Page 69 - Flytxt
P. 69
step, aspect terms are extracted using The emotions expressed by the
CRFs with the large set of features customers may not be always having
(e.g. word itself, context words, part- a direct tone and can be very complex
of-speech tags, word frequency, etc.). in nature, like irony or sarcasm. This
In second step, each aspect terms’ complicates the process of identifying a
sentiment/opinion is identified using clear sentiment. Though, advancement
CRFs. in technology will overcome these
The supervised model is applied on three challenges.
domains – i) Laptop, ii) Restaurant and The bottom line is that sentiment
iii) Amazon product reviews (e.g. coffee analysis is all about converting data into
machine, cutlery, microwave, toaster, meaningful and actionable information
etc.) The goal of this project is to build in hands of companies. No matter how
a generic model which can be applied complex it is, its benefits are massive
to any domain to discover relevant
aspect terms and sentiments. We are
also building a hybrid model using Flytxt has developed a domain
unsupervised and supervised approaches agonistic supervised Machine
towards each of the discovered aspect. Learning approach for Aspect
You can access the white paper here. Based Sentiment Analysis
Conclusion (ABSA). Conditional Random
With the ever-expanding data sets in Fields (CRFs) are being used to
deploy the supervised model
today’s world, tools like sentiment for extracting aspect terms
analysis open many gateways for
analyzing this data to derive meaningful and identifying sentiments
insights and gain a greater business on different aspects
value. However, there are many from customer reviews
challenges in the path of implementing and comments.
effective sentiment analysis.
INSIGHTZ - VOLUME 03, 2018 69
