Ana Oliveira Alves

Place Enrichment by Mining the Web

In this paper, we address the assignment of semantics to places. The approach followed consists on leveraging from web online resources that are directly or indirectly related to places as well as from the integration with lexical and semantic frameworks such as Wordnet or Semantic Web ontologies. We argue for the wide applicability and validity of this approach to the area of Ubiquitous Computing, particularly for Context Awareness. We present our system, KUSCO, which searches for semantics associations to a given Point Of Interest (POI). Particular focus is provided to the experimentation and validation aspects.

ASAP: Automatic Semantic Alignment for Phrases

In this paper we describe the ASAP system (Automatic Semantic Alignment for Phrases)1 which participated on the Task 1 at the SemEval-2014 contest (Marelli et al., 2014a). Our assumption is that STS (Semantic Text Similarity) follows a function considering lexical, syntactic, semantic and distributional features. We demonstrate the learning process of this function without any deep preprocessing achieving an acceptable correlation.

Tagging space from information extraction and popularity of points of interest

This paper is about automatic tagging of urban areas considering its constituent Points of Interest. First, our approach geographically clusters places that offer similar services in the same generic category (e.g. Food & Dining; Entertainment & Arts) in order to identify specialized zones in the urban context. Then, these places are analysed and tagged from available information sources on the Web using KUSCO [2,3] and finally the most relevant tags are chosen considering not only the place itself but also its popularity in social networks. We present some experiments in the greater metropolitan area of Boston.

Visualizing urban mobility

The goal of this research is understanding urban mobility through the visualization of the use of public transport systems. We focus on the visualization of anomalies regarding the number of passengers. To find patterns of use we analyze the raw data, which contains people counts for every bus stop in Coimbra. For each stop, and for each day of the week, we calculate the average number of passengers and its standard deviation for each 30 minute interval. This allows us to identify situations that deviate from the norm.

Making sense of location context

This paper presents a generic model for Semantic Enrichment of Location Context where the main attributes of places are described by tags. These tags are automatically extracted by applying natural language processing and information extraction techniques that have been thoroughly applied and tested using the World Wide Web as the primary source. Here, we are particularly focused on extracting information that allows an external system to distinguish one place from other places that are spatially or conceptually close. This is because the meaning of a place is a function of its most salient features, present in the textual descriptions found in online resources about that place. In the situation under investigation, places correspond to Points-of-Interest (POIs), as these are abundant on the Web. The applicability of such model is demonstrated through its implementation over real collected POIs. As an illustrative output of the system, a set of examples is also presented.

Place in perspective: extracting online information about points of interest

During the last few years, the amount of online descriptive information about places has reached reasonable dimensions for many cities in the world. Being such information mostly in Natural Language text, Information Extraction techniques are needed for obtaining the meaning of places that underlies these massive amounts of commonsense and user made sources. In this article, we show how we automatically label places using Information Extraction techniques applied to online resources such as Wikipedia, Yellow Pages and Yahoo!.

Comparing the Performance of Different NLP Toolkits in Formal and Social Media Text

Nowadays, there are many toolkits available for performing common natural language processing tasks, which enable the development of more powerful applications without having to start from scratch. In fact, for English, there is no need to develop tools such as tokenizers, part-of-speech (POS) taggers, chunkers or named entity recognizers (NER). The current challenge is to select which one to use, out of the range of available tools. This choice may depend on several aspects, including the kind and source of text, where the level, formal or informal, may influence the performance of such tools. In this paper, we assess a range of natural language processing toolkits with their default configuration, while performing a set of standard tasks (e.g. tokenization, POS tagging, chunking and NER), in popular datasets that cover newspaper and social network text. The obtained results are analyzed and, while we could not decide on a single toolkit, this exercise was very helpful to narrow our choice.

Mobile Crowd Sensing for Solidarity Campaigns

We present an ongoing project (This work is partially supported by the InfoCrowds project-PTDC/ECM-TRA/1898/2012 This work is supported by CISUC, via national funding by the FCT - Fundacao para a Ciencia e Tecnologia.) which has two separate strands, one refers to the technological study about the applicability of high performance and high availability technologies in Web Services and the other is directed to a practical application of these technologies to solidarity campaigns in collecting goods. The focus of this paper is in the first one, a technological study where several frameworks for building Web Services, databases of different types and libraries to assist in obtaining product codes (barcodes) and data are analyzed, this includes a study of performance, availability and reliability, as well as appraisals for each one. Besides this, we introduce an experimental setup and results obtained so far in a third sector institution, Caritas Diocesana of Coimbra (http://www.caritas.pt/site/nacional/ Portuguese Website (last visited in March 2015)), a non-profit organization part of Caritas (http://www.caritas.eu/ (last visited in March 2015)). As main contribution, we propose a distributed architecture for Mobile Crowd Sensing able not only to allow real time inventory through simultaneous campaigns but also it gives feedback to volunteers in order to instantly acquire information about which categories of goods are more needed.

Acquiring semantic context for events from online resources

During the last few years, the amount of online descriptive information about places and their dynamics has reached reasonable dimension for many cities in the world. Such enriched information can now support semantic analysis of space, particularly in which respects to what exists there and what happens there. We present a methodology to automatically label places according to events that happen there. To achieve this we use Information Extraction techniques applied to online Web 2.0 resources such as Zvents and Boston Calendar. Wikipedia is also used as a resource to semantically enrich the tag vectors initially extracted. We describe the process by which these semantic vectors are obtained, present results of experimental analysis, and validated these with Amazon Mechanical Turk and a set of algorithms. To conclude, we discuss the strengths and weaknesses of the methodology.

Towards the Improvement of a Topic Model with Semantic Knowledge

Although typically used in classic topic models, surface words cannot represent meaning on their own. Consequently, redundancy is common in those topics, which may, for instance, include synonyms. To face this problem, we present SemLDA, an extended topic model that incorporates semantics from an external lexical-semantic knowledge base. SemLDA is introduced and explained in detail, pointing out where semantics is included both in the pre-pocessing and generative phase of topic distributions. As a result, instead of topics as distributions over words, we obtain distributions over concepts, each represented by a set of synonymous words. In order to evaluate SemLDA, we applied preliminary qualitative tests automatically against a state-of-the-art classical topic model. The results were promising and confirm our intuition towards the benefits of incorporating general semantics in a topic model.