Skip to content

Commit

Permalink
Merge branch 'master' of github.com:googleinterns/step99-2020 into va…
Browse files Browse the repository at this point in the history 
…nilla-bars-backend
  • Loading branch information
@gracetian6
gracetian6 committed Aug 1, 2020
2 parents d6e5d05 + 16a7930 commit 232e799
Show file tree
Hide file tree
Showing 39 changed files with 8,614 additions and 200 deletions.
8 changes: 2 additions & 6 deletions .github/workflows/lint.yml
View file
Original file line number Diff line number Diff line change
Expand Up @@ -3,12 +3,8 @@
name: Lint

# Controls when the action will run. Triggers the workflow on push or pull request
# events but only for the master branch
on:
push:
branches: [ master ]
pull_request:
branches: [ master ]
# events
on: [push]

# A workflow run is made up of one or more jobs that can run sequentially or in parallel
jobs:
Expand Down
137 changes: 137 additions & 0 deletions client/analysis.html
View file
Original file line number Diff line number Diff line change
@@ -0,0 +1,137 @@
<!DOCTYPE html>

<html lang="en">
<head>
<meta charset="utf-8">
<meta name="description" content="STEP Capstone">
<link rel="stylesheet" href="css/base.css">
<link rel="stylesheet" href="css/analysis.css">
<title>Comment Analysis</title>
<base target="_blank">
</head>

<body>
<div id="content">

<div id="opening" class="center">
<div id="opening-wrapper">
<h1>Youtube Comment Analysis.</h1>
<ul>
<li>Enter a video name to see what people think!</li>
<li>Click <a onclick="toggleLightboxVisibility(true)">here</a> to learn more.</li>
</ul>
<div class="lightbox hide-lightbox" id="info">
<div id="info-box">
<h1>Welcome to my project!</h1>
<h3>Introduction</h3>
<p>
This is my portion of the 2020 Google STEP Capstone project.
With this feature, you can enter in a Youtube video name or ID
and get a solid estimate on how the video was received via
comment analysis.
</p>

<p>
In order to achieve this, I took advantage of the
<a href="https://cloud.google.com/natural-language">Google Natural Language (NLP)</a>
and <a href="https://www.perspectiveapi.com/#/home">Google Perspective</a> APIs
to analyze comments returned to me from the
<a href="https://developers.google.com/youtube/v3">Youtube Data API</a>.
</p>
<h3>Method</h3>
<p>
In order to ensure the most accurate analysis while handling the minimum amount
of data, this tool takes 20 random comments from the Youtube API, aggregates them
into a single string of sentences, and runs the chunk through both the NLP and
Perspective APIs. Here is what each API is responsible for:
</p>
<!--Potentially, we will take the most popular comments.-->
<p>
NLP
<p class="indent">
NLP returns a magnitude and score value which give us the Overall Response.
The score value contributes to the corresponding adjective:
Positive, Negative, or Mixed. If a video is "Clearly" one
of these values, NLP returned a high magnitude rating and thus is confident
that the given data supports its score. Click
<a href="https://cloud.google.com/natural-language/docs/basics#interpreting_sentiment_analysis_values">
here</a> for the official documentation.
</p>
</p>
<p>
Perspective
<p class="indent">
Perspective returns a series of probabilities which correspond to attributes such
as Toxicity, Insult, Identity Attack, etc. These probabilities indicate how likely
the text is to have these effects on the readers. It is important to note that all
attributes except for Toxicity are in the experimental phase right now. Click
<a href="https://github.com/conversationai/perspectiveapi/blob/master/2-api/models.md#all-attribute-types">
here</a> for further explanation.
</p>
</p>
<h3>Programming techniques</h3>
<p>
I've implemented techniques like caching, pattern-matching,
search, and string manipulation to provide the best possible experience
and just to have some fun with this.
The code is all open source and can be found
<a href="https://github.com/googleinterns/step99-2020">
here</a>.
</p>
<h3>Limitations</h3>
<p>
Unfortunately, there are some limitations on how the tool can work
since this technically is a Google product and I am under
the jurisdiction of the company. Here are the biggest ones:
<ul>
<li>
The API quota only allows for 94 search calls a day,
assuming no visitors use the Youtube API on one of my podmates' pages.
Please use a video ID if you can!
</li>
<li>
I can only fetch and parse through so many comments with the small quota.
The video matching and comment parsing algorithms I have are decent, but
they are constrained.
</li>
</ul>
</p>
<h3 id="right-header"> - Seun Omonije </h3>
<h4><a onclick="toggleLightboxVisibility(false)"">Click here to close.</a></h4>
</div>
</a>
</div>
</div>

<div class="flex-row center" id="upper-content">
</div>

<div class = "flex-row center">
<div id="header-wrapper">
<h1>Enter video id or name below:</h1>
</div>
</div>

<div class="flex-row center">
<div id="search-wrapper">
<input type="text" id="searchbar" />
<button id="sendbutton">Send</button>
</div>
</div>

<div class="flex-row" id="middle-content">
<div class="flex-column">
<h2 class="hidden" id="commentHeader">Featured comments:</h2>
<ol id="list"></ol>
<div id="videocard-wrapper">
</div>
</div>
<div class="flex-column">
<div class="grid-container" id="charts"></div>
</div>
</div>
</div>
</body>
<script src="js/analysis.js"></script>
<script src="js/caching.js"></script>
</html>
Loading

0 comments on commit 232e799

Please sign in to comment.