This function finds the most similar topics to a given query using semantic similarity analysis. It works with both semantic topic models and traditional STM models by creating topic representations using transformer embeddings and calculating cosine similarity scores.
Usage
find_topic_matches(
topic_model,
query,
top_n = 10,
method = "cosine",
embedding_model = "all-MiniLM-L6-v2",
include_terms = TRUE
)Arguments
- topic_model
A topic model object (semantic topic model or STM model).
- query
A character string representing the query topic.
- top_n
The number of similar topics to return (default: 10).
- method
The similarity method: "cosine", "euclidean", "embedding".
- embedding_model
The embedding model to use for query encoding (default: "all-MiniLM-L6-v2").
- include_terms
Logical, whether to include topic terms in the similarity calculation (default: TRUE).
Examples
if (interactive()) {
mydata <- TextAnalysisR::SpecialEduTech
united_tbl <- TextAnalysisR::unite_cols(
mydata,
listed_vars = c("title", "keyword", "abstract")
)
texts <- united_tbl$united_texts
topic_model <- TextAnalysisR::fit_embedding_topics(
texts = texts,
method = "semantic_style",
n_topics = 8
)
similar_topics <- TextAnalysisR::find_similar_topics(
topic_model = topic_model,
query = "mathematical learning",
top_n = 5
)
print(similar_topics)
}