Authors:
Muhammad Arslan
and
Christophe Cruz
Affiliation:
Laboratoire d’Informatique de Bourgogne (LIB), 9 Avenue Alain Savary, Dijon 21000, France
Keyword(s):
Artificial Intelligence (AI), Business Intelligence (BI), Dynamic Topic Modeling, Natural Language Processing (NLP), Named-Entity Recognition (NER).
Abstract:
Companies need the data from multiple sources for analysis and to find meaning in deriving valuable business insights. Online news articles are one of the main data sources that present up-to-date business news offered by various companies in the market. Topic modeling, i.e. a key player in the Natural Language Processing (NLP) domain, helps businesses to drive value from news articles. Also, it supports the extraction of business insights from news articles to facilitate the identification of trends (topics) in the market and their evolution over time. These insights can help businesses not only automate routine tasks but also in building new marketing policies, decision-making, and customer support. It is also important to find the linked semantic information (i.e. key persons, organizations, or regions called named entities) involved in generating these topics for the identification of news sources. This paper presents the application of a hybrid approach based on dynamic topic mo
deling and Named-Entity Recognition (NER) for extracting business trends along with the related entities. To show the functionality of the proposed approach, the news articles collected from the websites that published the content related to company interests were from 2017 to 2021 inclusive. The proposed approach can serve as the foundation for future exploratory trend analysis to study the evolution of information not only in the business domain but also applicable in other domains.
(More)