Awesome Sentiment Analysis!

It is more and more common for the need to extract a polarity of sentiment from a text, and we get more precise in this work. We are even able to distinguish different types of emotions, mainly anger, joy, disgust, fear, and sadness.

Therefore, we have tried to identify for you all the major contributors involved in this field, and summarize what they are able to provide today as results.

1. IBM Watson

▷ Sentiment features
✅ Sentiment Polarity (float ranges from -1.0 to 1.0: very negative to very positive)
✅ Emotion classification (5 classes: joy, fear, anger, disgust, sadness)

▷ Other features
✅ Concept detection (among 800k concepts from datasets)
✅ Categories detection (ex: “/Computers & Electronics”)
✅ Keywords detection

⚠️ Cost: free up to 30,000 requests / month

2. Google Cloud Natural Language

▷ Sentiment features
✅ Sentiment Score (float ranges from -1.0 to 1.0: very negative to very positive)
✅ Sentiment Magnitude (strength of the sentiment regardless of score, ranges from 0 to infinity)

▷ Other features
✅ Entities recognition (organizations, locations, events, persons, consumer goods, …)
✅ Syntaxic tagging (Dependency, Parse Label, Part of Speech, Lemma, Morphology)
✅ Categories detection (ex: “/Computers & Electronics”)

⚠️ Cost: free up to 5,000 requests / month

3. Aws Comprehend

▷ Sentiment features
✅ Sentiment Classification (4 classes: Mixed, Positive, Negative, Neutral): a probability is assigned for each class

▷ Other features
✅ Part of Speech Tagging
✅ Entity Recognition (Location, organization, date, person, …)
✅ Language Detection
✅ Keywords Detection

⚠️ Cost: free up to 50,000 requests the first year (free tier) Read more here

4. Microsoft Cognitive Services

▷ Sentiment features
✅ Sentiment Score (ranges from 0 to 100%)

▷ Other features
✅ Key phrases extraction
✅ Linked Entities recognition (gives links to wikipedia articles)

⚠️ Cost: free up to 5,000 requests / month

You prefer to write your own code , using python libraries?

5. TextBlob library

✅ Sentiment Polarity Score (ranges from -1.0 to 1.0: very negative to very positive)
✅ Subjectivity Score (ranges from 0 to 1.0: neutral to very subjective)
✅ Syntaxic Tagging (Part of Speech, Spell corrector, translation…)

6. Stanford NLTK library

✅ Sentiment Trees (5 classes: very negative, negative, neutral, positive, and very positive)

You want to train your own models and need to find dataset?

50 free Machine Learning datasets: Sentiment Analysis

Is there any actor we forgot to mention, or any useful dataset? Feel free to comment, and share with us your sources!👇



This website uses cookies for marketing, analytics, to enhance user experience. More info OK