Need Tagged Data sets for training ML algorithms to classify sentences/reviews into single/multi classes

Question

1. Need tagged text data of any domain. Preferably generalized and ubiquitous domains such as "Food & Hospitality" (Reviews about the quality of the food, comments about the ambience and service etc. This can be found in reviews for restaurant websites or food delivery and review sites like Zomato).
2. More Details : Each review might be classified into various classes*. This makes it a multi-class classification problem. Classes* - Important "aspects" that the review is talking about.

For example

Here I also mentioned the Sentiment Column (Which is 3 class, Positive, Negative and Neutral). Typically every domain, a broad class of aspects would be decided and tagged.

Can anyone link me to good data sets that meet the requirement? I will be needing hundreds of thousands data. The more, the better. Or suggest me an alternative way ?

Answer 1

Yelp holds a competition to find trends in its data. According to the website it contains 4.1 million reviews. Reviews include text and a rating (and other values).

You can also use the data for academic purposes (terms of use).

You can find the dataset (and more information) here: https://www.yelp.com/dataset_challenge

The fields in each review object:

{
  "review_id": "encrypted review id",
  "user_id": "encrypted user id",
  "business_id": "encrypted business id",
  "stars": star rating, rounded to half-stars,
  "date": "date formatted like 2009-12-19",
  "text": "review text",
  "useful": number of useful votes received,
  "funny": number of funny votes received,
  "cool": number of cool review votes received,
  "type": "review"
}

Answer 2

Not sure if this is exactly what you're looking for but you might want to check out the datasets here https://webhose.io/datasets, specifically under "online reviews" - where you can find large amounts of structured data divided into several broad categories. Data is in JSON format and you need to create a (free) account to download it.

Answer 3

UCI Machine Learning repository offers a small sample of sentiment analysis for reviews from

• imdb.com
• amazon.com
• yelp.com

https://archive.ics.uci.edu/ml/datasets/Sentiment+Labelled+Sentences

For each website, there exist 500 positive and 500 negative sentences. Those were selected randomly for larger datasets of reviews. We attempted to select sentences that have a clearly positive or negative connotaton, the goal was for no neutral sentences to be selected.