Project Proposal
Problem to be solved
The problem that this project aims to solve is the problem that business owners and/or brands cannot read through all of their audience’s opinions on products, for example, as there is an extremely high velocity of data on social media. If businesses and/or brands have the ability to better monitor and analyse public opinion on topics/phrases it will allow them to better market items and build their brand effectively that benefits the customer and the marketer.
My project aims to allow marketers to filter through large datasets of public opinion on their brand/product and gain an understanding on public opinion on their product/brand via sentiment analysis. The data will be derived from twitter, as it is a platform where users in large numbers tend to express their opinion, thus making it a gold mine for data on public outlook.
References to existing relevant work:
HubSpot’s ServiceHub
https://www.hubspot.com/
My project will differ to HubSpot’s Service Hub feature in a number of ways. While the intent of using sentiment analysis to improve marketing and customer relation is the same goal being strived for, HubSpot has more of a focus on breaking down qualitative survey responses, whereas my project is more focused on Twitter. The reason for this is that I believe due to the currently massive and continuously growing number of people on social media, this makes it possibly the greatest dataset when it comes to extracting public or customer opinion on a topic or product.
One possible future implementation that I could emulate from their service is their detailed outlay of graphs and charts that show levels of customer satisfaction.
Lexalytics
While Lexalytics is a powerful tool that uses natural language processing, they are more focused on parsing complex text documents, and therefore are more suited to businesses that have a focus on processing large volumes of text data. Therefore, the purpose of their product differs from mine in that theirs is not a marketing improvement tool, but a text parsing tool. They also produce structured overviews of the outcomes of their analysis.
Social Mention
http://www.socialmention.com
Social Mention has very well laid out insights into the analytics formed upon a search, and the format of input when searching for a term is very intuitive and could be something I use as inspiration when implementing my program.
¬Books:
Sentiment Analysis in Social Networks
By Federico Alberto Pozzi, Elisabetta Fersini, Enza Messina, Bing Liu:
This book is a deeper dive on sentiment analysis as an automated tool used to extract subjective data for the purpose of creating useful insights. The book explores a number of computing disciplines such as big data and natural language processing, core concepts in the topic of sentiment analysis. The book is relevant as it gives guidance on using data mining tools to extract sentiment for application.
Mining the Social Web: Data Mining Facebook, Twitter, LinkedIn, Instagram, GitHub and more
By Matthew A. Russell, Mikhail Klassen:
A book that explores mining the rich data encapsulated across a number of social sites such as Facebook, Instagram and Twitter. It teaches readers how to garner insights from these sites. It contains a few Python examples and exercises to solve.
Project objectives
The main objective of this project is to create a piece of software that will allow the user (business owners and brands) to gain a comprehensive understanding of public opinion on a particular brand/product. It will do this by allowing the user to search for a topic and it should return a concise summary of how positive tweets are generally vs how negative they are in relation to the topic.
There are sub-objectives for this project that should therefore be met so that the main objective is fully achieved, this includes:
1. I will first gain further understanding on the types of sentiment that define how people view brands/products, so the software can provide an accurate summary of public view on the chosen topic.
2. Following this I will review API and twitter datasets available.
3. Next, I will decide on the most appropriate source of Twitter data collection for my project.
4. Then I will review the possible Python libraries that allow for twitter API interaction and analysis of tweets/datasets.
5. Once the previous sub-objective is done, I will decide on the most useful library/libraries for my project.
6. The next stage of my project will be to implement the code for my program.
7. Then once it has been implemented, the code will be tested and reviewed, with the tests documented clearly.
Beneficiaries
The people who will benefit from the project are:
– End users:
o Business owners
o Influencers
o Brands
– Further developers
– Myself
The way they will benefit is:
End users:
– They can use this tool to predict products that will perform successfully before committing to creating/investing in them.
– Marketing strategies/campaigns can be adjusted and optimized to better fit the user’s target audience.
– Can be used to measure how well the user’s existing products/campaigns are performing.
– Can be used to improve customer service and keep customers loyal to the brand by responding to both positive and negative reviews in a timely manner.
Further developers:
– The project, once complete, could be developed further and made into a full service that generates revenue.
Myself:
– I will benefit from this project in a number of ways including the following:
o Growth in my research and documentation skills
o Growth in programming skills
o The project can be listed on my portfolio
Work plan
Activity Output Resources Start/End Date
Gain deeper understanding on the types of sentiment Enables me to implement useful summary/insights of data when implementing program Sentiment Analysis Wikipedia page
Websites such as MonkeyLearn have detailed breakdowns of the topics 31/10/19 – 07/11/19
Review API and Twitter datasets available online Will enable me to use the most appropriate dataset/API for my project Twitter has a developers site which provides information on the topic
Can also check forums of programmers who have used different APIs 07/11/19 – 14/11/19
Review Python twitter libraries such as ‘python-twitter’ (maintained by @bear), ‘tweepy’ (maintained by @applepie), and ‘Python Twitter Tools’ Will enable me to use the most appropriate Python twitter library for my project Twitter has a developers site which provides information on the topic
Programmer forums where users on the site have used different libraries and can review them 07/11/19 – 20/11/19
Implement my production using the Python language and libraries that access and manipulate twitter data as mentioned above Program that enables the user to carry out automated sentiment analysis on a twitter dataset Python online documentation
Code learning websites such as Codecademy
YouTube tutorials
24/11/19 – 15/2/20
Test and review the code Ensures that the code works correctly and is clear and concise Apply learnt principles of testing from university
Online tutorials and advice on testing code 15/2/20 – 28/2/20
Document the tests performed on the code Provides clear record of proof that the program functions as expected/highlights any amendments that need to be made Website pages specifically explaining how to document code tests well 15/2/20 – 28/2/20
Risks
** for likelihood and impact 1 refers to least likely/least impact and 5 is most likely/highest impact.
Risk Likelihood (1-5) Impact (1-5) Solution if risk occurs/Damage limitation
Big changes in requirements of the project at later stages 3 4 Making the PDD as extensive as possible so that the project is started with more of an understanding of the concept and how to carry out the project.
Making constant revisions to the PDD as research is done to keep it concise and accurate throughout.
Failures in the technologies used (e.g. Python libraries being third party) 2 4 Doing further research into each of the Python twitter libraries/APIs available that can help me when carrying out my project.
Researching what technologies were used in the creation of similar, currently available, sentiment analysis tools online.
If one particular researched library fails on me, switch to a different, appropriate, Python library to complete the project.
Run out of time for completing the project 2.5 3 Focus on core elements of the project to ensure the highest priority functions are completed so that the key objectives are completed mostly or entirely.
Incorrectly assuming certain parts of my planned program are feasible due to inexperience in the field 2.5 4 One way I can decrease the likelihood of this happening is by spending more time researching to a deeper level the programming field of sentiment analysis on a practical level.
One thing I can do to limit damage if it occurs would be to gain advice from a supervisor or using my research to find a solution.
2019-10-31-1572546966