Sentiment Analysis API
In this Post
Sentiment tracking involves the process of computationally identifying and categorizing the opinions expressed in textual content, such as news articles, social media posts, or online reviews. By analyzing language patterns and contextual cues, sentiment analysis algorithms can determine whether the expressed context of a keyword is positive, negative, or neutral. You can access this data via API.Table of Content
Intro
Sentiment tracking API gives you free, unlimited access to the results of computationally identifying and categorizing the opinions expressed in global news articles.
Sentiment search tab is an access point to a RESTful API that enables users to get historical media sentiment data and keyword centrality scores for any given word or topic of interest. Go on the front page of TextVisualization.app. Enter a word in Sentiment search and explore the word’s sentiment and role in media narratives
We collect data from mass media publications and use AI algorithms to calculate sentiment score averages for all instances of a keyword within a given timeframe. These sentiment scores quantify the overall tone surrounding the keyword, allowing the use to instantly see media coverage shifts.
Understanding the API Endpoint
The free, open access API endpoint that we offer is built for tracking the sentiment and narrative centrality of words in the English language. It can be used for monitoring the reputation of a public figure, tracking brand perception, or simply exploring trends. This API provides access to rich data that can inform various analyses.
Let’s consider an example usage of the API to track the sentiment and narrative centrality of a specific word, «test.» By querying the API endpoint with the word «test,» users receive links to CSV files containing historical frequency, sentiment, and narrative centrality data associated with the word. These datasets can be used as features for AI classification tasks. The data shows how the word «test» has been perceived and discussed in the news over time.
API example:
https://textvisualization.app/search/api/v1/?query=test
This will return JSON:
{"frequency_csv":"https://textvisualization.app/csv/word/frequency-test.csv","sentiment_csv":"https://textvisualization.app/csv/word/sentiment-test.csv"}
What data do you get with API?
Sentiment search API gives you a wealth of linguistic data that describes the search term, in csv format.
frequency_csv file headers:
word,Timestamp,TF,mean,variance,k,p,sqrt_kp,q_p,cmi,DF,DF_TF,TF_IDF,entropy,fisher
sentiment_csv file headers:
timestamp,word,mean_polarity,min_polarity,max_polarity,variance_polarity,mean_subjectivity,min_subjectivity,max_subjectivity,variance_subjectivity
These variables can be used as features for various classification models.
Polarity Sentiment and Narrative Centrality
The first measurement that the media sentiment tracker app shows is called ‘polarity sentiment’. It reflects the overall circumstances associated with a particular keyword over time. A higher polarity sentiment score indicates a generally optimistic, favorable, nonviolent, even a happy context, while a negative score suggests a pessimistic, unfavorable, unhappy sentiment. Such contexts are associated with suffering caused by war or crime, media outrage at a particular topic, and so on.
For instance, when tracking mentions of political figures like Trump or Biden in mainstream media (MSM), a higher polarity sentiment score would indicate positive sentiments expressed in the context surrounding these individuals, whereas a negative score would signify a more critical or unfavorable perception.
AI plots these scores over time, so that we can visualize trends and fluctuations in media coverage.
The second measure, narrative centrality, quantifies the significance or prominence of a word within news narratives over time. This metric can help users understand which topics or themes are currently dominating the news cycle and how specific words contribute to those narratives. The score is different from term frequency, but it is derived from word frequency distribution using the negative binomial distribution modelling. The measure is not correlated with frequency, instead it demonstrates how much attention is given to a term in discourse, measuring word aggregation inside texts within a corpus.
Sentiment search
Sentiment search API returns same results as the Sentiment search tab on the home page of textvisualization.org.
At the front page of TextVisualization.app enter a word in Sentiment search to view and explore the word’s sentiment and role in media narratives.
With a constantly expanding search base, encompassing a wide scope of mass media keywords, you can freely access a wealth of historical sentiment data and usage trends. This comprehensive and ever expanding database allows for thorough analysis and exploration of sentiment dynamics over time. Try search for countries - and you will get a kind of political geography guidebook of how this or that country is covered in global press. Use the search for keywords like, ‘gold’, ‘bitcoin’, ‘recession’ - you will see a quantified summary of media coverage over time.
Sentiment Search functionality allows to uncover patterns, correlations, and sentiment trends associated with specific keywords or topics. Whether tracking the sentiment surrounding a particular product launch, monitoring public perceptions of political figures or sentiment shifts in response to current events, this tool mines into the flow of public opinion. Furthermore, the inclusion of keyword centrality scores enhances the depth of analysis by identifying the prominence and relevance your search term within the media landscape.
