Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.

Sentiment Analysis for SEO

225 views

Published on

My slides from the recent Colchester Digital virtual Meetup where I discussed sentiment analysis, how it can be used for SEO, PR, reputation management and more and how you can perform it yourself for free using Google Sheets.

Published in: Marketing
  • Be the first to comment

Sentiment Analysis for SEO

  1. 1. Tell Us What They Really Think Sentiment Analysis for SEO Ben Johnston
  2. 2. HELLO I’m Ben Johnston • SEO & Data Analytics Director For ESV Digital • 13 years experience, 11 agency side • Focused around bringing data science techniques to digital marketing • Always looking to try new technologies • I love what I do • https://www.linkedin.com/in/ben-johnston-seo/ https://twitter.com/ben_johnston80 • https://www.ben-johnston.co.uk/
  3. 3. What We’ll Cover • What Is Natural Language Processing? • Why Does It Matter For SEO? • Introducing The Google Natural Language API • What Can We Use It For? – Analysing sentiment of URLs such as backlinks or mentions – Analysing sentiment of TrustPilot Reviews – Keyword Research • What This Teaches Us • Questions
  4. 4. WHAT IS NATURAL LANGUAGE PROCESSING?
  5. 5. What Is Natural Language Processing? • NLP is one of the holy grails of computing and data analysis • It’s about trying to make computers truly understand language, its nuances, its evolution and the intent behind the words rather than just the words themselves • Through this, we’re able to analyse and understand a lot more of the online ecosystem around our businesses or our clients • We can also use it to build better ways to engage with people, including automating those communications
  6. 6. What Is NLP Used For? • It’s used in more places than you might think – Google Translate – Chatbots – “AI” Twitter accounts etc • The possibilities are endless, but there’s one place that it’s important that doesn’t get talked about as much as I would like: search engines
  7. 7. NLP In Search • In 2018, Google said that they do not use sentiment as a direct ranking factor • However, we know that Google are not exactly trustworthy on statements like this and I question it • Updates such as BERT and EAT are very much focused around understanding the intent behind searches and content rather than just the words themselves • It would be brave to assume that this isn’t at least a small factor, especially with Google making their Natural Language API available to all • Bing have been public about it being a factor in their featured snippets
  8. 8. NLP In Search • In 2018, Google said that they do not use sentiment as a direct ranking factor • However, we know that Google are not exactly trustworthy on statements like this and I question it • Updates such as BERT and EAT are very much focused around understanding the intent behind searches and content rather than just the words themselves • It would be brave to assume that this isn’t at a small factor, especially with Google making their Natural Language API available to all • Bing have been public about it being a factor in their featured snippets • Ultimately, good sentiment around your business may not be a direct ranking factor at this stage, but it certainly influences a range of other factors
  9. 9. NLP In Search • Improving your brand’s sentiment online may not be a direct organic ranking factor, but it should be one of your core goals • Why? – It’s just good business – Word of mouth is still a strong decision-making factor – People reading good things about you online is likely to give them a nudge to engage, to speak about you or to link • By doing these things, you can improve your organic visibility and likely your conversion rates • Ultimately, providing good services in a way that makes people speak positively about you is one of the most authentic and effective ways to build and strengthen your brand • But how do we see how people are actually speaking, thinking and feeling about us?
  10. 10. HOW TO PERFORM SENTIMENT ANALYSIS
  11. 11. How To Perform Sentiment Analysis • There are lots of great tools with this function included: – URL Profiler – Sitebulb – SurferSEO – Brandwatch – IBM Watson • There are also a number of libraries for languages such as R, Python and Julia which can perform this very well and at scale • But what if there’s no budget for new tools or no developer resource?
  12. 12. The Google Natural Language API • Google have made their Natural Language API open to all and at very low cost – if you make 20 million queries a month, it will only cost $0.25! It will be free for up to 5,000 queries a month • The Natural Language API incorporates a large number of different elements, including entity analysis, translations and what we’re here to talk about today – sentiment analysis, the understanding of how positive or negative a piece of content is in tone • The best part? We’re going to do this in Google Sheets
  13. 13. Sentiment Analysis In Google Sheets • Here’s one I made earlier! • https://docs.google.com/spreadsheets/d/1Urw5NmT9vEo CS9JUbpJvkC9wNR7mwnuqsugws- KrN0E/edit?usp=sharing • Make a copy of it into your own Google Drive • It’ll let you do the following: – Analyse the sentiment of URLs – Analyse the sentiment of TrustPilot reviews – Perform sentiment analysis and keyword research with the SEMRush API • You don’t need to edit any scripts or formulas, but there is a little setup to do
  14. 14. SETTING UP THE GOOGLE SHEET
  15. 15. Getting A Google Natural Language API Key • Once you’ve made a copy of the sheet into your own Google Drive, you’ll need to sign up for Google Cloud Services – they’ll give you $300 credit for free • Enter the console and select New Project – it’s a nice big button! • In the new project, navigate to API’s & Services and find the Cloud Natural Language API. Enable that
  16. 16. Getting A Google Natural Language API Key • Now navigate to API’s & Services and click Create Credentials • From the dropdown, select “API key” and it will create an API key for you • Keep this somewhere safe. You’ll need it for the next step
  17. 17. Linking Your API Key With The Google Sheet • Once you’ve made a copy of the sheet, the only thing you need to do to get it running is to link your API key with the script • Enter the Script Editor (it’s under Tools) • Then go to Project Properties under File • Select Script Properties • Paste your API key into the Value box
  18. 18. SO WHAT CAN WE DO WITH THIS?
  19. 19. Sentiment Analysis Of Your Backlinks/ Mentions • The first tab of the sheet allows you to export all the written text from almost any URL and run the Natural Language API’s Sentiment Analysis model on it • Magnitude is the degree of salient words that have been analysed for sentiment and the score is how strong the overall sentiment is, scored between -1 and 1 • A score of zero is Neutral, below zero is negative and above zero is positive • The higher or lower the score, the stronger that sentiment is • All you need to do is paste your URLs into the URL column, the sheet will do the rest • It’s best to do this in smaller batches though – Google Sheets isn’t always the fastest
  20. 20. What Can We Do With This? • While backlinks are still a factor within SEO, we’re not quite sure if or how much of a factor the sentiment of them is (personally, I think it’s a component of the Trust section of Expertise, Authority and Trust) • However, even if there is no ranking benefit whatsoever, it’s important for SEO and PR teams to understand how people are speaking about us • This tab doesn’t only need to be using backlinks – it can be unlinked mentions, press coverage, almost any kind of page and if they’re speaking negatively of you, you can quickly identify them and look into why • From here, the opportunity to turn an unhappy customer into a happy one is easier to find than ever • You can also export that text and run it through a number of different tools such as word cloud generators in order to see what kind of phrases people are using the most to talk about your business, again, perhaps giving you valuable insights into the kinds of content you can create to serve your audience
  21. 21. Sentiment Analysis of TrustPilot Reviews • TrustPilot is one of the main places where people will gather to talk about your business • The reviews are usually more polarised than not – people are more likely to review a product or service when it’s either fantastic or terrible • On the second tab of the sheet, you can export the average review score and all the reviews from TrustPilot profiles and, again, run the Natural Language API on it to identify the sentiment of those reviews • Again, just paste the URL in, the sheet will do the rest
  22. 22. What Can You Do With This? • Keeping an eye on your own or your clients’ reviews is common sense • You can also use this to gain an overview of the market before a new venture or a new product, for example • It can also be used to keep an eye on the competition, seeing where they might be struggling and identifying opportunities for yourself • The possibilities are almost endless and there is a wealth of data there to acquire and analyse
  23. 23. Keyword Research With Sentiment Analysis • Keyword research is one of my favourite areas to use sentiment analysis • We always need to be looking beyond just search volumes and examining a range of other areas – especially the SERPs for our market • There are lots of different ways and tools that can help you do this and this Google Sheet isn’t necessarily the best way to scale that • What this will do, however, is give you an indication • Is the SERP around that keyword a toxic wasteland? Do we actually want to push our client into that area? Are there some unfavourable associations with our new product’s name that we haven’t thought of? This will help us find them
  24. 24. Keyword Research With Sentiment Analysis • In the Keyword Research tab, you can pull a range of useful data for a targeted keyword, including the search volume, cost per click, competition and the number of results for that term • The sheet also extracts the top ten URLs ranking for that term, downloads the content and analyses the sentiment of that URL, before giving you an average score on for that keyword • Ultimately, this will allow you to understand if a certain search term has a negative or positive connotation and give you the opportunity to make decisions around whether you want to target it • Just type your Keyword into the far-left Keywords column, and off it will go • You’ll need an SEMRush API key for this to work – there’s a cell to the right where you can paste it in • Also, only do five keywords at a time, otherwise you’ll be waiting a while
  25. 25. What Can We Do With This? • Understanding the search landscape and the market area you’re targeting is one of the most important parts of SEO • We need to be looking beyond just search volumes • We need to understand who our customer is, what they think about products or services, how they feel about them and what we can do to set ourselves apart • This process won’t do all of that for you, but hopefully it gives you enough inspiration to perhaps look into it further • I’m looking into adding social media elements into this sheet in the future, where you’ll be able to analyse Tweets and Facebook comments • The possibilities really are endless and I really hope more people will start using this as part of not just their keyword research, but their SEO, PR and marketing in general
  26. 26. WRAPPING UP
  27. 27. What Have We Learned From This? • Sentiment analysis and natural language processing isn’t perfect – nothing is – but it’s becoming an increasingly powerful weapon in the arsenal • By having an indication of how our market, our potential customers and our existing customers speak, think and feel about the offerings we’re marketing, we have vital information around how to reach them • When we know how they’re thinking and feeling about products or services, we can examine further and ensure that we are offering something which addresses those concerns, needs or feelings • We can create content that actually targets what people need to know rather than just creating content for content’s sake because it gets a decent number of searches a month • This is how we build trust – by speaking the customer’s language, by pre-empting their concerns and, crucially, by adapting and delivering on our promises • Sentiment analysis won’t do that for you – but it will point you in the right direction
  28. 28. QUESTIONS THANK YOU

×