Social media platforms can provide researchers with massive authentic data. In consumer studies, social media platforms have been recently used for linguistic and cultural data collection to gain insight into consumers' attitudes, responses, and expectations. The use of modern computational tools, such as Large Language Models and Generative Artificial Intelligence (GenAI) can automate the processes of data analysis, thus saving time and effort. The present study aims at investigating into traditional and local food consumption (case studies: snail dishes, and cheeses of the North-Aegean Sea), and at identifying key concepts and the relevant specific words used by Greek consumers. Instagram posts on the cases under study were collected, and using GenAI applications, sentiment analysis was performed on the posts to identify positively related concepts. The hashtags collected revealed patterns in word formation on social media. GenAI applications were used in an attempt to automate analysis tasks. Hashtag coappearances in posts also revealed key concepts, food trends, as well as conceptual networks. The methodology can be transferred to a) specialized lexicography (e.g. to compile domain-specific word lists), b) linguistic and cultural studies (e.g. to study word formation patterns and conceptual networks in linguistic/cultural studies and in comparative studies), and c) terminology (e.g. to identify neologisms and check term usage over time).
Despite high consumer liking scores in laboratory and in-house tests, product failures in the
marketplace are common. Consequently, sensory and consumer studies are increasingly employing
alternative methods to gather data from consumers in real-world settings. One such method
involves leveraging social media platforms. Online social media networks, content communities,
reviews, forums, and blogs offer a vast and rich source of qualitative data, which can be
quantitatively analyzed [
Manual handling is impractical for managing large datasets and cannot serve as a measure of
accuracy. However, it can offer insights into the precision of current Natural Language Processing
(NLP) tools [
Sentiment Analysis, a method initially used in politics and marketing, is also used in food and
consumer studies. It is the computational study of people's opinions, emotions, and attitudes
towards entities, topics etc. [
The objective of this study is to explore Greek consumers' conceptualizations of traditional and local food consumption through a novel methodology incorporating NLP models. The primary goals are: a) to discern how consumers think, feel, and express themselves on social media about traditional and local foods compared to new products with the same basic ingredient, b) to examine how specific aspects of food consumption —such as sensory attributes, geographical characteristics, nutritional value, and environmental concerns— influence consumer choices, and c) to identify pertinent concepts. The developed methodology was tested in case studies involving: a) snails, a traditional Mediterranean food and a sustainable meat alternative, and b) local cheeses from the North-Aegean Sea islands.
Instagram was the social media platform chosen for this study, because Instagram users interact
with companies more often than on other platforms in Greece, and cooking comes second (together
with health/ fitness) among the most common interests of Greek Instagram users [
For data collection, Apify [
Posts from April 2012 (the release of the Instagram platform) to March 2023 were collected. The posts were consolidated into a single file, and duplicates were removed. The data collected for each post included: post ID, type of post (image, sidecar, video), shortCode (shortened URL of the post), caption, hashtags, number of comments, number of likes, timestamp, and whether the post appeared on a professional account. The data were further refined by removing hashtags in languages other than Greek, Greeklish, or English, nonsensical words, parts of speech not relevant to the study (such as pronouns, articles, and prepositions), names of businesses (e.g., snail farming and selling companies, and restaurants), and the original hashtags used for data collection. In all, 1773 posts on snails containing 1844 hashtags, and 18,026 posts on the cheeses under study containing 54,780 hashtags were collected for further analysis.
ChatGPT [
The posts were finally categorized by sentiment (positive, neutral, negative) and by hashtag. The hashtags collected for both case studies were then merged into one file, put into alphabetical order, and duplicates were removed.
Within the hashtags repeated key concepts that pertain to food consumption were identified. These concepts were grouped using English monolingual dictionaries and a name was given to each group as follows: • • • • • •
Preparation: homemade, handmade, easy, simple Origin: traditional food, local food, slow food, Protected Designation of Origin-PDO, agriculture, fusion cuisine, Mediterranean cuisine Quality: healthy food/ diet, fresh, bio/ organic, quality, natural/ real food Content/ Diet: Mediterranean diet, clean eating, low-carb, (high-) protein, vegetarian, vegan, gluten-free Presentation: comfort food, gourmet
Sensory: tasty food, pleasure first
These key concepts were looked up on Google Trends [
Specific prefixes and suffixes involved in word formation were identified, and several word formation patterns were presented in Table 1. These prefixes and suffixes were further investigated using Google Trends to analyze user interest over time (Figure 3), interest by country, related topics, and related queries. -gasm -spotting -yearsstrong -yearsyoung foodstylist, foodinstagram, foodofinstagram foodporn, cheeseporn, pastaporn, burgerporn foodgasm, dessertgasm, beergasm, cretegasm buzzfeed, buzzfeedfood(s), buzzfeast tastespotting 50yearsstrong 50yearsyoung
ChatGPT was consulted to provide historical context, including their origins, trends and movements leading to their emergence, changes in meaning over time, and their representation in dictionaries. Each case pertains to a social media/ internet language trend. For example, regarding the word “buzz” ChatGPT provided the following response (on December 17th, 2024):
3. "buzz" as a Prefix
Modern Usage: The use of "buzz-" as a prefix is a relatively modern development (20th century onwards), and it reflects the word's association with chatter, popularity, and energy.
Common Examples:
Buzzword (first recorded in the 1940s): Refers to a trendy or popular term often used in specific fields or contexts, reflecting the "noise" around a concept.
Buzzkill (popularized in the 1980s–1990s): Describes a person or event that dampens enthusiasm or excitement, implying a disruption of the "buzz."
Buzzfeed (founded in 2006): Combines "buzz" with "feed" to suggest a platform generating and delivering shareable, attention-grabbing content.
Characteristics of Buzz as a Prefix: Often implies excitement, trendiness, or disruption.
Frequently used in contexts involving social or cultural dynamics (e.g., language, media, or technology). 4. "buzz" as a Suffix
Emergence of "-buzz": The use of -buzz as a suffix is less common than its prefixal form, but it is still productive, especially in informal or playful contexts.
Examples: Earbuzz: Used to describe a persistent sound or auditory phenomenon.
Buzzworthy (late 20th century): Denoting something that generates excitement or public interest. Characteristics of Buzz as a Suffix: Suggests association with excitement, noise, or attention.
Often conveys a sense of liveliness or trending appeal. 5. Cultural and Linguistic Influence
Buzz in Marketing and Technology: The metaphorical extension of "buzz" has been embraced by marketing and tech industries, emphasizing excitement and viral appeal (e.g., "buzz marketing" or "Google Buzz," a short-lived social platform).
Buzz in Social Media: In the age of digital communication, "buzz" is often used to describe online chatter, virality, or hype (e.g., "create buzz for a campaign").
6. Summary
The word "buzz" originated as an onomatopoeic term describing a sound, but it evolved to capture ideas of excitement, chatter, and activity. As a prefix, "buzz-" is commonly used to denote trendiness, popularity, or disruption (e.g., "buzzword," "buzzkill"). As a suffix, "-buzz" is less frequent but still conveys associations with attention, liveliness, or auditory phenomena (e.g., "buzzworthy"). Its rise in modern contexts, particularly in media and technology, reflects its ability to evoke the dynamic and fast-moving nature of trends and communication.” ChatGPT’s response was cross-checked as valid using other sources (e.g., dictionaries, social media posts). The same procedure was performed for other food-related words from the list.
Copilot [
These compound innovations often blend multiple words to create new, meaningful hashtags that convey specific ideas or themes.” Copilot’s response was quite satisfactory albeit not thorough as it only provided examples and not an exhaustive list. The same task was assigned to ChatGPT which was able to identify the following word formation patterns: a) blends (it identified hashtags that consisted of more than one word as blends, e.g. mummysrecipe, heresmyfood, aegeancuisine), b) compounds (it also identified hashtags that consisted of more than one word as compounds, e.g. greekcheese, healthyeating, instafood, loveforfood), c) code-switching [hashtags that contained words in different languages, e.g. myκουζινα (English (my) + Greek (kitchen)], d) innovative suffixation or prefixation (hashtags that ended in -lover, -addict, -holic, and-ista, e.g. cheeselover, cheeseaddict, cheesaholic, cheeseanista), and e) use of Greeklish. Further exploring of how to prompt and train AI to successfully identify innovative word formation is required.
Attempts were made to process the list of hashtags using ChatGPT and Copilot, giving the prompt: “I will give you a list of hashtags. I need you to group them into 3 groups: certainly not neologisms, certainly neologisms, and not sure whether neologisms or not. Remember that hashtags are not always one word but can contain two or more words written as one. Some hashtags are in English, some in Greek, and others in Greeklish (meaning Greek transcribed in Latin alphabet based on how they are written or pronounced)”. These attempts were not successful. The AI applications used were unable to successfully process hashtags in Greeklish and hashtags that contained more than one word. As a result, the list of hashtags had to be manually checked for neologisms. In the future, this task needs to become automated.
The methodology for analyzing the hashtag data involves constructing a co-occurrence network
based on hashtags extracted from posts. The hashtag frequencies are calculated to identify the most
used hashtags. A graph is then constructed using NetworkX [
The sentiment analysis of posts reveals whether the words under study are predominantly used in positive, negative, or neutral contexts within a specific domain. For example, the hashtags containing buzz-, -porn, and -gasm exclusively appeared in positive posts. This information is valuable when identifying and studying neologisms.
This study highlights the importance of leveraging advanced methodologies, such as NLP models and sentiment analysis, to gain deeper insights into food consumption trends and language use on social media. By analyzing hashtag co-appearances and sentiment, we can identify key concepts and understand the context in which food-related terms are used. Although current AI applications faced challenges in processing certain hashtags, the manual verification process provided valuable data. Future efforts should focus on automating these tasks to enhance efficiency and accuracy. For this purpose, an open access manually annotated dataset for LLM validation is being prepared. Overall, the integration of linguistic analysis and AI tools offers a promising approach to studying consumer behavior and language trends in the digital age.
Declaration on Generative AI During the preparation of this work, the authors used Chat-GPT-4 in order to: Grammar and spelling check. After using this tool, the authors reviewed and edited the content as needed and take full responsibility for the publication’s content.