Social Network Analysis: Twitter by Sarah Hewitt

Everything?

sarah_tiggy21 — 2015-07-08 11:03:14 UTC

'All tweets' doesn't mean all tweets. And let's not forget that what happens on twitter isn't necessarily a reflection of the real world. It's just a snapshot, and has to be considered as representative when measured against other factors. Which is probably why DEMOS linked their analysis with news events.

Bias

sarah_tiggy21 — 2015-07-09 12:54:02 UTC

"One drawback of the Streaming API is that it only returns at most 1% of the tweets on Twitter at a given moment. Once the volume of the query surpasses 1% of all of the tweets on Twitter, the response is sampled. The way in which Twitter samples the data is unpublished. "

Sample API: "...the Sample API takes no parameters and returns a 1% sample of all of the Tweets produced on Twitter."

Search API:

Pew Internet Research

sarah_tiggy21 — 2015-07-09 12:58:57 UTC

Find this.... DONE!

NodeXL

sarah_tiggy21 — 2015-07-09 13:22:53 UTC

No tweets more than a week old.

The algorithm used by twitter to match tweets is not the same one it uses on its own home page. Results may vary!

Language & Bias

sarah_tiggy21 — 2015-07-09 14:00:28 UTC

How relevant is bias (see post-it above) when looking at the use of mosygynist language on twitter?

The Sample API would seem to be the most appropriate source. What about taking samples on various days and at different times, and carrying out a comparison?

Also: geo data is unreliable. Are there other ways of assessing if the tweet originates from the UK?

Remember

sarah_tiggy21 — 2015-07-10 11:35:29 UTC

What people say in public doesn't necessarily reflect what they REALLY think, e.g. Mary Beard confronting her abuser, and deinviduation

Message Invectives Scale

sarah_tiggy21 — 2015-07-10 11:42:35 UTC

An effective way of surveying what other people might regard as misogynist, not misogynist, of how much misogyny is present. The people surveyed (age, level of education etc. might also have an impact on results). And gender.

Data Sampling

sarah_tiggy21 — 2015-07-11 10:01:50 UTC

Snowball, or 'breadth-first search'; an algorithm for traversing graph data structures.

IDDFS: Iterative deepening depth-first search. Both well explained in wikipedia.

Random walk: Markov Chain "...a random process that undergoes transitions from one state to another...". " It must possess a property that is usually characterized as "memoryless": the probability distribution of the next state depends only on the current state and not on the sequence of events that preceded it." Lattice path. Self-avoiding walk. There are others.

F-Test

sarah_tiggy21 — 2015-07-11 13:49:19 UTC

"An F-test is any statistical test in which the test statistic has an F-distribution under the null hypothesis." The analysis of variance.

Scaling Up

sarah_tiggy21 — 2015-07-11 14:13:54 UTC

Does this mean that if you apply a 'test' to data set x, and then apply the same test to the considerably larger data set y, the results will contain errors that make it unreliable? Is this a bit like scaling up a digital image and finding it's pixelated, and the software you're using has had to guess the best content for the new pixels based on their neighbours?

Grammar

sarah_tiggy21 — 2015-07-11 15:06:03 UTC

Identify how the words you've selected are used e.g. nouns, adjectives, adverbs, verbs etc.

User's Intentions

sarah_tiggy21 — 2015-07-13 09:32:09 UTC

Ethics: What did the user intend with their tweet?

What do users intend anyway? Will impact on research.

Natural Language Processing

sarah_tiggy21 — 2015-08-16 13:55:46 UTC

... is a field of computer science, artificial intelligence, and computational linguistics concerned with the interactions between computers and human (natural) languages.

SENTIMENT ANALYSIS (also known as opinion mining) is part of this.

Sentiment analysis

sarah_tiggy21 — 2015-08-16 13:58:47 UTC

...refers to the use of natural language processing, text analysis and computational linguistics to identify and extract subjective information in source materials.

Machine Learning: Opinion as a Quintuple

sarah_tiggy21 — 2015-08-19 10:20:39 UTC

e = an entity representing the target of the opinion.

a = an aspect of e (e.g. a's political perspective, or an action performed by e that has led to the tweet response).

s = the sentiment towards a. a level of tension relating to a, e or both (ranging from e.g. 0 to 3).

h = the holder of the opinion (the twitter user that posted the tweet) .

t = the time it was posted.

Machine Learning: n-Gram

sarah_tiggy21 — 2015-08-19 11:09:52 UTC

In machine learning, naive Bayes classifiers are a family of simple probabilistic classifiers based on applying Bayes' theorem with strong (naive) independence assumptions between the features.

An NB machine classifier supports multiple n-gram classifiers (uni-gram: one word).

Machine Learning: SVM

sarah_tiggy21 — 2015-08-19 11:13:59 UTC

In machine learning, support vector machines (SVMs, also support vector networks^[1]) are supervised learning models with associated learning algorithms that analyze data and recognize patterns, used for classification and regression analysis. Given a set of training examples, each marked for belonging to one of two categories, an SVM training algorithm builds a model that assigns new examples into one category or the other, making it a non-probabilistic binary linear classifier.