Abstract:
"Causal Inference combines methodologies of evaluating the causal impact of treatment
along with outcomes influenced by confounders where they need to be measured. This
concept is an important research area topic in several areas such as computer science,
economics, statistics etc. Recently, causal effects estimation on observations’ data is
growing as an upcoming research field since it has large size data sets in many domains
especially measuring confounders from text become an interesting research topic. For
example, how a product review majorly impacts it sales increment? As the review
content in majorly in a text format this will be a bit much challenging to measure causal
effects of text since it has a high dimensional structure. Various causal impact
estimating methods for text have arisen in the wake of the fast evolution of machine
learning. But still, it faces the selective biased problems where it produces invalid
causal effects estimations. As mentioned in the example the major reviews systems
analysis the impact of sales increments based on numerical ratings where it ignores
important facts hidden from review content.
To overcome above mentioned issues, by combining causality with machine learning
are sufficient for forecasting and understanding the cause and effect of the results. This
research is utilized systematic approach to identify which causal machine learning
methods are used to enhance the causal inference problems related to textual reviews
system to validating how it is impact on the products sales/demand using amazon
platform as an example. This research also focuses on the combine use of advanced
causal models like Double Machine Learning with machine learning approaches to
conduct causality analysis using amazon online review data. "
