Lecture notes (2) Flashcards
Methodology validation
Ensure that the consumer topic is amenable for metaphor analysis
Metaphor dictionary creation
Create and validate a dictionary for each metaphor that is empirically frequent and/or theoretically interesting. Depending on the topic/domain, find keywords (can be automated). Extract context from a sample, code metaphors, and build an automated way to do so of the full sample. Then, validate the automation and adjust if necessary
Metaphor characterisation
Describe the structure of each metaphor and what it does to marketplace sentiments on the consumer topic. Assign meaning to the metaphors, which can be done in several ways
Metaphor-enabled market place sentiment analysis
Detect reach, prevalence, and changes over time of marketplace sentiments regarding the consumer topic by studying its metaphors. Apply the method and related score and identify the importance of the metaphor through frequency or impact over time (dispersion). Brand strategies may change if new metaphors are trending
Conceptual metaphors
Link two domains through a shared relationship
Chatbot (consumer side)
Consumers self-select or intiate use. Goal is often to make consumer life easier
Chatbot (company side)
Consumers are confronted (no choice). Goal is meant to cut cost or create revenue
Anthropomorphising
Attributing human characteristics to non-human entities. Doing this to AI can make people perceive AI more positively; more trustworthy, higher engagement/sentiment, and more friendly
AI receptivity
Refers to perceptions on (future) AI usage
Respondent simulation
There is a tool that can simulate respondents. (1) Create respondent backstory (distribution for age and gender, and other characteristics), (2) Use backstory to answer survey (Responds to experimental conditions), and (3) Results can be saved as a .csv exports
Generative AI
A tool rather than a solution. Can support in time-consuming tasks, but we should remain critical of its output. It is based on public data sources, which can be biased or incorrect, and it usually does not disclose where information comes from. Thus, it can be useful as a starting point, or to generate ideas
Removal rate
Conversion (e.g. C1) removed / total conversion. The higher, the more is removed when you take away a channel
Conv
Means consumer buys something
Null
Means that they don’t buy anything
Total conversion
The total of probabilities of paths that lead to conversion
Weighting the removal rate
Leads to attribution through; removal rate C1 / sum of all removal rates. It makes comparison across removal rates possible, e.g. which channel is most important
Markov model
The goal is to determine which component is most important in the customer journey
Item-based framing
Looks at the similarity between items and makes a recommendation
Positive lift
The words co-occur more often than expected. Lift is larger than 1, so there is a positive association between words
