Israeli's emotions become more negative after the truce
Summary
๏ปฟ
Sentiment analysis is a natural language processing (NLP) technique that involves determining the sentiment or emotional tone expressed in a piece of text. The goal of sentiment analysis is to understand the attitude, opinions, and emotions expressed in the text, whether they are positive, negative, or neutral.
In this study we try to analyze the sentiment of Israeli audience during the war in Gaza 2023. Emoji based sentiment analysis showed that Israeli audience become more negative after the truce (23-30 Nov.). They were more positive during almost all days of the war and we notice how the positivity drop and negativity increase after few days when truce stopped. On 15th of December, the negative opinions and feelings increased significantly and reached more than 90% of the overall feelings in that day.
Approach:
We applied Emoji based sentiment analysis of Telegram channels in our analysis. When a message - usually news - is being published in a Telegram channel, the subscribers express their feeling and emotions on the message by either a comment, or an emoji. We use the emojis selected on each message and counts to infer the sentiment distribution in the message. Aggregating the sentiments of all messages during a period of time - such as a day - can help us understand the sentiment of that period.
Data Collecting and Cleaning
Data source:
Before we collect emoji data, we have to select a Telegram channel for this analysis. Selecting the proper Telegram channel was based on few criteria:
- The selected telegram shall be a news agency, well known, and popular in Israel.
- The channel allows any type of emojis to be used on the messages, and does not restrict using specific and selective set of emojis.
According to this criteria, we selected YediotnewsIL channel which has around 500K subscribers.
Data retrieval:
To retrieve data from Telegram channel, we used telethon Python library. The code we developed to read Emoji data from Telegram channel will be shared soon.
Data preprocessing:
In this step the messages retrieved from Telegram is being preprocessed and prepared for analysis tasks as following:
- The input file is a messages file that every row contain all reactions relevant to a specific message. Where the final file being generated contains rows of reactions, where every row contains count for every emoji on a specific message. This will help run aggregation tasks on emoji, emoji category, sentiment, and message levels
- It is hard to do the analysis using a large set of emojis, so we tried to categorize them into two levels of groups. Populating the emoji category and sentiment for every reaction based on the following table:
| Emoji | Category | Sentiment |
|---|---|---|
| '๐', '๐' ,'โค' ,'๐ฅ', '๐', '๐', '๐ซก', '๐ฏ', 'โค๐ฅ' | Like | Positive |
| '๐คฃ', '๐', '๐', '๐พ', '๐คฉ', '๐ฅฐ', '๐','๐', '๐ค', '๐', '๐ค', '๐' | Happy | Positive |
| '๐', '๐คฎ', '๐ฉ', '๐คก', '๐ด', '๐ฅฑ', 'โ', '๐ฅด' | Dislike | Negative |
| '๐คฌ', '๐', '๐ก', '๐คฏ' | Angry | Negative |
| '๐ข', '๐', '๐ญ' | Sad | Negative |
| , '๐คท', '๐คจ', '๐คทโ' | Fear/suspecious | Negative |
| Other emojis | Others | Others |
- "Others" group contains many emojis that are least frequent and not explicitly classified into Positive or Negative
- There is no Neutral group as we managed to categorize the above emojis to either Positive or Negative
Calculations
The plot presented above shows proportion of the reactions of every category in each day starting from 7th of October until 15 December 2023. The proportions are calculated on every message first, and then aggregated to calculate the overall proportions of that day using formulas (1) and (2):
Results:
๏ปฟ
We can notice that the sentiment of Israeli audience become more negative after the 23-30 Nov. truce. They were more positive during almost all days of the war and we notice how the positivity drop and negativity increase after few days of truce end. On 15th of December, the negative opinions and feelings increased significantly and reached 90% of the overall feelings in that day.
