Topic modelling of diverse events in the 21st century public library: A case analysis of New York Public Library

Research background

Alongside social changes that have occurred in the 21st century, information technology has progressed, and people's behaviour and consciousness have changed in light of new technologies. For example, people who use social networking services have more opportunities to encounter radical ideas, and the influence of COVID-19 has caused division through the ubiquity of conspiracy theories. Public libraries, which are open to everyone for free, are becoming more important as places to resolve these divisions (Klinenberg, 2018; Wiegand, 2015). In particular, in the United States, disparities have arisen related to racial conflict. Policies to cope with social division are currently being advanced. For example, the New York State Library has reviewed library policies approximately once every five years in response to a serious deterioration in social conditions (New York State Library, 2021; Yamagishi et al., 2022).

Barchas-Lichtenstein et al. (2019) illustrated that public libraries prioritised events to contribute to individuals and communities. In particular, recently, events focusing on social inclusion have attracted attention. It has been revealed that New York State is promoting various events to deal with this situation. The New York Public Library held about 100,000 per year in 2019 (The New York Public Library, 2020, p. 48). A typical event at the New York Public Library is the summer reading programme (The New York Public Library, 2021). This event, which has been held for a long time, supports children's reading and learning during the summer vacation. Children who cannot afford their own materials will be provided with study kits for this program. Furthermore, due to the COVID-19 pandemic, many library events using online systems (e.g., Zoom) have been planned and held. Library events are one of the most direct approaches to users among the services of public libraries. In other words, holding events is a service that can directly meet the needs of users beyond libraries. Specific examples are IT support for the elderly and storytelling events for parents and children.

As the four-space model shows (Jochumsen et al., 2010, 2012), the functions of public libraries are diverse, but nothing has been clarified about the functions including online events. Igarashi et al. (2021) analysed events of the Helsinki Central Library Oodi to clarify the characteristics of the events, including online holding. However, the United States has more serious social problems, and it can be said that the situation is different from that of Northern European countries.

Previous research

Mathiasson and Jochumsen (2019) investigated posted events of public libraries on Facebook for creating a new approach of researching public library programs. They collected many events through Facebook posts, but their purpose is a creation of a new research method, and they did not analyse events in detail to understand the topics and characteristics of events. Recently, much research on events has been done and revealed interesting and significant insights regarding public library events. However, they focused on each specific event and case, such as reading clubs (Luyt et al., 2011), language programs for immigrants (Johnston, 2016; Vårheim, 2011), health information programs (Luo, 2018), information literacy programs (Hedemark and Lindberg, 2018; Lopatovska et al., 2018), and services for children with disabilities (Adkins and Bushman, 2015). But, we have not captured the entire picture of public library events yet, since they have so many events every year, as is the case of NYPL. In order to solve this research problem, Igarashi et al., (2021) comprehensively analysed the events at Helsinki Central Library Oodi, a prominent library in the Nordic countries, and identified traits of Finnish library events from broader perspectives. However, it has not been described how these events relate to traditional public library services, nor has the relevance of the library events been explained. Besides, there is no such research on public libraries in the United States.

Research purpose

The purpose of this research is to clarify the role that public libraries in the 21st century play in their communities. This is done through an analysis of the events that public libraries provide to citizens. New York Public Library (NYPL) was selected as a representative case. Public libraries in the 21st century such as NYPL have changed from traditional collection-based libraries to user-centred libraries. Therefore, the role of public libraries in the 21st century will be illustrated by analysing the entire picture of user-centred events at NYPL using a topic modelling method.

Three research questions are addressed in this study:

• RQ1. What kinds of events are held by NYPL?
• RQ2. What are the topics of the events held by 21st century public libraries?
• RQ3. What is the relationship among the event topics as exemplified by NYPL?

NYPL has a large library network and is one of the most outstanding libraries in the world. RQ2 seeks to understand the topics of events held by 21st century libraries by generalizing beyond NYPL. In RQ3, we consider how different or similar each topic is, and we create topic groups.

Research method

Case selection and unit of analysis

The case selected for this study was the New York Public Library (NYPL) in the United States. Event information was collected from the library website every day from November 10, 2020, to November 30, 2021. Due to an error, events on July 12, 2021, were not collected. The total number of events recorded was 21,411. The information collected about each event included event title, category, date, time, location, description, and audience. Latent Dirichlet Allocation (LDA) was used to analyse the words in each event description. This analysis did not extract data from the full event description when descriptions exceeded 500 words, but the word analysis process was sufficient to gain an overview of each event. Though some events may have been cancelled for various reasons, what was important was to analyse the events that were planned. Therefore, all planned events were included in the analysis.

Analysis method - LDA

Overview of LDA and previous research in the field of library and information science

Latent Dirichlet Allocation (LDA) is a typical method of topic model for estimating the topic of text groups (Blei et al., 2003; Griffiths and Steyvers, 2004). In LDA, the topic of a document is estimated by dividing a sentence into words and calculating the Dirichlet distribution from the co-occurrence of words. In the field of Library and Information Science, LDA has been conducted for abstracts of academic papers, and most of them identified transitions of the topic in the academic papers.

As described above, by using LDA, the topics contained in the text can be estimated. In addition, the similarity between the topics can be measured. In other words, by applying LDA to the description of an event held in NYPL, the topics of that event can be estimated and the relationships between topics can be analysed. This allows for description of an overall picture of NYPL events in detail. In this study, by extracting topics from the content described in the event title and introductory text and calculating the closeness of the topic, the content of the event held in the modern public library can be explained in more detail.

Analysis process

In this study, LDA was implemented using NLTK and Gensim, which are Python libraries. First, morphological analysis of the descriptions was performed using NLTK. At this point, the default stop words and symbols/signs in NLTK were excluded. In addition, those symbols and words that are cardinal numbers, prepositions or subordinating conjunctions, or personal pronouns were also excluded. Then, stemming was performed using Porter's stemming algorithm to create the Bag of Words (BoW). Next, LDA was applied to the created BoW. At this time, in order to improve the accuracy of LDA, a calculation was performed by deleting the words that appeared in fewer than 3 of all events and the words that appeared in 90% or more of all events. LDA was conducted by specifying the number of topics from between 10 and 30. The authors analysed and discussed the details of these results. It was decided that 24 topics was most appropriate because this number of topics best explained NYPL events. Regarding the determination of the number of topics, the optimum number of topics may be calculated quantitatively by calculating Perplexity and Coherence. At this time, however, there is no previous research on how best to determine these values, so the authors instead decided on the number of topics qualitatively.

Due to the nature of New York Public Library events, and because some events were held in various languages, there were cases where event information included foreign language characters, such as Chinese. Since LDA focuses on words, it may not properly analyse languages other than English. However, since holding events in multiple languages is an important element of library services, it was decided to regard foreign-language events as a topic of its own, rather than parse the meaning of foreign-language descriptions. The descriptions of this set of events therefore were not translated.

The value of the relevance metric was set to 0.6, which is generally considered appropriate for LDA analysis. This value may be anywhere from 0 to 1, where the smaller the value, the higher the probability that the word appears only in the selected topic, and the larger the value, the more likely it appears generally across multiple topics. In this study, 0.6 was used as the standard when preparing the list of words in each topic. Additionally, and aside from the relevance metric used, the topics were discussed after examining the words that appear in each topic. A concern with LDA in this study is that a small number of characteristic events may be overlooked because topics are extracted by the distribution of word frequency. However, it is possible to analyse very large datasets that are practically difficult to qualitatively analyse, therefore providing a more accurate overview of events in public libraries. LDA is a quantitative method, but we checked the event titles and so on, and we decided that 24 topics were enough to give a complete picture of the event.

Results

Labelling of topics

The results of LDA are shown in Table 1. It gives the top five words and the characteristic words in each topic. Topics were labelled by examining the words that appeared within each topic and the event information related to each topic.

Topic 1 includes events that provide employment support or literacy support. Topic 2 is a topic related to the use of online services of the library because it includes words such as ‘password’ and ‘e-book’. Topic 3 is best described as citizen learning from words such as ‘citizenship’ and ‘studi (es)’. Topic 4 can be presumed to include IT support from words such as ‘softwar(e).’ Topic 5 features events in foreign languages because languages other than English are frequently used. It is conducted in English with foreign languages such as Spanish to meet the needs of immigrants.

The word ‘career’ characterises Topic 6. This topic includes events that provide life support such as work and school for citizens. Similarly, Topic 7 is characterised by words such as ‘financi(al)’ and ‘money,’ and it can be presumed to be a topic that includes events related to financial support for the poor. Topic 8 does not show the characteristic words like Topics 6 and 7, but it contains words that are found frequently in online events. Topic 9 is a topic about creative activities that have recently become more important than traditional library services such as collection lending and reading. There are other topics related to creative activities, but the words that make up each are slightly different, and Topic 9 features activities centred on knitting. In Topic 10, words such as ‘mask’ and ‘requir(e)’ come to the top. These words imply that the event is held face-to-face. In addition, this topic includes events such as board games such as chess using “library as place”.

Topic 11, with words such as ‘research’ and ‘consult’ appearing at the top of the list, is an event about supporting research. This includes events that support more advanced IT skills such as ‘python’. Topic 12 is likely to include a wide range of book-based events, such as ‘book’, ‘read’ and ‘discuss’. Topic 13 includes events related to the New York Public Library's exhibitions, as indicated by the words ‘hall’ and ‘tour’. Here, the words ‘architectur(e)’ and the names of the buildings of the New York Public Library e.g. ‘astor’, ‘schwarzman’, are included, although less frequently. Topic 14 is characterised by words such as ‘children’ and ‘parent’, and is an event for children, such as ‘storytelling’.

Topic 15 is an event regarding literature about American history and African American, since the words ‘poetri(es)’, ‘black’, and ‘history’ are incorporated. Topic 16 is an event similar to a creative activity, characterised by the word ‘lego’. Topic 17 is considered to be an educational event for young people, ‘teen’ and ‘comput(er)’ are typical words. Topic 18 is dominated by words such as ‘trivia’ and ‘anim(e)’. However, the inclusion of Chinese characters and the inclusion of a library guide in Chinese suggest that this is an event about Asian multiculturalism, which is a growing population in the United States.

Topic 19 is an event about online reading that includes the words ‘discuss’ and ‘onlin(e)’. It also contains specific words such as ‘manga’ and ‘naruto’, suggesting that it is a Young Adults’ event. Topic 20 is characterised by words such as ‘Japanes(e)’, ‘journal’, and ‘caribbean’. This topic has cultural diversities. Topic 21, with words such as ‘colleg(e)’ and ‘scholarship’ on the top of the list, is considered to be educational support that is not homework support. For example, this could include how to write an application for university. Topic 22 is the topic of health and medical care, with words such as ‘mental’ and ‘healthcar(e)’. Topic 23 is about language teaching, such as ‘English’, ‘practic(e)’ and ‘learner’, and includes events which are considered to be aimed at non-native speakers of English. Topic 24 contains a wide range of events such as ‘music’ and ‘painting’. And, it also contains events in Chinese. That means this topic is culturally diversified.

Grouping topics

There are similar topics among the 24 identified. Similar topics were therefore grouped to show the relationship between topics more clearly. The Figure 1 below is a visualisation of the results of LDA using PCoA. PCoA is the most commonly used algorithm for visualising LDA results. When illustrating the results of LDA, some information is omitted to simplify the visual, so it is not always the most accurate. However, this result sufficiently expresses the grouping.

The twenty-four topics were grouped into five groups. The five groups are 1) Collection-based reading and learning, 2) Exchange through creative activities, 3) Multiculturalism and respect for history, 4) Support for all, and 5) Events held online during the COVID-19 pandemic.

Collection-based reading and learning (Topics 3, 12, 14, 19)

The first group is traditional collection-based events held in the library. This group consists of topics 3, 12, 14, 19 and includes reading clubs and events related to citizens' lifelong learning. The target audience of the events range from children to adults. Reading aloud, especially for children, accounts for a relatively large proportion (Topic 14). There were not many events other than storytelling for children.

Exchange through creative activities (Topic 9, 10, 16, 18, 24)

The second group includes events that promote interaction through creative activities. This group does not use the collection and includes events for creative activities such as games and knitting. Each topic has its own distinctive activities, with Topic 9 being knitting, Topic 10 being board games such as chess, and Topic 16 being events using educational toys such as Legos. Although Topics 18 and 24 had no clear characteristics, it was inferred from words such as anime, movie, music, and paint that they mainly focused on creative activities.

Multiculturalism and respect for history (Topics 5, 13, 15, 20, 23)

The third group incorporates events that discuss multiculturalism and American history. In the United States, where people with diverse roots live, multicultural understanding is an important issue, so many such events are held. Topic 5 is an event that is held in both foreign languages and English, and Topics 13 and 15 include events related to the history of NYPL itself and the history of the United States. In particular, Topic 15 includes events related to learning about African Americans and literary works. Topic 23 includes events that provide language instruction. It is important for non-native English speakers, immigrants, and others to understand American history and culture.

Support for all (Topics 1, 4, 6, 7, 11, 17, 21, 22)

The fourth group assembles events designed to support all people, specifically vulnerable people, and they are indispensable for addressing social disparities and divisions. In particular, language was important in both life and culture, and it is essential for libraries to be aware of this when holding events. The New York Public Library offers a variety of support, and when taken together, it can be seen that these types of events reach everyone. Topics 1 and 6 relate to job hunting and career support and counselling sessions. Topics 17 and 21 are mainly for student support and educational events. In addition, some events support people who are said to be socially vulnerable, such as IT support (Topic 4), which is thought to be mainly aimed at the elderly, and financial support counselling meetings (Topic 7). Provision of health and medical information is also included in this group (Topic 22). This group also includes learning events to prevent overdose, approaching social issues such as the opioid epidemic.

In addition, research support utilising the specialised research collection of New York Public Library also emerged as a topic (Topic 11). Though the target audiences differ, each topic in this category supports and improves the lives of citizens in some way.

Events held online during the COVID-19 pandemic (Topics 2, 8)

The fifth topical group differs from the other groups. Topic 2 includes an event that guides participants about how to use online services that have become more frequent after COVID-19. Topic 8 summarises words that are included in events held online. Since these types of events account for a relatively large proportion of events recorded, a distinct category was formed to account for them.

The event information collection period for this study lasted one year, beginning November 2020. During this period, COVID-19 became a global pandemic, and many activities were restricted. At the beginning of the pandemic, services in public libraries were severely restricted, but since then, online events have become more prevalent. The fifth topical group indicates that online events reflect the current social situation. For this reason, online events were categorised as a distinct event group.

5.3 Relationships between topic groups

The five topic groups summarise in a qualitative way the types of events held at the library. The relationships between the five groups are roughly as shown in Figure 1 above. Specifically, Group 1 Collection-based reading and learning and Group 2 Exchange through creative activities are very close to each other, both of which are carried out to fulfil the cultural creative function of the library. In particular, Group 1 is a basic event held at any library and was the most prevalent online event under the influence of COVID-19. Group 3, Multiculturalism and respect for history, are important roles of public libraries, and they have become particularly important in recent years, and they represent a group of events designed to help overcome social divisions. Some of the topics in Group 3 include learning through literary works, and they have similar properties to Group 1 in that they utilise the collection. Group 4 Support for all is the group with the most topics. It overlaps with Group 3 Multiculturalism and respect for history in, for example, language instruction (Topic 23), and it is necessary for immigrants to become familiar with American society, but it is also support for immigrants who are vulnerable to social vulnerabilities. Group 5 Events held online during the COVID-19 pandemic overlaps with all groups. During COVID-19, events on any topic were required to be held online.

Discussion and conclusion

The purpose of this study was to clarify the roles played by the public library in the 21st century by analysing what kind of events were offered to the citizens of NYPL. If the number of topics is large, the words that make up each topic will all be similar, and as a result, the topics will not be informative. Therefore, in this research, LDA was tried with the number of topics ranging from 10 to 30. Each number was considered until the set of 24 topics was found to be the most informative and meaningful.

At the stage of deciding the number of topics for the implementation of LDA, about 20,000 events were examined and it was found that NYPL has a great variety of events. This addresses Research Question 1. For instance, the event called “Overdose Prevention with Gracie Square Hospital” provides a virtual presentation by addiction experts free of charge for any citizen. The description of this event is as follows:

During this virtual presentation you will hear an addiction expert from Gracie Square Hospital discuss substance use, treatment options, the trajectory of the opioid epidemic, and how COVID-19 has changed the landscape. You can learn about addiction and recovery before and after COVID. Attendees will also receive training in how to use a naloxone kit and have the option to have one mailed to their home.

This event has not been shown in a specific topic since it is for a very limited/minority group, but it is critically important for public libraries to have such an event. The importance of having this event in a public library is that anyone can attend for free. The library has always had the idea of accepting all people, but the removal of space constraints by holding online will lead to the expansion of services including future events for libraries in the 21st century. On the other hand, with the outbreak of COVID-19, even in the uncertain situation of seeking a response in accordance with the policies of the government and WHO, the public library requested from the citizens that these events were planned and held. At the same time, it is considered to suggest that it is a resilient public facility that can take various approaches to society.

Events in the 21st century public library could be represented as 24 topics and 5 topic groups, as noted above. This addresses Research Question 2. The results of LDA topic analysis indicate that there are diverse events to approach community needs. For instance, the event called "Virtual Book Discussion: Kindred by Octavia Butler" tells a brave story where an African American woman overcame/conquered racism. At the same time, the event incorporates many significant topics relevant to current societies, since the event makes people understand and respect diverse communities and multicultural histories. One critical event may impact multiple domains of the world. For instance, it can be seen that language instruction (Topic 23) has directly impacted everyday communications among diverse ethnic groups as well as indirectly bridged social divisions, created new cultures, and united society.

Each of the five topic groups has overlaps or is strongly related. This addresses Research Question 3. Group 1 and Group 4 have in common that they are traditional events. For example, the Summer Reading programme includes both Group 1 and Group 4 and is a typical event that has been held for a long time. At the same time, this programme provides opportunities to study for vulnerable children the same as ordinary people. Group 1 and Group 2 include events that lead to the creation of culture. There are many events for children, and reading and playing are indispensable topics for children's growth. In particular, events that utilise places of interaction rather than collections such as Group 2 play an important role in modern times. Similarly, respecting diversity within Group 3 and understanding the cultures of the United States and other countries is also a topic that will lead to overcoming social issues in the 21st century. During COVID-19, the new online delivery mode has increased for events in both groups.

The COVID-19 pandemic has transformed our lives into more restrictive ones. However, public libraries have begun to further develop online services as an important institution that supports society as a whole. This makes it easier for some people to access library services. Even after COVID-19, it will be desirable to provide various library services that take advantage of this option. Until now, library events had the restriction of having to be physically present, but now it will be possible to participate in library events from all over the world while online. The various events of public libraries in the 21st century will be expected to function as new places to connect and interact with people from all over the world.

Finally, as a future direction for this research, it is hoped that events can be studied using the voices of those people who actually participated in them. In particular, it will be important to understand how people viewed events held online compared to conventional events. It is important to consider the role of public libraries in the 21st century in facilitating interactions among people from around the world and between those who previously had no access to the NYPL events before they were held online.

Acknowledgements

This work was supported by JSPS KAKENHI Grant Numbers JP20H04479, JP21J10661. We thank Prof. Kei Wakabayashi at the University of Tsukuba for assistance with the LDA method in this research.

About the authors

Motoko Yamagishi is a master’s student at the Graduate School of Comprehensive Human Sciences, University of Tsukuba, 1-2 Kasuga, Tsukuba, Ibaraki 305-8550, Japan. She has written extensively on topics related to cultural policy, public libraries and library spaces and architecture. She can be reached at: s1711578@klis.tsukuba.ac.jp
Masanori Koizumi is an associate professor at the Faculty of Library, Information and Media Science at the University of Tsukuba, 1-2 Kasuga, Tsukuba, Ibaraki 305-8550, Japan. He has written extensively on topics related to library strategy, information specialists and the public sphere. Institution of highest degree, PhD. Keio University, Library and Information Science in 2013. He can be reached at: koizumi@slis.tsukuba.ac.jp
Michael M. Widdersheim is assistant professor in the School of Library and Information Management at Emporia State University, United States. His research interests include the study of communicative structures related to public library development. His contributions include a new research design, historical case study; new models of the public sphere, the tessellation model and decision cycle; and a new explanatory theory of library development. He can be reached at mwidders@emporia.edu
Tomoya Igarashi is PhD candidate in Graduate School of Comprehensive Human Sciences, University of Tsukuba, 1-2 Kasuga, Tsukuba, Ibaraki, 305-8550, Japan. His current research interests are evaluation of public libraries and public libraries as space for overcoming social division. He can be contacted at igarashi@klis.tsukuba.ac.jp

How to cite this paper