Anastasiia Filatova’s research while affiliated with ITMO University and other places

Highlights What are the main findings? Enhanced Event Detection Accuracy: The introduction of the SemConvTree model, which integrates improved versions of BERTopic, TSB-ARTM, and SBert-Zero-Shot, enables a significant enhancement in the detection accuracy of urban events. The model’s ability to incorporate semantic analysis along with statistical evaluations allows for discerning and categorizing events from social media data more precisely. This results in approximately a 40% increase in the F1-score for event detection compared to previous methods. Semantic Analysis for Event Identification: The SemConvTree model leverages semi-supervised learning techniques to analyze the semantic content of social media posts. This approach helps in understanding the nuanced contexts of urban events, improving the identification process. The model not only recognizes the occurrence of events but also categorizes them into meaningful groups based on their semantic characteristics, which is crucial for effective urban management and planning. What are the implications of the main findings? The increased accuracy in event detection ensures that urban planners and emergency services can respond more effectively to both planned and unplanned urban events. More accurate data leads to better resource allocation, ensuring that services are deployed where they are most needed. This could lead to enhanced safety, improved traffic management, and better crowd control during events, ultimately enhancing urban living conditions. By effectively categorizing urban events based on their semantic characteristics, city administrators can gain insights into the types of events that are prevalent in different areas of the city. This can inform more targeted community engagement strategies, help in the planning of public services and facilities, and ensure that urban policies are closely aligned with the actual dynamics of the city. Additionally, this can aid in long-term urban development strategies by identifying evolving trends and shifts in urban activity patterns. Abstract The digital world is increasingly permeating our reality, creating a significant reflection of the processes and activities occurring in smart cities. Such activities include well-known urban events, celebrations, and those with a very local character. These widespread events have a significant influence on shaping the spirit and atmosphere of urban environments. This work presents SemConvTree, an enhanced semantic version of the ConvTree algorithm. It incorporates the semantic component of data through semi-supervised learning of a topic modeling ensemble, which consists of improved models: BERTopic, TSB-ARTM, and SBert-Zero-Shot. We also present an improved event search algorithm based on both statistical evaluations and semantic analysis of posts. This algorithm allows for fine-tuning the mechanism of discovering the required entities with the specified particularity (such as a particular topic). Experimental studies were conducted within the area of New York City. They showed an improvement in the detection of posts devoted to events (about 40% higher f1-score) due to the accurate handling of events of different scales. These results suggest the long-term potential for creating a semantic platform for the analysis and monitoring of urban events in the future.

The digital world is increasingly invading our reality, which leads to the formation of a significant reflection of the processes and activities taking place in the smart city. Such activities include well-known urban events, celebrations, and those with a very local character. Due to the mass occurrence, events have a comparable influence on the formation of the spirit and the urban atmosphere. This work presents an enhanced semantic version of the ConvTree algorithm - SemConvTree. It allows considering the semantic component of the data obtained by using semi-supervised learning of topic modeling ensemble (consisting of improved models BERTopic, TSB-ARTM, SBert-Zero-Shot). We also present an improved event search algorithm based on both statistical evaluations and semantic analysis of posts. This algorithm allows fine-tuning the mechanism of discovering the required entities with the specified particularity (such as a particular topic). Experimental studies were conducted within the area of New York City. They showed an improvement in the detection of posts devoted to events (about 40% higher f1-score) due to the accurate handling of events of different scales. These results lead in the long term to talk about the potential perspective in creating a semantic platform for the analysis and monitoring of urban events in the future.

Social networks are a source of data that can be useful to researchers in a variety of fields. However, an important limitation to the applicability of this kind of data is the presence of private and sensitive information in it. Therefore, it is important to anonymize it before using it in research to ensure that it is ethical and does not lead to the leak of private information. Most anonymization methods involve removing named entities and identifiers from text. However, in most cases, this level of anonymization is not enough since the authorship of texts can be determined by the style of writing. We propose a new text generation method that takes into account the features of social networks and allows effective anonymization of text style. We evaluate the effectiveness of the proposed model on social media texts, including the Sentiment 140 dataset. The result shows that our model outperforms the state-of-the-art model for authorship obfuscation and semantics preservation.

Accurate geolocation of users’ posts in social networks plays a vital role in a wide area of research devoted to analysing the urban environment based on social media data. It allows us to effectively correlate information about a real urban object with its description in social networks. Information analysis about extensive urban sites is challenging in such studies because publications correlate extremely unevenly with marked geotags relating to such objects. In addition, users of social networks often intentionally indicate incorrect geolocations to increase their publications’ popularity. This article proposes a solution to the problem of accurate geolocation reconstruction for extensive urban sites. It is proposed to use a combined method that enriches the list of initial geolocations of the social network with the help of external sources. The transformer model is then applied to recognise named entities that help to detect mentions of events or locations in user posts. After detection, posts are redefined to the new locations from the extended list according to the semantic similarity calculated between publications and locations.

The rapid growth in the popularity of social media allows researchers in various fields to extract useful knowledge for diverse applications. One of the popular areas of such research is event detection based on social media data analysis. In the scope of this direction, one of the major challenges is to evaluate the quality of the results and to provide a fair comparison of an event detection approach with existing solutions. In this work, we evaluate recently published articles on event detection from the comparability perspective and analyze the ways in which the authors evaluate their algorithms. Next, we review publicly available event detection evaluation datasets and highlight the limitations of their use. As a result, we come to the conclusion that there is a strong need to create a public universal dataset that would allow authors to compare algorithms with each other. Lastly, we describe characteristics of such a dataset and suggest potential ways how it can be created.

This study describes the semi-automated pipeline created for the comprehensive analysis of the urban areas with the extremely low and extremely high popularity levels. It includes the geo-frequency analysis of the Russian-language Instagram publications for the St. Petersburg area and selection of areas with the extreme values of the popularity level according to the number of publications in them. Semantic analysis of the urban areas with an extremely low number of publications includes comparing of algorithms for descriptions extraction and classification for these areas and results of such descriptions extraction and classification using TF-IDF vectorization technique and most valuable words extraction. Semantic analysis of areas with an extremely high number of publications includes the structure description of such areas, comparing of algorithms for advertisement publications extraction, results of the advertisement extraction using BigARTM model and further development and implementation of the algorithm for extracting events related to the the points of attraction in extremely popular urban areas, which is based on the strong time binding hypothesis and the idea of similarity queries using combination of LDA models for revealing semantic structure and algorithm based on frequency analysis. Developed algorithm was tested to extract events in the urban area of St. Petersburg where Ice Palace is placed and showed interpretable results and allow us to correctly extract 89 events out of 102 which occurred in this area in 2019. Finally, SeSAM pipeline for comprehensive urban analysis was created that combined the described algorithms.