Home Screen
Add New
In-System Channels Listing
Add New
Channel Details Page
Sentiment Log
Running Sentiment Analytics On Video
In-System Videos Listing

Inspiration

Youtube is one of the largest video sharing sites in the world. A study by forbes.com shows reveals that there are currently 50 million content creators spread across Youtube, Instagram, TikTok and other social media. In this figure, Youtube has 12 Million amateur and 1 Million professional (full time) content creators illustrated by the image below: Social Media Content Creators Pie Chart

Furthermore, youtube statistical data provided by review42 suggests that:

There are over 1.9 Billion active users each month visiting Youtube
Content Creators upload more than 500 hours of content every minute
User watch over a billion hours of content daily.

From the 3 points mentioned above this implies that content creators have need to track the feedback on their work and going through every comment to ascertain sentiment manually would be time-consuming and or inefficient. This is especially true when a content creator has over a million subscribers and the average number of comments per video spans into the thousands. More so, professional content creators who make youtube videos for a living like Mr Beast have more than one channel.

Enterprises worldwide also have youtube channels where they showcase products and concept they would want to birth into the world. Users usually give feedback in the comment section and it's usually in the thousands. Running analytics of this feedback data could help product development and decision making on what the market reception is likely to be

The goal is to create a platform that allows Youtubers/ Content Creators/ Enterprises to efficiently monitor the Channel Brand Health and provide Realtime Sentiment Analytics powered by Artificial Intelligence.

What it does

Youtube Sentimeter allows Content Creators to have a consolidated view of all their channels providing valuable A.I generated insights such as overall channel brand health and a breakdown of sentiment analytics by video. It combines a rules-based algorithm with Modzy A.I models to deliver brand health insights per channel and sentiment analytics per video. These two metrics allow the Content Creator to ascertain the impact of their work(and additional work they upload) on the overall channel health allowing them to forecast their current market reputation over time.

How we built it

Youtube Sentimenter is a Django Web App that uses the Youtube API to fetch comments from a video. Each video belongs to a channel which the Content Creator has to designate. When the Content Creator navigates to the video details page and clicks on analyse this video button, the App uses the Youtube API to fetch related comments, cleans them and then feeds them into a Modzy A.I Sentiment Analysis model. When the Modzy A.I models returns the inference, the resulting data is fed into a built- in rules-engine that calculates and converts sentiments to percentages that the Content Creator can understand.

Algorithm to conduct sentiment Analytics on Video Code Snippet

job = client.jobs.submit_text('ed542963de', '1.0.1', sources)
                result = client.results.block_until_complete(job, timeout=None)
                results_json = result.get_first_outputs()['results.json']
                class_predictions = results_json['data']['result']['classPredictions']
                for api_object in class_predictions:
                    if api_object['class'] == "positive":
                        positive_score = api_object['score']
                    elif api_object['class'] == "negative":
                        negative_score = api_object['score']
                    elif api_object['class'] == "neutral":
                        neutral_score = api_object['score']

                if positive_score > negative_score and positive_score > neutral_score:
                    posfeed+=1
                elif negative_score > positive_score and negative_score > neutral_score:
                    negfeed+=1
                elif neutral_score > negative_score and neutral_score > positive_score:
                    neutfeed+=1
                elif positive_score == negative_score and positive_score > neutral_score:
                    neutfeed+=1
                elif neutral_score == negative_score and neutral_score > positive_score:
                    negfeed+=1
                elif neutral_score == positive_score and neutral_score > negative_score:
                    posfeed+=1
                elif neutral_score == positive_score and neutral_score == negative_score:
                    neutfeed+=1
                i+=1

        tota = int(posfeed+negfeed+neutfeed)
        pospercnt = round((posfeed/tota)*100)
        negpercnt = round((negfeed/tota)*100)
        neutpercnt = round((neutfeed/tota)*100)

Algorithm to calculate Channel Brand Health Code Snippet

    for a in q:
        c = Sentiments.objects.filter(video_id = a.video_id)
        for b in c:
            posi += int(b.sent_positive)
            nege += int(b.sent_negative)
            neut += int(b.sent_neutral)
            svcnt+=1
        pos_avg = round(posi/svcnt)
        neg_avg = round(nege/svcnt)
        neut_avg = round(neut/svcnt)
        if pos_avg >= neut_avg and pos_avg > neg_avg:
            chanhealth = "Good"
        elif neut_avg > pos_avg and neut_avg >= neg_avg:
            chanhealth = "Neutral"
        elif pos_avg <= neut_avg and pos_avg < neg_avg:
            chanhealth = "Critical"
        elif pos_avg == neut_avg and pos_avg == neg_avg:
            chanhealth = "Neutral"
        fstring = str(pos_avg)+'|'+str(neg_avg)+'|'+str(neut_avg)+'|'+str(chanhealth)

Potential Value

With the rapid increase in content consumption on Youtube as mentioned above, Youtube content creation has become a lucrative business with some Youtubers becoming millionaires. Success stories such as PewDewPie and MrBeast ( to name a few) have motivated more users to become content creators too. The YouTube Sentimeter App comes in handy to YouTubers as it can help them track their efforts across all their channels from one consolidated view. This can help them carry out what went wrong and what went right analysis hence understanding what content appeals to their subscribers.

Accomplishments that we're proud of

A tool to complement the efforts content creators wordwide has been successfully created powered by Modzy A.I

What we learned

How easy it is to interact with the modzy A.I plaform. A.I is really a powerful tool and can do much of the needed heavy-lifting for solving business problems.