The document discusses Google's ML APIs versus OpenAI's APIs and their applications for SEO and digital marketing tasks. It provides examples of how natural language processing APIs from Google and OpenAI can be used for tasks like text analysis, sentiment analysis, document classification, translation and content transformation. While both Google and OpenAI APIs are useful, the document recommends choosing the right API for each specific task based on its capabilities and limitations in order to get the best results.

1. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
#pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
Google's ML APIs vs OpenAI's APIs
(SEO Use Cases)
Machine Learning and Search

2. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
● SEO & Data Science Lead
● Intrepid Digital - enterprise
digital marketing agency
● Client Portfolio Focus: B2B
Enterprise in Big Tech
● MSc, specialising in ML/NLP

3. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
My main goal:
to inspire you to think creatively
about available ML APIs and how you use
them in your day-to-day.

4. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
There are several areas,
where ML automation can help
boost efficiency and improve outcomes for
digital marketers.

5. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
You will also see why it’s important to
choose your automation allies carefully
depending on the task.

6. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
SEOs
OpenAI
AWS, Azure
Google Cloud

7. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
10
APIs by Google Cloud
(each with multiple applications for digital marketing)

8. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
3
APIs by OpenAI
(again, each with multiple applications for digital marketing)

9. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.

10. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
#pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
Understanding text
Entity Recognition

11. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
Entity recognition
Named Entity Recognition (NER) is a field
of computer science and natural language
processing that deals with the
identification and classification of
named entities in text.
The goal of NER is to automatically
extract information from unstructured
text, such as names of people,
organizations, locations, and so on
tl;dr entity = recognised thing/concept

12. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
Google’s Natural Language API vs ChatGPT

13. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
Google’s Natural Language API vs ChatGPT
53 16

14. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
With Google’s Natural
Language API, you can
also do Syntax analysis
● Dependency
● Parse label
● Part of speech
● Lemma
● Morphology

15. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
How important is text understanding for
marketers?

16. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
Internal link opportunity identification
quickly find pages to link to
Anchor text identification
quickly identify the text of use for
the links you’ve identified

17. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
Content understanding
quickly understand what
content is about, even at
scale
Content Gap analysis
Check whether the
content on the site aligns
with the business
direction desired
Competitor analysis
quickly understand
competitors’ content
catalogs

18. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
Topic understanding
understand the entities that make up
a topic, based on analysing the
content that ranks
SERP understanding
entities in titles, meta descriptions,
URLs, breadcrumbs

19. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
#pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
Clustering / Classification
Keyword Research, Topic Research, Content Research

20. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
Classification vs Clustering - What’s the difference?
Classification sorts data into specific
categories using a labeled dataset.
Clustering is partitioning an
unlabeled dataset into groups of
similar objects.

21. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
With Google’s Natural Language API, you can classify
documents in 700+ predefined categories
(out of the box, can be custom-trained, too with AutoML)

22. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
With Open AI’s GPT-3 or with ChatGPT, you can do both,
but results are a hit or miss.

23. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
Can you guess what can go great?
(and what - horribly wrong)

24. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
✅Predictable categories
✅Controlled training of model
✅Accuracy indicated
✅Great for scale and
beckmarking
✅Can map the information to a label
or assign a plausible such, provided it
has this information in its training set
✅Very adaptive
✅Great for small projects, one offs

25. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
❌Can’t be used for uses outside
of the main task
❌Can’t be given custom lists
(...unless)
❌Requires time and data for
custom training models with
AutoML
❌Non-predictable results
❌Direction might not followed
❌Model not trained for task
❌Limited knowledge
❌Unsuitable for niche industries

26. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
How important is classification for marketers?

27. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
Content understanding
quickly understand what topics the site is
covering with the content
Content Gap analysis
check whether the content topics on the site
aligns with the business direction desired
Competitor analysis
quickly understand the topics that
competitors’ content talks about

28. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
Topic understanding
quickly understand what topics the
keyword universe you have consists of
Keyword clustering
quickly understand how other parameters
of keyword research relate to the clusters
identified

29. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
#pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
Understanding emotions
Sentiment Analysis, Entity Sentiment

30. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
Sentiment analysis
Understand the
overall opinion,
feeling, or attitude
sentiment expressed
in a block of text.

31. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
With Google’s Natural
Language API, you can
quickly get
● Document and sentence
level sentiment analysis
● Salience = importance
● Sentiment Score
● Sentiment Magnitude =
strength
● # of mentions = entity
prominence
● Entity Sentiment

32. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.

33. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
How important is sentiment analysis for
digital and search marketing?

34. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
Analyse feedback
Identification of competitive
advantages, based on users
(entities where your brand
excels)
Online Reputation
Management
Analysing social media
comments, and other user
interactions
Content idea mining
Content opportunity
identification, including FAQ
content, new content ideas,
etc.

35. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
#pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
Image SEO Hygiene
Image captioning, image generation

36. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
Generate images from text (+edit, +variate)

37. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
…but no captioning?
(yet)

38. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
“combines a vision CNN with a language-generating
RNN so it can take in an image and generate a fitting
natural-language caption”

39. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.

40. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
#pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
Content Transformation
Speech to text, text to speech, text summarization

41. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
Omni-presence
and Accessibility
Both users and search engines
want to see multi-modal
presence for high-value sites.
Meaning:
● Text to video
● Videos to text
● Text to audio
● Audio to text
● Text summaries for longer
texts (e.g. FAQs, headings)

42. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
Shorten ✔
Lengthen ✔

43. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
Answer ✔
(with some
limitations)

44. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
Multi-modal
Transformation ✔

45. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
Needless to say scaling production here is
pretty significant for organic growth.

46. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
You have a library of
videos on YouTube but
no blog?
→ Scale it’s launch
quickly through
transcription.

47. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
You have a library of
high-performing blog
posts but no presence on
YouTube/TikTok?
→Scale production
with text to speech.

48. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
Want to optimize for
high-intent visitors or improve
accessibility of content?
→ Provide summaries of
sections, improve headings,
add FAQs, at scale.
(...)

49. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
#pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
Content Localization
Translation

50. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.

51. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.

52. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
1. DeepL : 0.19001633345126925
2. ModernMT : 0.17788391513374424
3. Amazon : 0.16483921567053203
4. Neuralspace : 0.163133354485786
5. Google : 0.16280640903935437
6. OpenAI : 0.15934198508564865

53. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
#pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
Performance Analysis
Time series analysis, anomalies detection, forecasting

54. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
#pubcon @Pubcon | @lazarinastoy Lazarina Stoy.

55. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
so, who won?

56. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
Entity Recognition, including Text analysis 👑 ✔
Sentiment Analysis 👑 ❌
Document classification 👑 ✔
Keyword Clustering ✔ ✔
Image Captioning ✔ ❌
Image Generation ❌ 👑
Content Transformation (multimodal) ✔ ❌
Content Transformation (text-based) ❌ ✔
Content Localization (translation) ✔ ✔
Timeseries Analysis - Anomalies Detection, Forecasting 👑 ❌
Text Generation ❌ 👑

57. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
If there’s anything that you should take
away from your talk…

58. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
ML-based APIs are amazing for SEO.
You should absolutely use them.

59. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
They can help with:
✔ boosting your efficiency and speed

60. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
They can help with:
✔ skyrocketing insight-generation speed
for large-scale projects and sites

61. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
They can help with:
✔ career development and growth.

62. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.

63. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
Don’t just use a blanket model for all your
tasks just because you’ve seen it go viral.

64. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
Research what each model you use has been
trained for, and what’s it ideal use before
you start.

65. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
Then relate that back to the task at hand.

66. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
Expect limitations…
regardless which API you choose.

67. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.

68. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.

69. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.

70. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.

71. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
We’re near the end now…

72. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.

73. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.

74. #pubcon @Pubcon | @lazarinastoy Lazarina Stoy.
what I do my website
seo@lazarinastoy.com
social handle
best way to get in touch