Mood Analysis Training Data

Question

Where can I get a corpus of documents that have already been classified as positive / negative for moods in the corporate domain? I need a large body of documents that provides reviews for companies, such as company reviews provided by analysts and the media.

I find those cases that have product and movie reviews. Is there a case for a business domain, including company reviews that match the language of the business?

Answer 1

http://www.cs.cornell.edu/home/llee/data/

http://mpqa.cs.pitt.edu/corpora/mpqa_corpus

You can use twitter with its smiles, for example: http://web.archive.org/web/20111119181304/http://deepthoughtinc.com/wp-content/uploads/2011/01/Twitter-as-a-Corpus- for-Sentiment-Analysis-and-Opinion-Mining.pdf

Hope you get started. There is more to the literature if you are interested in specific subtasks, such as denial, the realm of feelings, etc.

To focus on companies, you can associate the method with topic discovery, or cheaply just a lot of mentions about a given company. Or you can get your data annotated with mechanical turkers.

Answer 2

This is a list I wrote a few weeks ago from my blog . Some of these datasets have recently been included in the Python NLTK platform.

dictionaries

• Lexicon Opinion by Bing Liu

  • URL : http://www.cs.uic.edu/~liub/FBS/sentiment-analysis.html#lexicon
  • PAPERS : Development and synthesis of customer reviews
  • NOTES . Included in the NLTK Python platform.

• MPQA Subjectivity Lexicon

  • URL : http://mpqa.cs.pitt.edu/#subj_lexicon
  • PAPERS : Recognition of contextual polarity in sentiment-level sentiment analysis (Theresa Wilson, Janyce Wiebe, and Paul Hoffmann, 2005) .

• SentiWordNet

  • URL : http://sentiwordnet.isti.cnr.it
  • NOTES . Included in the NLTK Python platform.

• Harvard Principal Investigator

  • URL : http://www.wjh.harvard.edu/~inquirer
  • PAPERS : General Query: A Computer-Aided Approach to Content Analysis (Stone, Philip J; Dexter C. Dunphry, Marshall S Smith, and Daniel M. Ogilvy. 1966)

• Linguistic Query and Word Count (LIWC)

  • URL : http://www.liwc.net

• Vader Dictionary

  • URL : https://github.com/cjhutto/vaderSentiment , http://comp.social.gatech.edu/papers
  • PAPERS : Vader: an analytical model for analyzing moods in social networks (Hutto, Gilbert, 2014)

---

Datasets

• MPQA Datasets

  • URL : http://mpqa.cs.pitt.edu

  • NOTES : GNU Public License.

    • Political debate
    • Product Product Data
    • Sense Annotations Subjectivity

• Sentiment140 (Tweets)

  • URL : http://help.sentiment140.com/for-students
  • PAPERS : Twitter Sent by remote control (Go, Alec, Richa Bhayani and Lei Huang)
  • URL : http://help.sentiment140.com , https://groups.google.com/forum/#!forum/sentiment140

• STS-Gold (Tweets)

  • URL : http://www.tweenator.com/index.php?page_id=13
  • DOCUMENTS : Evaluation datasets for analyzing moods on twitter (Saif, Fernandez, On, Alani)
  • NOTES : like Sentiment140, but the dataset is smaller and with human annotators. It comes with three files: tweets, entities (with their mood) and an aggregate set.

• Customer Data Set (Product Reviews)

  • URL : http://www.cs.uic.edu/~liub/FBS/sentiment-analysis.html#datasets
  • PAPERS : Development and synthesis of customer reviews
  • NOTES : review title, product function, positive / negative label with opinion power, other information (comparisons, pronoun resolution, etc.).

  Included in the Python NLTK Platform

• Pros and Cons of Dataset ( Pros and Cons )

  • URL : http://www.cs.uic.edu/~liub/FBS/sentiment-analysis.html#datasets
  • DOCUMENTS : Opinion in comparative sentences (Ganapathibhotla, Liu 2008)
  • NOTES : list of sentences with <pros> or <cons> tags

  Included in the Python NLTK Platform

• Comparative offers (reviews)

  • URL : http://www.cs.uic.edu/~liub/FBS/sentiment-analysis.html#datasets
  • PAPERS : Identification of comparative sentences in text documents (Nitin Jindal and Bing Liu) , Features of Mining Opinion in customer reviews (Minqing Hu and Bing Liu)
  • NOTES : Sentence, POS-labeled sentence, entities, type of comparison (not equal, equivalent, excellent, non-gradable)

  Included in the Python NLTK Platform

• Sanders Analytics Twitter Sentiment Corpus (Tweets)

  • URL : http://www.sananalytics.com/lab/twitter-sentiment

  5513 manual tweets on 4 different topics. Because of Twitters ToS, a small Python script is included to download all tweets. The classification of feelings itself is provided free of charge and without restrictions. They can be used for commercial products. They can be redistributed. They are subject to change.

• Spanish tweets (tweets)

  • URL : http://www.daedalus.es/TASS2013/corpus.php

• SemEval 2014 (tweets)

  • URL : http://alt.qcri.org/semeval2014/task9

  You SHOULD NOT redistribute tweets, annotations, or the resulting corpus (from the readme file)

• Various Datasets (Reviews)

  • URL : https://personalwebs.coloradocollege.edu/~mwhitehead/html/opinion_mining.html
  • DOCUMENTS : Creating a model of intelligent analysis of moods using general purpose (Whitehead and Yaeger) , Moods using models of classification of ensembles (Whitehead and Yaeger)

• Various Datasets # 2 (Reviews)

  • URL : http://www.text-analytics101.com/2011/07/user-review-datasets_20.html

---

Literature:

• Keenformatics - analytic analytic vocabulary and data sets (my blog)
• Personal experience

Answer 3

Here are a few more:

http://inclass.kaggle.com/c/si650winter11

http://alias-i.com/lingpipe/demos/tutorial/sentiment/read-me.html

Answer 4

If you have resources (media channels, blogs, etc.) about the domain that you want to explore, you can create your own enclosure. I do this in python:

• using Beautiful Soup http://www.crummy.com/software/BeautifulSoup/ to analyze the content that I want to classify.
• separate these offers from positive / negative opinions about companies.
• Use NLTK to process these sentences, tokenize words, mark POS, etc.
• Use PMI NLTK to calculate bitrams or mos trigrams often in the same class

Creating corpus is a difficult job of preprocessing, checking, marking, etc., but it has advantages in preparing a model for a specific domain many times, increasing accuracy. If you can get an already prepared case, just continue to analyze the mood;)

Answer 5

I don’t know which such package is available, but you can try an uncontrolled method for an unlabeled dataset.

Answer 6

You can get a large selection of online reviews from Datafiniti. Most reviews come with rating data, which would provide more granular mood than positive / negative. Here is a list of enterprises with reviews , and here is a list of products with reviews .