Documents Webpages Online Reviews

Tokenizing (bag-of-words) Tagging parts-of-speech Filtering Stemming

Sentiment analysis Topic modeling Language style

Designing Content I Flashcards by Marisol Ann Ciovacco

Content marketing is

a form of non-traditional marketing communications whereby a brand produces or designs content in various forms (e.g., text, images, video, audio)
and disseminates that content to targeted audiences and/or customers

How well did you know this?

Not at all

Perfectly

T/F

Content marketing is NOT new

How well did you know this?

Not at all

Perfectly

Content marketing emphasizes the creation of

relevant and valuable online content to drive

business results

How well did you know this?

Not at all

Perfectly

Features of Content Marketing

An integral part of social media marketing, SEO, PR strategies

ROI may be difficult to quantify

Attribution problem under the the multi-channel multi-touch environment
 $, frequency, reach of content, can be easily
measured

How well did you know this?

Not at all

Perfectly

Content features including formats…

…links, promotional nature, topics,

messages, quality of writing, creativity, humor …

How well did you know this?

Not at all

Perfectly

__ __ __ brings unique opportunities to content marketing research

Unstructured big data

How well did you know this?

Not at all

Perfectly

CONTENT IS __ BIG DATA

UNSTRUCTED

How well did you know this?

Not at all

Perfectly

The 4 “V”s of Big Data

Volume
Veracity
Variety
Velocity

How well did you know this?

Not at all

Perfectly

Unstructured data will account for more than __ of the data collected by organizations.

80%

How well did you know this?

Not at all

Perfectly

Unstructured data (or unstructured information) is

information that either does not have a pre-defined data model or is not organized in a pre-defined
manner.

How well did you know this?

Not at all

Perfectly

Unstructured data files often include

text and multimedia content, examples include e-mail messages, word processing documents, videos, photos, audio files, presentations, webpages and many other kinds of documents.

How well did you know this?

Not at all

Perfectly

Unstructured data are not new–previous forms:

 Qualitative research (e.g., in-depth interviews, focus groups, open-ended
questions in consumer surveys)

 Qualitative analysis attempts to reduce the vast amount of verbal or observational
data to a set of well-defined and clearly explained patterns and themes

How well did you know this?

Not at all

Perfectly

unstructured big data opportunities and challenges

 Need tools to extract useful information from unstructured data in a scalable
way

 Find related social media posts (for example,
using Twitter API to extract data)
 Sentiment analysis on text
 Generate word cloud to summarize the text

How well did you know this?

Not at all

Perfectly

Text Mining

 What to mine?

 Online reviews
 Social media posts
 News
 Entertainment products(e.g., movie scripts, books)
 Brand/product descriptions

How well did you know this?

Not at all

Perfectly

Text Mining

 Purpose?

 Sentiment analysis
 Measuring consumer preferences/motivation
 Understanding and assisting the creation of
creative content

How well did you know this?

Not at all

Perfectly

Text Mining Process

Collect Data

Preprocessing

Applying text mining techniques

Analysis of text

Discovery of knowledge

How well did you know this?

Not at all

Perfectly

Collect Data

Documents
Webpages
Online Reviews

How well did you know this?

Not at all

Perfectly

Preprocessing

Tokenizing (bag-of-words)
Tagging parts-of-speech
Filtering
Stemming

How well did you know this?

Not at all

Perfectly

Applying text mining techniques:

Keyword methods: LIWC, SentiWordNet, etc.

* Statistical methods: LDA, SVM, Naïve Bayes, etc.

How well did you know this?

Not at all

Perfectly

Analysis of text

Sentiment analysis
Topic modeling
Language style

How well did you know this?

Not at all

Perfectly

Why focus content marketing on social

media?

 Control (over paid and owned media)
 Cost
 Audience engagement
 Virality
 Feedback (earned media)

How Does Content Drive Business

Results?

Exposure
Influence
Engagement
Action

Schweidel and Moe (2014) demonstrates online venues

(e.g., blogs, forums, social networks, micro-blogs) differ in:

 Extent of social interaction
 Amount of information
 Audience attracted
 Focal product/attributes
 Develop a joint model of brand sentiment and venue format
choice
 Unstructured data: sentiment and topics are manually coded

Results show “significant variation in average sentiment across venue formats, highlighting

the importance of separating the effects of venue format from any sentiment measure”

Tucker (2015) tackles the question

whether “popular” contents are effective contents in driving sales

Crowd sourced persuasiveness measure

“by randomly exposing half of these consumers to a video ad and half to a similar placebo video ad, and then surveying their attitudes towards the focal product”

Video ads designed to be “viral” are in general

less persuasive

relative ad persuasiveness is on average

10% lower for every one million views that the video ads achieve

Exceptions are ads that

generated views and large number of comments, and ads that attracted comments that mentioned the product by name. These ads tend to be perceived as “funny” rather than “outrageous”.

Li and Xie (2020): studies the relationship between

imagery content and social media engagement (likes and shares) using Twitter and Instagram data

Theoretic Framework

Imagery Content Text Content Control variables:

Imagery Content

* Mere Presence * Image characteristics: * Colorfulness * Human face and emotional state * Image Source * Image Quality

Text Content

``` • Sentiment • Topic • Linguistic Content Category • Behavioral Drivers • Number of Hashtags (#), Mentions, (@), Emojis, and Words ```

Control variables:

* Posting Time | * Account Characteristics

Li and Xie (2020) Research Findings

 Processing unstructured data SVM to exact topics and sentiment from text content  Google vision API and manual coding to extract information from imagery content  Manual coding to determine the relevancy between text content and imagery content Modeling: zero-inflated bivariate negative binomial model

How to make people pay attention and engage with your content? What should we do to figure out the type of content (e.g., content features) that make people pay attention and engage with your content

Lab or field experiment

Component of an Experiment

(At least) one independent variable  Independent variable is what experiment manipulates (At least) one dependent variable  Dependent variable is what researcher is interested in explaining (At least) one manipulation  Independent variables are manipulated in some systematic way

(__ __) causes a change in (__ __) and it isn’t possible that (__ __) could cause a change in (__ __).

Independent variable dependent variable dependent variable independent variable

A/B testing

one independent variable with no more than two levels or aspects and one manipulation

Factorial design requires us to

figure out potential interactions between independent variables

Factorial Designs

Allows for manipulation of two or more independent variables at the same time Each variable has two or more levels or aspects Allow to test main effects as well as interactions.

A main effect in Factorial Designs

is the separate influence of each IV on the DV

Interaction occurs (in Factorial Designs) when

the simultaneous effect of two or more IVs is different from the sum of their independent effects.

Using observational data

Start with your own social audit: Go back in the past and look at everything you posted on your various social media accounts in the last 3-6 months Conduct a social audit on your competitors: Look at everything they posted on various social media accounts in the last 3-6 months and ask the same questions

Independent variables (things you control)

 Content features  Posting time  Promoted or not  Posting platform

Dependent variables (feedback received)

```  Views  Likes  Shares  Comments: counts, sentiment, topics  Engagement rates ```

Construct a regression model to

figure out the type of content (e.g., content features) that tend to receive higher engagement  Helps us predict the engagement level of a specific post

Correlation:

measuring the closeness of the relationship or | joint variation between two variables

Regression:

derive an equation that relates the dependent variable to one or more independent variables  Correlation ≠ Causation

Correlation only measures

the nature and degree of association or co-variation between variables.

Estimation method:

Ordinary least squares (OLS). Parameters are estimated from the sample data so that the sum of squared errors is minimized.

Brainscape's Knowledge GenomeTM

Designing Content I Flashcards

Brainscape's Knowledge Genome^TM