLUSIONS

The primary problem addressed in this work is the

growing need to manage labeled data in the health-

care sector, particularly in hospital residency settings,

where structured access to historical and categorized

information is essential for quality care and clinical

decision-making. The lack of an efﬁcient infrastruc-

ture to collect, organize, and use these data in a secure

and scalable way presents a considerable challenge in

these environments.

As a solution, an architecture has been developed,

consisting of a native application, a reverse proxy, a

modern API, and a non-relational database. This plat-

form enables structured management of labeled data,

facilitating access and analysis for the creation of de-

tailed health proﬁles and the identiﬁcation of patterns

in residents’ health status. Integration capability with

IoT devices also optimizes automated data collection,

beneﬁting healthcare staff by providing constant ac-

cess to organized and categorized information, thus

improving operational efﬁciency and data security.

To assess the impact and expand the utility of

the platform, future work should focus on conduct-

ing studies on clinical and operational effectiveness.

These studies would provide objective metrics on its

contribution to healthcare delivery and organization.

Additionally, usability tests aimed at users with min-

imal training would help ensure that the system is

accessible and easy to use, conﬁrming its applicabil-

ity across various clinical settings and levels of staff

ICT4AWE 2025 - 11th International Conference on Information and Communication Technologies for Ageing Well and e-Health

238

training.

ACKNOWLEDGMENTS

This result has been partially supported by

grant PID2021-127275OB-I00 funded by MICI-

U/AEI/10.13039/501100011033 and by “ERDF A

way of making Europe”, grant PDC2023-145863-I00

funded by MICIU/AEI/10.13039/501100011033 and

by “European Union NextGenerationEU/PRTR”,

and grant M.2 PDC_000756 funded by Consejería

de Universidad, Investigación e Innovación and by

ERDF Andalusia Program 2021-2027.

REFERENCES

Alowais, S. A., Alghamdi, S. S., Alsuhebany, N., Alqahtani,

T., Alshaya, A. I., Almohareb, S. N., Aldairem, A.,

Alrashed, M., Bin Saleh, K., Badreldin, H. A., et al.

(2023). Revolutionizing healthcare: the role of arti-

ﬁcial intelligence in clinical practice. BMC medical

education, 23(1):689.

Bhuiyan, M. N., Rahman, M. M., Billah, M. M., and Saha,

D. (2021). Internet of things (iot): A review of its en-

abling technologies in healthcare applications, stan-

dards protocols, security, and market opportunities.

IEEE Internet of Things Journal, 8(13):10474–10498.

Chang, J. C., Amershi, S., and Kamar, E. (2017). Re-

volt: Collaborative crowdsourcing for labeling ma-

chine learning datasets. In Proceedings of the 2017

CHI Conference on Human Factors in Computing Sys-

tems, CHI ’17, page 2334–2346, New York, NY, USA.

Association for Computing Machinery.

Dang, L. M., Min, K., Wang, H., Piran, M. J., Lee, C. H.,

and Moon, H. (2020). Sensor-based and vision-based

human activity recognition: A comprehensive survey.

Pattern Recognition, 108:107561.

Davenport, T. and Kalakota, R. (2019). The potential for

artiﬁcial intelligence in healthcare. Future healthcare

journal, 6(2):94–98.

Dong, X., Gabrilovich, E., Heitz, G., Horn, W., Lao, N.,

Murphy, K., Strohmann, T., Sun, S., and Zhang, W.

(2014). Knowledge vault: a web-scale approach to

probabilistic knowledge fusion. In Proceedings of

the 20th ACM SIGKDD International Conference on

Knowledge Discovery and Data Mining, KDD ’14,

page 601–610, New York, NY, USA. Association for

Computing Machinery.

Duan, Y., Edwards, J., and Dwivedi, Y. K. (2019). Artiﬁcial

intelligence for decision making in the era of big data

- evolution, challenges and research agenda. Int. J. Inf.

Manag., 48:63–71.

Haug, C. J. and Drazen, J. M. (2023). Artiﬁcial intelligence

and machine learning in clinical medicine, 2023. New

England Journal of Medicine, 388(13):1201–1208.

Huang, Q. and Zhao, T. (2024). Data collection and label-

ing techniques for machine learning. arXiv preprint

arXiv:2407.12793.

Jobanputra, C., Bavishi, J., and Doshi, N. (2019). Human

activity recognition: A survey. Procedia Computer

Science, 155:698–703.

Johnson, K. B., Wei, W.-Q., Weeraratne, D., Frisse, M. E.,

Misulis, K., Rhee, K., Zhao, J., and Snowdon, J. L.

(2021). Precision medicine, ai, and the future of per-

sonalized health care. Clinical and translational sci-

ence, 14(1):86–93.

Ke, S.-R., Thuc, H. L. U., Lee, Y.-J., Hwang, J.-N., Yoo, J.-

H., and Choi, K.-H. (2013). A review on video-based

human activity recognition. Computers, 2(2):88–131.

Kumar, Y., Koul, A., Singla, R., and Ijaz, M. F. (2023). Ar-

tiﬁcial intelligence in disease diagnosis: a systematic

literature review, synthesizing framework and future

research agenda. Journal of ambient intelligence and

humanized computing, 14(7):8459–8486.

Laghari, A. A., Wu, K., Laghari, R. A., Ali, M., and Khan,

A. A. (2021). A review and state of art of internet of

things (iot). Archives of Computational Methods in

Engineering, pages 1–19.

Miller, D. and Brown, E. (2017). Artiﬁcial intelligence in

medical practice: The question to the answer? The

American journal of medicine, 131 2:129–133.

Mouha, R. A. R. A. et al. (2021). Internet of things (iot).

Journal of Data Analysis and Information Processing,

9(02):77.

Ramanujam, E., Perumal, T., and Padmavathi, S. (2021).

Human activity recognition with smartphone and

wearable sensors using deep learning techniques: A

review. IEEE Sensors Journal, 21(12):13029–13040.

Ratner, A. J., Bach, S. H., Ehrenberg, H. R., and Ré, C.

(2017). Snorkel: Fast training set generation for infor-

mation extraction. In Proceedings of the 2017 ACM

International Conference on Management of Data,

SIGMOD ’17, page 1683–1686, New York, NY, USA.

Association for Computing Machinery.

Shaikh, M. B. and Chai, D. (2021). Rgb-d data-based action

recognition: a review. Sensors, 21(12):4246.

Wang, J., Kraska, T., Franklin, M. J., and Feng, J. (2012).

Crowder: crowdsourcing entity resolution. Proc.

VLDB Endow., 5(11):1483–1494.

Wu, R., Sakala, P., Li, P., Chu, X., and He, Y.

(2021). Demonstration of panda: a weakly super-

vised entity matching system. Proc. VLDB Endow.,

14(12):2735–2738.

Yu, K.-H., Beam, A. L., and Kohane, I. S. (2018). Artiﬁ-

cial intelligence in healthcare. Nature biomedical en-

gineering, 2(10):719–731.

Zhang, C. and Lu, Y. (2021). Study on artiﬁcial intelligence:

The state of the art and future prospects. Journal of

Industrial Information Integration, 23:100224.

StreamTag: A Platform for Flexible Tagged Data Management

239